P Baudrenghien

CERN BE-RF

LLRF 2013 Lake Tahoe Oct 2013

LLLRF FOR THE LHC CRAB CAVITIES

Many thanks to R Calaga for help material and comments

Content

bull Why Crab Cavities in the LHCbull Cavity prototypesbull Proposed layout in the LHC tunnelbull LLRF challengesbull Operational scenariobull Planningbull Conclusions

Oct 1st 2013 LLRF13 2

WHY CC IN THE LHC

Oct 1st 2013 LLRF13 3

LHC performances

Oct 1st 2013 LLRF13

After LS1 (2015-2020) we want 60 fb-1yr Thanks to the reduced bunch spacing (25 ns -gt up to 2760 bunches vs 1320) increased energy (65 TeV) and reduced transverse emittances (injectors upgrade) We keep b in the 04 ndash 055 m range

~56 fb-1 totalIncreasing nbr Bunches during the year

~23 fb-1 totalReduced b and increasing bunch intensity during the yearAnnouncement

of the Higgs Boson identification

Courtesy BE-OP

4

High Lumi LHCbull HiLumi-LHC (2020-2030) -gt

upgrade to 250-300 fb-1yrbull The crossing angle is required to

limit the effect of long-range interactions on both sides of the IP We have 32 interactions per IP Experiments have shown that significant losses arise when the separation in the common region is below 10 s

Oct 1st 2013 LLRF13

gt10 s

reduced b

Increased crossing angle

R Calaga

5

LLRF13

bull The crossing angle reduces the luminosity The reduction factor F is easily computed from the half crossing angle f the bunch length sz and the transverse size in the crossing plane s

Geometric Luminosity FactorOct 1st 2013

2

1

12

z

F

2012HiLumi

2011

R Calaga

6

Crab cavitiesbull Luminosity can be recovered by applying a rotation of the bunches

around its center to align the colliding bunches in the centre of the detector This is achieved with an RF dipole cavity (crab cavity) with the RF phase aligned with bunch centre

bull On the other side of the IP at ~ half a betatron period distance from the initial crabbing an opposite kick is applied (uncrabbing) so that the bunch rotation is limited to the IP Else a much increased machine aperture would be required

Oct 1st 2013 LLRF13

R Calaga

7

CAVITIES

Oct 1st 2013 LLRF13 8

Parametersbull 400 MHz (fundamental) and 800 MHz cavities were originally

considered But the length of the LHC bunch (4 s = 13 ns) favored the 400 MHz design

bull We need 10 MV crabbing kick -gt 3 cavities per IPside (CMS and ATLAS) Total 12 cavities per ring

bull The two LHC beams are separated by 194 mm only The apertures in the IP are 84 mm The cavity outer radius must thus be below 150 mm so that it does not block the non-interacting beam pipe

Oct 1st 2013 LLRF13 9

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Content

bull Why Crab Cavities in the LHCbull Cavity prototypesbull Proposed layout in the LHC tunnelbull LLRF challengesbull Operational scenariobull Planningbull Conclusions

Oct 1st 2013 LLRF13 2

WHY CC IN THE LHC

Oct 1st 2013 LLRF13 3

LHC performances

Oct 1st 2013 LLRF13

After LS1 (2015-2020) we want 60 fb-1yr Thanks to the reduced bunch spacing (25 ns -gt up to 2760 bunches vs 1320) increased energy (65 TeV) and reduced transverse emittances (injectors upgrade) We keep b in the 04 ndash 055 m range

~56 fb-1 totalIncreasing nbr Bunches during the year

~23 fb-1 totalReduced b and increasing bunch intensity during the yearAnnouncement

of the Higgs Boson identification

Courtesy BE-OP

4

High Lumi LHCbull HiLumi-LHC (2020-2030) -gt

upgrade to 250-300 fb-1yrbull The crossing angle is required to

limit the effect of long-range interactions on both sides of the IP We have 32 interactions per IP Experiments have shown that significant losses arise when the separation in the common region is below 10 s

Oct 1st 2013 LLRF13

gt10 s

reduced b

Increased crossing angle

R Calaga

5

LLRF13

bull The crossing angle reduces the luminosity The reduction factor F is easily computed from the half crossing angle f the bunch length sz and the transverse size in the crossing plane s

Geometric Luminosity FactorOct 1st 2013

2

1

12

z

F

2012HiLumi

2011

R Calaga

6

Crab cavitiesbull Luminosity can be recovered by applying a rotation of the bunches

around its center to align the colliding bunches in the centre of the detector This is achieved with an RF dipole cavity (crab cavity) with the RF phase aligned with bunch centre

bull On the other side of the IP at ~ half a betatron period distance from the initial crabbing an opposite kick is applied (uncrabbing) so that the bunch rotation is limited to the IP Else a much increased machine aperture would be required

Oct 1st 2013 LLRF13

R Calaga

7

CAVITIES

Oct 1st 2013 LLRF13 8

Parametersbull 400 MHz (fundamental) and 800 MHz cavities were originally

considered But the length of the LHC bunch (4 s = 13 ns) favored the 400 MHz design

bull We need 10 MV crabbing kick -gt 3 cavities per IPside (CMS and ATLAS) Total 12 cavities per ring

bull The two LHC beams are separated by 194 mm only The apertures in the IP are 84 mm The cavity outer radius must thus be below 150 mm so that it does not block the non-interacting beam pipe

Oct 1st 2013 LLRF13 9

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

WHY CC IN THE LHC

Oct 1st 2013 LLRF13 3

LHC performances

Oct 1st 2013 LLRF13

After LS1 (2015-2020) we want 60 fb-1yr Thanks to the reduced bunch spacing (25 ns -gt up to 2760 bunches vs 1320) increased energy (65 TeV) and reduced transverse emittances (injectors upgrade) We keep b in the 04 ndash 055 m range

~56 fb-1 totalIncreasing nbr Bunches during the year

~23 fb-1 totalReduced b and increasing bunch intensity during the yearAnnouncement

of the Higgs Boson identification

Courtesy BE-OP

4

High Lumi LHCbull HiLumi-LHC (2020-2030) -gt

upgrade to 250-300 fb-1yrbull The crossing angle is required to

limit the effect of long-range interactions on both sides of the IP We have 32 interactions per IP Experiments have shown that significant losses arise when the separation in the common region is below 10 s

Oct 1st 2013 LLRF13

gt10 s

reduced b

Increased crossing angle

R Calaga

5

LLRF13

bull The crossing angle reduces the luminosity The reduction factor F is easily computed from the half crossing angle f the bunch length sz and the transverse size in the crossing plane s

Geometric Luminosity FactorOct 1st 2013

2

1

12

z

F

2012HiLumi

2011

R Calaga

6

Crab cavitiesbull Luminosity can be recovered by applying a rotation of the bunches

around its center to align the colliding bunches in the centre of the detector This is achieved with an RF dipole cavity (crab cavity) with the RF phase aligned with bunch centre

bull On the other side of the IP at ~ half a betatron period distance from the initial crabbing an opposite kick is applied (uncrabbing) so that the bunch rotation is limited to the IP Else a much increased machine aperture would be required

Oct 1st 2013 LLRF13

R Calaga

7

CAVITIES

Oct 1st 2013 LLRF13 8

Parametersbull 400 MHz (fundamental) and 800 MHz cavities were originally

considered But the length of the LHC bunch (4 s = 13 ns) favored the 400 MHz design

bull We need 10 MV crabbing kick -gt 3 cavities per IPside (CMS and ATLAS) Total 12 cavities per ring

bull The two LHC beams are separated by 194 mm only The apertures in the IP are 84 mm The cavity outer radius must thus be below 150 mm so that it does not block the non-interacting beam pipe

Oct 1st 2013 LLRF13 9

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

LHC performances

Oct 1st 2013 LLRF13

After LS1 (2015-2020) we want 60 fb-1yr Thanks to the reduced bunch spacing (25 ns -gt up to 2760 bunches vs 1320) increased energy (65 TeV) and reduced transverse emittances (injectors upgrade) We keep b in the 04 ndash 055 m range

~56 fb-1 totalIncreasing nbr Bunches during the year

~23 fb-1 totalReduced b and increasing bunch intensity during the yearAnnouncement

of the Higgs Boson identification

Courtesy BE-OP

4

High Lumi LHCbull HiLumi-LHC (2020-2030) -gt

upgrade to 250-300 fb-1yrbull The crossing angle is required to

limit the effect of long-range interactions on both sides of the IP We have 32 interactions per IP Experiments have shown that significant losses arise when the separation in the common region is below 10 s

Oct 1st 2013 LLRF13

gt10 s

reduced b

Increased crossing angle

R Calaga

5

LLRF13

bull The crossing angle reduces the luminosity The reduction factor F is easily computed from the half crossing angle f the bunch length sz and the transverse size in the crossing plane s

Geometric Luminosity FactorOct 1st 2013

2

1

12

z

F

2012HiLumi

2011

R Calaga

6

Crab cavitiesbull Luminosity can be recovered by applying a rotation of the bunches

around its center to align the colliding bunches in the centre of the detector This is achieved with an RF dipole cavity (crab cavity) with the RF phase aligned with bunch centre

bull On the other side of the IP at ~ half a betatron period distance from the initial crabbing an opposite kick is applied (uncrabbing) so that the bunch rotation is limited to the IP Else a much increased machine aperture would be required

Oct 1st 2013 LLRF13

R Calaga

7

CAVITIES

Oct 1st 2013 LLRF13 8

Parametersbull 400 MHz (fundamental) and 800 MHz cavities were originally

considered But the length of the LHC bunch (4 s = 13 ns) favored the 400 MHz design

bull We need 10 MV crabbing kick -gt 3 cavities per IPside (CMS and ATLAS) Total 12 cavities per ring

bull The two LHC beams are separated by 194 mm only The apertures in the IP are 84 mm The cavity outer radius must thus be below 150 mm so that it does not block the non-interacting beam pipe

Oct 1st 2013 LLRF13 9

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

High Lumi LHCbull HiLumi-LHC (2020-2030) -gt

upgrade to 250-300 fb-1yrbull The crossing angle is required to

limit the effect of long-range interactions on both sides of the IP We have 32 interactions per IP Experiments have shown that significant losses arise when the separation in the common region is below 10 s

Oct 1st 2013 LLRF13

gt10 s

reduced b

Increased crossing angle

R Calaga

5

LLRF13

bull The crossing angle reduces the luminosity The reduction factor F is easily computed from the half crossing angle f the bunch length sz and the transverse size in the crossing plane s

Geometric Luminosity FactorOct 1st 2013

2

1

12

z

F

2012HiLumi

2011

R Calaga

6

Crab cavitiesbull Luminosity can be recovered by applying a rotation of the bunches

around its center to align the colliding bunches in the centre of the detector This is achieved with an RF dipole cavity (crab cavity) with the RF phase aligned with bunch centre

bull On the other side of the IP at ~ half a betatron period distance from the initial crabbing an opposite kick is applied (uncrabbing) so that the bunch rotation is limited to the IP Else a much increased machine aperture would be required

Oct 1st 2013 LLRF13

R Calaga

7

CAVITIES

Oct 1st 2013 LLRF13 8

Parametersbull 400 MHz (fundamental) and 800 MHz cavities were originally

considered But the length of the LHC bunch (4 s = 13 ns) favored the 400 MHz design

bull We need 10 MV crabbing kick -gt 3 cavities per IPside (CMS and ATLAS) Total 12 cavities per ring

bull The two LHC beams are separated by 194 mm only The apertures in the IP are 84 mm The cavity outer radius must thus be below 150 mm so that it does not block the non-interacting beam pipe

Oct 1st 2013 LLRF13 9

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

LLRF13

bull The crossing angle reduces the luminosity The reduction factor F is easily computed from the half crossing angle f the bunch length sz and the transverse size in the crossing plane s

Geometric Luminosity FactorOct 1st 2013

2

1

12

z

F

2012HiLumi

2011

R Calaga

6

Crab cavitiesbull Luminosity can be recovered by applying a rotation of the bunches

around its center to align the colliding bunches in the centre of the detector This is achieved with an RF dipole cavity (crab cavity) with the RF phase aligned with bunch centre

bull On the other side of the IP at ~ half a betatron period distance from the initial crabbing an opposite kick is applied (uncrabbing) so that the bunch rotation is limited to the IP Else a much increased machine aperture would be required

Oct 1st 2013 LLRF13

R Calaga

7

CAVITIES

Oct 1st 2013 LLRF13 8

Parametersbull 400 MHz (fundamental) and 800 MHz cavities were originally

considered But the length of the LHC bunch (4 s = 13 ns) favored the 400 MHz design

bull We need 10 MV crabbing kick -gt 3 cavities per IPside (CMS and ATLAS) Total 12 cavities per ring

bull The two LHC beams are separated by 194 mm only The apertures in the IP are 84 mm The cavity outer radius must thus be below 150 mm so that it does not block the non-interacting beam pipe

Oct 1st 2013 LLRF13 9

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Crab cavitiesbull Luminosity can be recovered by applying a rotation of the bunches

around its center to align the colliding bunches in the centre of the detector This is achieved with an RF dipole cavity (crab cavity) with the RF phase aligned with bunch centre

bull On the other side of the IP at ~ half a betatron period distance from the initial crabbing an opposite kick is applied (uncrabbing) so that the bunch rotation is limited to the IP Else a much increased machine aperture would be required

Oct 1st 2013 LLRF13

R Calaga

7

CAVITIES

Oct 1st 2013 LLRF13 8

Parametersbull 400 MHz (fundamental) and 800 MHz cavities were originally

considered But the length of the LHC bunch (4 s = 13 ns) favored the 400 MHz design

bull We need 10 MV crabbing kick -gt 3 cavities per IPside (CMS and ATLAS) Total 12 cavities per ring

bull The two LHC beams are separated by 194 mm only The apertures in the IP are 84 mm The cavity outer radius must thus be below 150 mm so that it does not block the non-interacting beam pipe

Oct 1st 2013 LLRF13 9

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

CAVITIES

Oct 1st 2013 LLRF13 8

Parametersbull 400 MHz (fundamental) and 800 MHz cavities were originally

considered But the length of the LHC bunch (4 s = 13 ns) favored the 400 MHz design

bull We need 10 MV crabbing kick -gt 3 cavities per IPside (CMS and ATLAS) Total 12 cavities per ring

bull The two LHC beams are separated by 194 mm only The apertures in the IP are 84 mm The cavity outer radius must thus be below 150 mm so that it does not block the non-interacting beam pipe

Oct 1st 2013 LLRF13 9

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Parametersbull 400 MHz (fundamental) and 800 MHz cavities were originally

considered But the length of the LHC bunch (4 s = 13 ns) favored the 400 MHz design

bull We need 10 MV crabbing kick -gt 3 cavities per IPside (CMS and ATLAS) Total 12 cavities per ring

bull The two LHC beams are separated by 194 mm only The apertures in the IP are 84 mm The cavity outer radius must thus be below 150 mm so that it does not block the non-interacting beam pipe

Oct 1st 2013 LLRF13 9

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

PrototypesOct 1st 2013 LLRF13

Cockcroft 4R Jlab ndash SLAC Dipole Cavity

BNL DoubleQWR

10

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

PROPOSED LAYOUT

Oct 1st 2013 LLRF13 11

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Oct 1st 2013 LLRF13

LHC layout Main (ACS)RF system

Crab CavityRF

Crab CavityRF

12

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

TX and LLRF in new service galleries

Oct 1st 2013 LLRF13

bull We have ~300 m distance between crabbing and un-crabbing cavities

bull We propose to dig-out a short gallery running parallel to the tunnel on both sides of IP1 and IP5 for the RF power (TX) and LLRF electronics

bull Such a gallery exists in IP4 (ex LEP klystron gallery)

13

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

LLRF CHALLENGES

Oct 1st 2013 LLRF13 14

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Challengesbull Fine-phasing of the CC kick with the individual bunch centrebull Reduce the cavity impedance at the fundamental to prevent

transverse Coupled-Bunch instabilitiesbull Prevent transverse emittance blow-up caused by RF noisebull Manage beam loss following a klystron trip or cavity quench

Oct 1st 2013 LLRF13 15

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

16

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Phasing the CC with the bunch centre

bull A phase error (wrt bunch centre) causes an offset of the bunch rotation axis This results in a transverse offset Dx at the IP

bull 1 deg RF phase (7 ps) leads to Dx = 05 mm (110th transverse beam size) Worryinghellip

bull We will use phase compensated links from SR4 (main RF) to IP1 and IP5

bull The LLRF will use PU signal to fine-adjust the CC phase for each bunch

bull This is strictly required with the phase modulation scheme foreseen for the accelerating cavities

Oct 1st 2013 LLRF13

500 1000 1500 2000 2500 3000 3500bunch

40

30

20

10

10

20

tim e ps

Proposed ACS phase modulation in physics The phase slip is 60 ps pk-pk along the turn

See talk by TMastoridis

RFRF

cx

17

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Machine Impedance bull The HighLumi LHC will accelerate 11 A DC current per beam (compared

to 035 A DC in 2012)bull The Crab Cavities will introduce a series of Narrow-Band resonators in the

machine (fundamental plus HOMs)bull Control of the fundamental is the responsibility of the LLRF bull At the fundamental one cavity presents a transverse impedance around

25 GWm

to be reduced below 05 MWm (threshold calculated for physics conditions assuming 12 cavities and a 60 ms damping time from the active transverse damper)

Oct 1st 2013 LLRF13

LRR QQc

18

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

RF feedbackbull With accelerating cavities in high beam current machines the problem

of (in)stability caused by the cavity impedance at the fundamental is now routinely cured by active feedback The amplifier driven by a feedback system feeds a current into the cavity which just cancels the beam current

bull The cavity impedance is effectively reduced by the feedback gainbull The limitation comes from the unavoidable delay in the loop Above

some gain level the delay will drive the feedback into electrical oscillations (not related to the beam)

Oct 1st 2013 LLRF13 19

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

bull With strong RF feedback the minimal effective transverse impedance scales with the loop delay T

bull Assume 300 W RQ and 1ms loop delay we get

Rmin = 63 MWm per cavity

bull The impedance is reduced by 400 linear bit still way too much compared to the instability threshold

bull Fortunately for a resonator whose BW covers several revolution frequency lines what counts is not just the impedance but the growth rate

Similarly for the Crab CavitiesOct 1st 2013 LLRF13

0min 0

RR TQc

0

0

1 2

L

L

R QQZ

c iQ

20

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Instability growth rate

Oct 1st 2013 LLRF13

bull The growth rate is the difference between real impedance on the two plusmn(l wrev+wb) sidebands of the wRF

bull The growth rate is much reduced if the effective impedance is symmetric around wRF

Real part of the cavity impedance with RF feedbackbull Left situation in physics with CC on tunebull Right situation with CC detuned by -100 kHz

Very important reduction to be expected in physics

Smaller reduction when the cavity is detuned but impedance budget larger (factor 2)

0

0

2

Re Re2

l l RF rev b RF rev bb rev

l RF rev b RF rev bb rev

c q Ij Z l Z l

E T

c q IZ l Z l

E T

21

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Transverse Betatron Comb filter

bull One-turn delay filter with gain on the betatron bands

bull To keep 10 dB gain margin the gain on the betatron lines is limited to ~ 6 linear (16 dB)

bull 2 N Poles on the betatron frequenciesbull Reduction of noise PSD where the beam

respondsbull Reduction of the effective cavity impedance

thereby improving transverse stability

bull Zeros on the revolution frequency linesbull No power wasted in transient beam loading

compensation with off centered beam

Oct 1st 2013 LLRF13

Betatron comb filter response with a=3132 and non-integer Q=03 Observe the high gain and zero phase shift at (n plusmn03) frev

NQiNQi

NN

zeazea

zzzH

22 11

1)(

We can further reduce the growth rate by selectively correcting the effective impedance on the betatron lines

22

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Betatron Comb filter

Oct 1st 2013

bull The filter is identical to the double peak comb filter used at PEPII on the accelerating cavities for the reduction of the longitudinal impedance at the fundamental frequency It had resonances on the synchrotron side-bands (Qs=005)

Frequency response of the PEPII double peak comb filterF Pedersen RF Cavity Feedback CERN-PS-92-59-RF

LLRF13 23

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Cavity RF Noise

Oct 1st 2013 LLRF13

RF feedback noise sources The RF reference noise nref

The demodulator noise (measurement noise) nmeas

The TX (driver) noise ndr (process noise) It includes also the LLRF noise not related to the demodulator

The Beam Loading Ib Dx

We get

0

1 1

sZ

cav set ref meas b drs s RQLQ

K G e Z s Z sV V n n x I n

K G e Z s K G e Z s

Closed Loop response CL(s) Equal to ~1 in the CL BW Increase of K increases the BW Within the BW reference noise and

measurement noise are reproduced in the cavity field

Beam Loading response = effective cavity impedance Zeff(s)

Equal to ~1KG in the CL BW Increase of K decreases Zeff within the CL BW Within the CL BW TX noise and beam loading

are reduced by the Open Loop gain KG

0

1 2

with

L

L

R QQZ s

sc Q

s j

Main coupler

24

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

RF phase noise without beam ACS

bull SSB phase noise Power Spectral Density in dBcHz RF sum of the 8 cavities beam 1 No beam but physics conditions (12 MV) The grey trace corresponds to injection conditions (6 MV) Out of loop measurement

Oct 1st 2013 LLRF13

50 Hz and harmonics come from the klystron HV ripples The klystron loop reduces them by 50 dB up to 600Hz Fs crosses 50 Hz during the ramp [55 Hz ndash 26 Hz] ndr noise

In the 10kHz-400 kHz range the noise level is defined by the demodulation of the antenna signal in the RFfdbk and OTFB loops Flat between -135 and -140 dBcHz nmeas noise

The dips at multiple of Frev are the action of the OTFB that increases regulation gain and decreases noise level

In the transverse plane the first betatron band is at 3 kHz Reference noise is not an issue The performances will be defined by TX noise and measurement noise

Noise in the 10Hz-1kHz range is caused by the VCXO limited Q (-20 dBdecade) nref noise

s

radfS

fL 210

)(

in102 )(

Hz

dBcfL in)(

25

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Oct 1st 2013 LLRF13

bull The beam will sample the noise spectrum in all betatron side-bands bull The integrated effect of the measurement noise increases with the

closed loop BW that is proportional to the feedback gainbull The effect of the TX noise is inversely proportional to the feedback gainbull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hz

then summing the noise PSD from DC to + 300 kHz over all betatron bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull A ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise that is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

bull But the CC TX (50 kW tetrode) will be less noisy than the ACS klystronbull And an additional reduction (16 dB) will come from the betatron comb

Scaling the ACS to crab

26

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Trade-off

bull A weak coupling (large QL)bull Reduces the effect of TX and LLRF noise Goodbull But we want to keep BW sufficient so that microphonics are located

within the cavity BW

bull A large feedback gainbull Reduces the instability growth rates Goodbull Limits the effect of TX and LLRF noise Goodbull Increases the integrated effect of measurement noise by increasing

the BW Bad

bull Trade-off requiredbull Will depend on the TX noise

Oct 1st 2013 LLRF13 27

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

RF Power vs QL

Oct 1st 2013 LLRF13

TX linked to the cavity via a circulator 3 MV RF RQ = 300 W 11 A DC current 1 ns 4s bunch length (18 A RF component of beam current) Cavity on tunebull Yellow trace beam centeredbull Blue trace x=+1 mm offset bull Red trace x=-1 mm offset

2

(Panofsky Wenzel)

1

2 22L

L

zx

x RFg

dVi ep

dx

V IRP Q xQ cR QQ

For 1 mm offset the range of acceptable QL (P lt 30 kW) is

bull 25 105lt QL lt18 106 for RQ=300 W

bull 8 104lt QL lt6 105 for RQ=900 W

Low values preferred for tuning (microphonics)

Candidate TX 50-100 kW tetrode as used in the SPS 3522 MHz system in the 1990rsquos

28

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

KEEPING CRABBING AND UNCRABBING KICKS EQUALImproves precision and compensates for a single-cavity fault

Oct 1st 2013 LLRF13 29

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Coupled feedbackbull Following a CC quench or a TX trip the crabbing will not be limited to the IP

region but will continue around the ringbull The tails of the bunch transverse distribution will be collimated out resulting in

huge losses and the beam dump will be triggered It takes up to 3 turns to reactbull The LLRF can help minimizing the losses during that transient The idea is to

implement a coupled feedback acting on all CC cavities and keeping crabbing and un-crabbing equal

bull In the case of a crabbing cavity trip it would ramp the un-crabbing voltage down tracking the total crabbing voltage

Oct 1st 2013 LLRF13 30

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

The Crab Cavity systems (IP1 and IP5)Oct 1st 2013 LLRF13

A slower loop ( 5 ms response time) combines the antenna signals f rom the 3 crabbing and 3 un-crabbing cavities and acts on the individual set-points to minimize the non-closure of crabbing gymnastics

Phase PU

MIMO feedback6-In 6-Out spider

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

IP1 or IP5One beam

Phase PU

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

TX

Ant

Crab

Cavity Controller

The 400 MHz RF and Frev Bx references received on Fiber Optic links f rom SR4

For each cavity a f ast local loop (lt 1 ms response time) keeps the voltage at the desired set-point f or each bunch A local Phase PU would improve bunch per bunch accuracy

300 m

Voltage change following a klystron trip is defined by QL Voltage variations following a quench is not well known To be studiedexperimented (SPS test-bench)

31

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Operational scenario

Oct 1st 2013 LLRF13 32

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Operational scenario (revisited)bull Strong RF feedback and tune controls are ON at all timebull During filling ramping or operation with transparent crab cavities we detune

the cavity but keep a small field requested for the active Tuning system As the crabbing kick is provided by three cavities we can use counter-phasing to make the total field invisible to the beam The RF feedback is used with the cavity detuned to provide stability and keep the Beam Induced Voltage zero if the beam is off-centered We can use the demanded TX power as a measurement of beam loading to guide the beam centering

bull ON flat topbull Reduce the detuning while keeping counter-phasing so that the total

voltage is zero The RF feedback keeps the cavity impedance small (beam stability) and compensates for the beam loading as each cavity moves to resonance

bull Once the cavity detuning has been reduced to zero we drive counter-phasing to zero Any luminosity leveling scheme is possible by synchronously changing the voltage in all crab cavities as desired

Oct 1st 2013 LLRF13 33

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Planningbull 2013-2014 Cavity Testingbull 2016- SPS tests 2 cavities in one cryostatbull 2015-2017 (LS2) Prototype Cryomodulebull 2018-2020 (LS3) Productionbull In operation after LS3

Oct 1st 2013 LLRF13

SPS testbench with 2 Crab cavities in the LSS4 The cryomodule can be moved in and out of the beam line (2016-)

Proposed SPS cryomodule with 2 Crab Cavities The cryomodule is designed to accept all three CC makes

Courtesy EN-MME

34

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

CONCLUSIONS

Oct 1st 2013 LLRF13 35

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

bull All requirements appear very demanding but not unrealisticbull We have possible LLRF solutions to be tested in the SPS test-bench

This will give indications on the phase error impedance and RF noisebull We will also observe the behavior of the RF field following a quench

and validate the coupled-cavity feedback idea

Oct 1st 2013 LLRF13

Conclusions

36

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

THANK YOU FOR YOUR ATTENTION

Oct 1st 2013 LLRF13 37

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

BACK-UP SLIDES

Oct 1st 2013 LLRF13 38

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

State-space model

bull We start with the simplest modelbull Cavities on-tune represented as first order linear difference

equation (LPF)bull TX represented by a constant gain

bull In matrix formbull We want to design a linear regulator (matrix K) that is

a proportional feedback generating corrections on I(n) from measurement of V(n)

bull With the regulator the state equations become

bull The feedback changes the state-transition matrix bull The choice of the K matrix coefficients can be done

using the Linear Quadratic Regulator (LQR) theory

Oct 1st 2013 LLRF13

TXZ(s)

Cavity 1

Cavity 2

I1

TXZ(s)

I2

2-Input 2-Output Regulator

V1

V2

2221

1211

2

1

2

1

0

0

0

0

with

1

kk

kkK

g

gb

a

aA

ni

ninI

nv

nvnV

nIbnVAnV

nignvanv

nignvanv

222

111

1

1

nVKnI

nIbnVbKAnV 1

39

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Linear Quadratic Regulator LQRbull Assume that the state is displaced at time zero (non-zero initial cavity

voltage) and observe the transient while the system is brought back to the zero state

bull Let us define a cost function that is a quadratic function of the state variables (V) and the regulation input (I)

bull With the matrix Q and R we give different weight to the state error (Q) and the needed regulation power (R)

bull We will use a diagonal R matrix so that the second term is proportional to the klystron power needed in the restoring transient

bull With the Q matrix elements we can give more or less coupling between the two feedbacks With a diagonal matrix the feedbacks are fully decoupled

bull By balancing q0 and q1 in the matrix below we give more importance to the voltage difference

Oct 1st 2013 LLRF13

22112

202

10

101

110

nvnvqnvqnvqnVQnV

qqq

qqqQ

t

0n

tt nIRnInVQnVJ

40

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Oct 1st 2013 LLRF13

10

01Q

01100100

10001100Q

Cavity voltage following a unit step of both klystrons at time zero and a half-unit step reduction of klystron 2 alone at time 50When observing one loop at the time the regulation with independent klystrons is better It is 3 times faster and the static error is three times smallerWhen the transient is on one klystron only (kly2 red) there is no compensation on cavity 1 for the independent feedbacks resulting in a static error of ~2 With the coupled feedbacks kly1 (blue) reacts to the drop in cavity 2 voltage and the final voltage difference is ~ 05

Independent feedbacks

Strongly coupled feedbacks

Cavity 1Cavity2

Diagonal A matrix

Large off-diagonal values -gt strong coupling

Kly1 ignores Cav2 voltage drop

Kly1 tries to track Cav2 voltage drop

Study ongoinghellip Commission in the SPS

41

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

RF phase modulationbull In physics

bull We will accept the modulation of the cavity phase by the beam current (transient beam loading) and adapt the voltage set point for each bunch accordingly

bull The klystron drive is kept constant over one turn (amplitude and phase)

bull The cavity is detuned so that the klystron current is aligned with the average cavity voltage

bull Needed klystron power becomes independent of the beam current For QL=60k we need 105 kW only for 12 MV total

bull Stability is not modified we keep the strong RFfdbk and OTFB

bull The resulting displacement of the luminous region is acceptable

bull During filling bull It is desirable to keep the cavity phase constant for

clean capture Thanks to the reduced total voltage (6 MV) the present scheme can be kept with ultimate

Oct 1st 2013 LLRF13

Modulation of the cavity phase by the transient beam loading in physics 2835 bunches 17 1011 pbunch 15 MVcavity QL=60k full detuning (-78 kHz)

42

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

EMITTANCE GROWTH MEASUREMENT

Oct 1st 2013 LLRF13 43

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

bull The beam will sample the noise in all betatron side-bands

bull The integrated effect of the measurement noise increases with the closed loop BW that is proportional to the feedback gain

Oct 1st 2013 LLRF13

bull Assume a SSB phase noise of L= -135 dBcHz or S= 63E-14 rad2Hzbull Now summing the noise PSD from DC to + 300 kHz over all betatron

bands we get 30011 x 2 x 63E-14 rad2Hz =34E-12 rad2Hz from DC to the revolution frequency

bull Conclusion a ldquocopyrdquo of the LHC ACS design (300 kW klystron ) would generate 37E-8 rad2 white noise

bull That is 2E-4 rad rms or 11E-2 deg rms phase noise 400 MHz

SSB phase noise in an ACS cavity with varying feedback gains Measured and calculated [3]

Scaling the ACS to crab

44

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

bull How can we measure the effect of RF noise if it is not dominant in the emittance growth

bull We faced a similar problem in the LHC as longitudinal emittance growth driven by IBS is much stronger than the growth caused by RF noise

bull Measurements were done with ions at 35 Z TeV in Nov 2010 with four equi-distant bunches per ring

Oct 1st 2013 LLRF13 45

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

bull Phase noise was injected in one RF cavity with a bandwidth of 10 Hz centered on the first revolution band frevplusmnfs

bull The power of the injected noise was varied during the test

bull Bunch length was monitoredbull With large noise power the

effect became dominant allowing for a good calibration of emittance growth (slope of bunch lengthening) vs RF noise power

Oct 1st 2013 LLRF13

Will be done on the SPS test-bench

frev

)(4

22

fSdt

ds

46

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

With an RF feedback the minimal effective impedance Rmin and closed-loop single-sided bandwith Dw scale with loop delay T

For the LHC we have T=650 ns

Rmin = 75 kWcavity 06 MW total

Dw2p = 320 kHz

Strong RF feedback The LHC ACS

Oct 1st 2013 LLRF13

TQ

RR 0min

T

31=ωΔ

Measured Closed Loop response with the RF feedback QL=60000 without feedback (~7 kHz 2-sided BW) With feedback we get 700 kHz BW The effective impedance is reduced by ~ 35 The LHC cavities are equipped with movable couplers and QL can be varied from 10000 to 100000 But with feedback Qeff ~600 in all positions

The loop delay T was kept low in the LHC

With the RF feedback we have reduce the ACS cavity impedance at resonance by ~35 linear (QL=60k) The 216 MW are reduced to 06 MW

47

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

HARDWARE

Oct 1st 2013 LLRF13 48

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

For each Crab Cavity we have a Cavity Controller includingbull An RF Feedback Loop for noise and beam loading controlbull A TX Polar Loop to reduce the TX noise and stabilize its gainphase shiftbull A Tuner Loop to shift the cavity to a detuned position during filling and ramping Then smoothly bring the

cavity on-tune with beam for physicsbull A field Set Point for precise control of the cavity field

Tuner Processor

Dir Coupler

Fwd

Rev

DA

C

Ic fwd

Ic rev

TUNER LOOP

CAVITY LOOPS

TX

Circ

Ig fwd

TX Polar Loop (not needed)

Feed-forward

Tuner Control

Ic fwd

CONDITIONING DDS

SWITCH amp LIMIT

SWITCH

Ana

log IQ

Mo

dulator

IQ Rotator ampGain Control

LO

Var G

ain RF

A

mpifier

DDS AM Chopper

Main Coupler Vacuum

FAST LIMIT

RF Drive permitted

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Fwd

Ant

CRAB CAVITY

Voltage functions

I0Q0

DIGITAL IQ DEMOD

DIGITAL IQ DEMOD

Ant

SUM

Version 20111111

AFF for Beam Loading

compensation

SinCos CORDIC

Gain amp Phase

IC revFrom Tuner

Loop

Gain Set

IC rev

Crab Cavity Servo Controller Simplified Block Diagram

Technology DSP

CPLD or FPGA

Analog RF

SignalsDigital

Analog baseband

Digital IQ pair

Analog IQ pair

RF 3522 MHz

Ant (from paired cavities)

DIGITAL IQ DEMOD

PU

Digital RF feedback

with Cavity Coupling

and Betatron

comb

Oct 1st 2013 LLRF13

Field control fully integrated with the rest of the LHC by using standard FGCs (slides 22-23)

Interconnection with the paired cavity(ies) Coupled feedback

Measurement of bunch phase modulation

Gain increased on the betatron bands

49

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Oct 1st 2013 LLRF13

SSB Modulator

IF (I amp Q) asymp25MHz

Dual TxDac 16 bits

RF Demodulator

RFampLO mixing

IF asymp25MHz

14 bits ADC

Fs=4IF asymp100MHz

4x

LO Distribution

2 x Duplex Optical Serial Links

2 in amp 2 out

2Gbitss

(le32Gbitss)

SRAM 2x8 Mbyte for diagnostics

VME P1 backplane for

slow controlsreado

ut

Dedicated backplane (P2)

bull Power Supplybull Clocksbull Interlocksbull hellip

Xilinx Virtex 5 SX

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

50

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Oct 1st 2013 LLRF13

2 x Duplex Optical Serial links

2 inputs

2 outputs

Status LED

RF output

RF inputs (4x)

LO input

G Hagmann BE-RF-FBdesigner

SPS 800 MHz TWC prototype

51

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Oct 1st 2013 LLRF13

ACS LLRF 1 rack 2 VME crates per cavity in a Faraday Cage in the UX45 cavern

52

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Oct 1st 2013 LLRF13

ACS RF in the UX45 cavern

Faraday cages with LLRF electronics

Klystrons

Waveguides to cavities

Concrete shielding around the beam line

53

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

The accelerating (ACS) system (IP4)Oct 1st 2013 LLRF13

Faraday Cages

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Kly

Ant

to SUM

Cavity Controller

Cav

Phase PU

SUM

Fiber (400 M

Hz and F

rev ref)

Tunnel IP4

Beam 1

UX45 cavern

Kly

Ant

Cav

Kly

Ant

Cav

Kly

Ant

Cav

Phase PUBeam 2

to SUM

Cavity Controller

to SUM

Cavity Controller

to SUM

Cavity Controller

SUM

RF Synchronization Beam Control beam 2Beam Control beam 1

Fiber (400 M

Hz and F

rev ref)

Fib

ers

to S

PS

Surface building SR4

cabl

e

cabl

e

cable

cabl

e

Distance ~ 500 m

The 400 MHz and Frev B1 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

One Beam Control system per ring I t averages over all bunches I t updates once per turn I t generates a fixed amplitude RF reference signal that tracks the

momentum ramp I t uses signals f rom a phase PU to minimize the eff ect of RF noise

The 400 MHz and Frev B2 references Are sent to Crab Cavities I P1 and I P5 on Fiber Optic links

For each cavity a f ast local loop keeps the voltage at the desired set-point f or each bunch

54

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Longitudinal One-Turn delay feedback (OTFB)bull The One-Turn delay Feedback (OTFB) produces

gain only around the revolution frequency harmonicsbull It further reduces the transient beam loading and

effective cavity impedance (factor of 10)

Oct 1st 2013 LLRF13

Effective Cavity Impedance with RF feedback alone (smooth trace) and with the addition of the OTFB (comb) The cavity centre frequency is 400789 MHz We look at a band offset by +200 kHz to +300 kHz Frev= 11 KHz The OTFB provides ~ 20 dB additional impedance reduction on the Frev lines

With the One-Turn delay feedback on the LHC ACS cavities we have gained another factor 10 in impedance reduction on the revolution sidebands resulting in a 350-fold reduction (QL=60k) The 216 MW are reduced to 006 M W

55

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Beam Control plus Cavity Controller

bull Two major noise sourcesbull The RF reference noise from the Beam Control introduced during the

modulationdemodulation process in the Cavity Controller This noise is coherently injected in all eight cavities It will be the reference 400 MHz for CC as well

bull The noise injected in the Cavity Controller electronics and the Klystron noise This noise is uncorrelated from cavity to cavity

Oct 1st 2013 LLRF13 56

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation

Why 10 sigma separation

Oct 1st 2013 LLRF13

Footprints calculated for nominal and upgradeparameters with 10σ and 12σ beam separation with headoncollisions at 2 IPs and 16 or more long-range interactionsPer IP

57

• LLLRF for the LHC Crab CAVITies
• Content
• Why CC in the LHC
• LHC performances
• High Lumi LHC
• Geometric Luminosity Factor
• Crab cavities
• cavities
• Parameters
• Prototypes
• Proposed layout
• LHC layout
• TX and LLRF in new service galleries
• LLRF challenges
• Challenges
• The Crab Cavity systems (IP1 and IP5)
• Phasing the CC with the bunch centre
• Machine Impedance
• RF feedback
• Similarly for the Crab Cavities
• Instability growth rate
• Transverse Betatron Comb filter
• Betatron Comb filter
• Cavity RF Noise
• RF phase noise without beam ACS
• Scaling the ACS to crab
• Trade-off
• RF Power vs QL
• Keeping crabbing and uncrabbing kicks equal
• Coupled feedback
• The Crab Cavity systems (IP1 and IP5) (2)
• Operational scenario
• Operational scenario (revisited)
• Planning
• conclusions
• Conclusions
• Thank you for your attention
• Back-up slides
• State-space model
• Linear Quadratic Regulator LQR
• Slide 41
• RF phase modulation
• emittance growth Measurement
• Scaling the ACS to crab (2)
• Slide 45
• Slide 46
• Strong RF feedback The LHC ACS
• HARDWARE
• Slide 49
• Slide 50
• Slide 51
• Slide 52
• Slide 53
• The accelerating (ACS) system (IP4)
• Longitudinal One-Turn delay feedback (OTFB)
• Beam Control plus Cavity Controller
• Why 10 sigma separation