cms tracker upgrade thanks to many cms tracker collaborators, past and present, too numerous to...

CMS Tracker Upgrade

Thanks to many CMS Tracker collaborators, past and present, too numerous to acknowledge individually

Tracker web pageshttp://cmsdoc.cern.ch/Tracker/Tracker2005/TKSLHC/index.html

Tracker Upgrade Wiki pageshttps://twiki.cern.ch/twiki/bin/view/CMS/SLHCTrackerWikiHome

http://cmsdoc.cern.ch/Tracker/Tracker2005/TKSLHC/index.html

https://twiki.cern.ch/twiki/bin/view/CMS/SLHCTrackerWikiHome

Geoff Hall ATLAS workshop Nov 2008 2

CMS Compact Muon Solenoid

ECAL

Tracker

HCAL

4T solenoid

Muon chambers

Total weight: 12,500 tOverall diameter: 15 mOverall length 21.6 mMagnetic field 4 T

First operation during 2008

pp collisions at 10 TeV CM energy


Upgrade to CMS

• CMS was designed for 10 years operation at L = 1034 cm-2.s-1

– Max L1 trigger rate 100kHz & decision latency ≈ 3.2µs

• To operate at L = 1035 cm-2.s-1

– most of CMS will survive & perform well with few changes• But expect to upgrade trigger electronics & DAQ

• Notable exception is tracking system– Higher granularity is required to maintain current performance– Greater radiation tolerance, especially sensors

• ASIC electronic technologies will be adequate but

0.25µm CMOS, pioneered by CMS, will probably not be accessible

– L1 trigger using tracker data is essential

Reminder of why this is needed

• Limited statistics – eg:– and time to reduce errors

• However, the environment is very challenging:


1035103510341034

Full LHC luminosity~20 interactions/bx

Proposed SLHC luminosity~300-400 interactions/bx

HZZ ee, MH= 300 GeV vs luminosity

Physics requirements

• Essentially unknown until LHC data make it clear– general guidance as for LHC – granularity, pileup, …– but improving statistics in rare and difficult channels could be vital

• eg: whatever Higgs variant is discovered, more information on its properties than LHC can provide will be needed– accessible by trilinear coupling

• An excellent detector is essential…• …even better than LHC to cope with particle density & pileup

– which should also be flexible to adapt to circumstances


• Expected HH production after all cuts in 4W -> l+/-l+/- + 4j mode– = 0.07-018 fb-1 for mH = 150 – 200 GeV

– with 3000fb-1 ≈ 200 – 600 signal events– plus significant background


TOBTOB

TIDTIDTIBTIB

TECTEC

PDPD

Current Tracker system

• Two main sub-systems: Silicon Strip Tracker and Pixels– pixels quickly removable for beam-pipe bake-out or replacement

Microstrip tracker Pixels

~210 m2 of silicon, 9.3M channels ~1 m2 of silicon, 66M channels

73k APV25s, 38k optical links, 440 FEDs 16k ROCs, 2k olinks, 40 FEDs

27 module types 8 module types

~34kW ~3.6kW (post-rad)

ATLAS workshop Nov 2008 7

Barrel & Forward Pixels July /Aug

Geoff Hall


A better tracker for SLHC?

• Present detector looks to be very powerful instrument• No physics reason to improve spatial and momentum

measurement precision– Key point is to maintain tracking and vertexing performance

• Heavy ion tracking simulations are encouraging:– Track density similar to SLHC– Extra pixel layer would restore losses

• Must optimise layout of tracker for– CPU-effective track finding– Trigger contributions

• Weakest point in present system is amount of material– Electron & photon conversions– Hadronic interactions


Ferenc Sikler

• Heavy ion performance of present tracker is remarkably good– Pixel seeding using 3 layers loses ~10% – but some pp events are more demanding, especially jets

• Granularity of tracker must increase anyway– because of leakage current/noise after irradiation as well as tracking

Material budget

• Reducing power is only way known to improve this– present requirements: pixels ~3.7 kW.m-2, µstrips < 0.1 – 0.4 kW.m-2

• From Tracker perspective, least mass should be at lowest radii



Tracker services

• Major constraint on upgraded system– Complex, congested routes

– Heat load of cables must be removed– Pcable = Rcable(PFE/Vs)2

– Cable voltage drops exceed ASIC supply voltages

• limited tolerance

to voltage excursions

Installation of services was one of the most difficult jobs to complete CMS

It will probably be impossible to replace cables and cooling for SLHC

PFE ≈ 33kW I=15,500A PS = 300kVA


Planning an Upgrade Project

• The SLHC planning assumption– Phase I to 2.3 x 1034 (peak) ~ 2013– Phase II to 1035 from ~2017

• How long is really needed?Developing and building a new Tracker requires ~10 years, eg– 5 years R&D– 2 years Qualification– 3 years Construction– 6 months Installation and Ready for Commissioning

• NB – this is an aggressive schedule– System design and QA important and cannot be downscoped

• we found problems during the later phases which needed time– Cost was a driver for LHC detectors from day one– Construction & Commissioning were major tasks, still ongoing

CERN-AB-2008-065

Working Group organisation

• CMS Tracker R&D structure - active for ~18 months– recently added Task Forces on Layout, Power Options, Trigger

• Objective: develop R&D while Tracker layout is being defined


State of activities

• Sensors– So far, strongly dependent on common R&D efforts – eg RD50

• several R&D projects approved in CMS

– Wide range of sensor designs in common mask set close to submission with HPK

• Engineering, cooling, material,…– Completion of installation has recently released effort– Expect to aim at Phase I pixel replacement as test bed for Phase II

developments• new cooling, power issues, data links• all highly constrained by existing services & YB0

• Readout– simulations in 0.13µm CMOS, studying link options– new readout architecture evolving


Defining a new layout

• Present design suffered from limited simulations– we did not know how many layers would provide robust tracking– pixel system was a late addition, which has an important impact– the material budget estimate was not as accurate as desired

• A new tracker might be “easy” to design based on experience– but provision of trigger information adds a major complication– and the tools to model CMS at L = 1035 were not in place– major uncertainties in power delivery, sensor type, readout architecture

• Hence big emphasis on simulation of new detector


Possible scenario

• The first upgrade will involve the pixel system– simply because the inner layer must be replaced as it degrades

• The design for easy replacement simplifies our task– we can replace when we are ready to do so– and evolve it gracefully into the Phase II system

• How to upgrade:– design pixel system with 4 barrel layers and expanded endcap

• in first step install what we believe is needed, at the right time• maximise benefits to CMS, including smooth transition• fix mechanical envelope for future

– study PT (doublet) layers to contribute to trigger– design outer tracker to match cost, power and performance needs


BPIX Options for 2013 replacement/upgrade

Option

0

1

2

3

4

5

Cooling

C6F14

C6F14

CO2

CO2

CO2

CO2

Readout

analog 40MHz

analog 40MHz

analog 40MHz

analog 40MHz-tw-pairs

digital 320MHz-tw-pairs

digital 640 MHz-tw-pairs

Pixel ROC

PS46 as now

2x buffers

2x buffers

2x buffers

2xbuffer, ADC160MHz serial

2xbuffer, ADC160MHz serial

Layer/Radii

4, 7, 11cm

4, 7, 11cm

4, 7, 11cm

4, 7, 11cm

4, 7, 11cm

4, 7, 11, 16cm

Modules

768

768

768

768

768

1428

Power

as now

as now

as now

as now

as now

DC-DCnew PS

as 2

008

R Horisberger May 2008

Geoff Hall 18ATLAS workshop Nov 2008

Doubling the buffer size

new ROC size

0.4mm

mounting screw whole

• Doubling buffer size just ok in 0.25 CMOS need extra 0.8mm chip periphery!

• No R&D need but verification with high rate test beam

• Design time & verification & testing at high beam rates 8-10 month

L < 2.5x1034cm-2s-1


Barrel Pixel Option 5

4 layers of BPIX with or without ROC data buffers

Build for sure: ~1300 perfect pixel modules 2 new barrel mechanics ( design & fabrication) 4 new supply tube halves ( design, fabrication & testing) ~2000 px-AOH fiber links

new power supplies 2 more pxFED’s

With present power cables, optical readout and cooling pipes absolute impossible ! cables are not there !

4 layer system has 1.8x more modules than present 3 layer system

Unless: • have DC-DC step down converters to bring more power through cables• high speed links to transmit 3.6x more data through same fibers• have advanced bi-phase cooling ( e.g. CO2) in same pipe x-section

R Horisberger May 2008


Outer readout

• Present architecture– analogue, unsparsified, analogue optical links, synchronous– external digitisation, cluster finding, zero suppression– 0.25µm CMOS, FP edge-emitting lasers, single-mode fibres

• Pros– works extremely well and easy to use with excellent diagnostic

capability and noise robustness– occupancy insensitive – few power fluctuations– synchronous system easy to model and understand– cost effective, despite customisation

• Possible cons for future– new optolinks links required – analogue not an option– if analogue information to be preserved, on-detector ADC


22

binary architecture – un-sparsified

digital digital

O/Pdriver

FE amp comp. digital pipeline digitalMUX

off-chip

what about binary un-sparsified?

much simpler than “digital APV”particularly for pipeline and readout side

need fast front end and comparator => more power here

but no ADC power and simpler digital functionality will consume less

allows retention of features we likesimpler synchronous systemno FE timestampingdata volume known, occupancy independent(so no trigger-to-trigger variation)

but less diagnostics (can measure front end pulse shape on every channel in present system

some loss of position resolution, common mode immunity)

binary, un-sparsified is an option we are considering

vth

vth

vth

vth

slow control,bias,

test pulse,……

M RaymondTWEPP 08

ATLAS workshop Nov 2008

23

0.13m preamp/shaper – 2 supply rails only

IPRE

IPSF

ISHA

ISSF

CSENSOR

Cfp

CC

Cfs

Cload

simulated FE amplifier performance

for short strips (CSENSOR ~ 5 pF) choose preamp andshaper input device currents (and Rfs) to achieve50 and 20 nsec CR-RC pulse shapes

0.96

0.94

0.92

0.90

0.88

0.86

volts

3002001000nsec.

50 ns 20 ns

simulated pulse shapes (CSENSOR = 5 pF)

Rfs

peaking time

50 ns 20 ns

IPRE [uA] 40 90

IPSF [uA] 15 15

ISHA [uA] 10 30

ISSF [uA] 35 15

total [uA] 100 150

power [uW] 120 180

noise [e] 800 890

=> for short (~few cm) strips can get quite goodpreamp/shaper noise performance for > factor 5 less than APV (~1 mW) even with only 2 rails

speed

pipe capacitance

0.13 m simulation example

M RaymondTWEPP 08


Simulations status

• First emphasis on developing tools and ensuring they work– significant discoveries along the way– much effort needed, mostly part time– unknown errors, absence of certain expected features (pileup!), bugs,

CPU time, hard-wired detector descriptions,…

• Software team set ambitious goals (next slide)

• Now have two strawmen models for detailed studies– plus layout tool to evaluate quickly basic features of different designs



Goals of the Simulations Group

Perform simulations & performance studies: Must simulateThe physical geometry, including numbers & location of layers, amount of material, “granularity” (e.g. pixels, mini-strips, size and thickness)

The choice of readout, (e.g. technology, speed, latency, numbers of bits)

Types of material, or technology (e.g. scattering, radiation hardness, noise)

Tracking strategy and tracking algorithm

Trigger strategy, trigger technology, and trigger algorithm

Develop a common set of software tools to assist these studies For comparisons between different tracking system strategy/designs

For comparisons with different geometries, and with CMS@LHC

To include sufficient detail for optimization (realistic geometry, etc.)

For comparisons between different tracking trigger strategy/designs

Develop set of common benchmarks for comparisons

Maximize the overlap of these common software tools with those in use for CMS@LHC (assist current efforts where possible)

Get good integration between Tracker and (Tracking) Trigger design


More Realistic Strawman A

A working idea from Carlo and AlessiaTake current Strawman A and remove 1 “TIB” and 2 “TOB” layers

Strawman A r-phi view Strawman A r-phi view

(RecHit ‘radiography’)(RecHit ‘radiography’)

4 inner pixels

2 TIB strixelsAdjust chn count

2 TIB short stripsRemove 1

2 TOB strixelsAdjust chn count

4 TOB short stripsRemove 2


More Realistic Strawman B

Adjust granularity (channel count) of Strawman B layersKeep the TEC for now until someone can work on the endcaps

Strawman B r-phi view Strawman B r-phi view (RecHit ‘radiography’)(RecHit ‘radiography’)

r-z view r-z view


Why tracker input to L1 trigger?

• Single µ and e L1 trigger rates will greatly exceed 100kHz– similar behaviour for jets

• increase latency to 6.4µs but maintain 100kHz for compatibility with existing systems, and depths of memory buffers

Single electron trigger rate

<pT> ≈ few GeV/bx/trigger tower

Isolation criteria alone are insufficient to reduce rate at L= 1035 cm-2.s-1

5kHz @ 1035

L = 2x1033

L = 1034 muon L1 trigger rate

The track-trigger challenge

• Impossible to transfer all data off-detector for decision logic so on-detector data reduction (or selective readout) essential– The hit density means high combinatorial background– Trigger functions must not degrade tracking performance

• What are minimum track-trigger requirements? (My synthesis)– single electron - an inner tracker point validating a projection from the

calorimeter is believed to be needed – single muon - a tracker point in a limited window to select

between ambiguous muon candidates & improve pT

• because of beam constraint, little benefit from point close to beam

– jets – information on proximity/local density of high pT hits should be useful

– separation of primary vertices (ie: 300-400 in ~15cm) – a combination of an inner and outer point would be even better


readoutelectrodes


Possible approaches

• Limited number of proposals so far– try to send reduced data volume from

detector for further logic– eg factor 20 with pT > few GeV/c?

• Use cluster width information to eliminate low pT tracks (F Palla et al)– simple but thinner sensors may limit

• Compare pattern of hits in contiguous sensor elements in closely spaced layers– pT cut set by angle of track in layer

– simple logic but with unrealistically small elements

• can it be applied with coarser pixels? – understanding power & speed issues requires

more complete electronic design

J Jones et al

31

possible PT module for inner layer

2 x 2.5mm

CorrelatorASIC

data

25.6

mm

64 x 2

data

2 layer stacked tracking approach80 mm x 25.6 mm sensors segmented into 2.5 mm x 100 m pixelstiled with readout chips – could be wire bonded for easy prototyping

readout chip ideas (see *)each chip deals with 2 x 128 channel columnsuse cluster width discrimination to reduce data volume

rea

do

ut

chip

12.8mm

12

8 c

ha

nn

els

PT module

x32

*http://indico.cern.ch/getFile.py/access?contribId=15&sessionId=2&resId=1&materialId=slides&confId=36581

80 mm

correlatorcompares hit pattern and address from both layersif match then shift result off-detector

data volumes need to transmit all correlated hit patterns every BXpredicted occupancy + reduction from correlation=> 1 link can serve 2 PT modules

link powerneed ~ 3000 PT modules for 3m length cylinder, r=25cm so 1500 links (@2.56 Gbps) => 3 kW @ 2W / link

readout power50 W / pixel (extrapolate from current pixels)=> 2.4 kW for 8192 x 2 x 3000 channels

=> this will not be a low power layer

G Hall, M PesaresiM Raymond


32

Pt - Trigger for TOB layers

2mm

Strip Read Out Chip2 x 100 pitch withon-chip correlator

Hybrid

50mm strips

1mm

2mm

2 x DC coupled Strip detectorsSS, 100 pitch ~8CHF/cm2

wire bonds

spacer

W.E. / R.H.

track angular resolution ~20mrad good Pt resolution

Two-In-One Designbond stacked upper and lowersensor channels to adjacent

channels on same ASICno interlayer communication

no extra correlation chipjust simple logic on readout chip, looking

at hits (from 2 layers) on adjacent channels

R Horisberger*W Erdmann

32http://indico.cern.ch/getFile.py/access?contribId=3&sessionId=0&resId=0&materialId=0&confId=36580*



Correlation Algorithm

Stub generation

A stub is created when both the row and column difference lie within a given range.

e.g. row offset = 30 ≤ row window ≤ +10 ≤ column window ≤ +1

Upper

Lower

Pass Fail

100μm

100μm


Algorithm Performance

Cuts optimised for high efficiency:

Row window = 2 pixelsColumn window = 3 pixels @ 1mm, 2mm; 4 pixels @ 3mm; 6 pixels @ 4mm

pT discriminating performance of a stacked layer at r=25cm for various sensor separations using 10,000 di-muon events with smearing

Sensor separation is again an effective cut on pt – as with the stacked strips

Again, the width of the transition region increases with separation. Due to:

- pixel pitch- sensor thickness- charge sharing- track impact point

Efficiencies decrease with sensor separation due to the larger column window cuts – sensor acceptances and fake containment are issues

Future power estimates

• Some extrapolations assuming 0.13µm CMOS– Pixels 58µW -> 35µW/pix

• NB sensor leakage will be significant contribution

– Outer Tracker: 3600 µW -> 700µW/chan• Front end 500µW (M Raymond studies)• Links 170µW (including 20% for control)

– PT layers: 300µW/chan - most uncertain• Front end 50µW (generous extrapolation from pixels)• Links 100µW (including 20% for control)• Digital logic 150µW (remaining from 300µW)• 100µm x 2.5mm double layer at R ≈ 25cm => 11kW

• More detailed studies needed– sensor contribution not yet carefully evaluated– internal power distribution will be a significant overhead


Power delivery

• Perhaps the most crucial question– although estimates of power are still imprecise, overall requirements

can be estimated– we must reduce sensor power with thin sensors

• finer granularity should allow adequate noise performance

– and attempt to limit channel count to minimum compatible with tracking requirements (simulations!)

• total readout power expected to be ~25-35kW – in same range as present system so larger currents required

• Radical solutions required– serial powering or DC-DC conversion– Task Force to answer soon if we should adopt a baseline solution


Conclusions

• CMS is trying systematically to develop a new Tracker design– using simulations to define new layout

• We are very satisfied with the prospects for the present detector– but would like to reduce the material budget– and achieve similar performance

• The largest challenges are– power delivery and distribution– provision of triggering data– the schedule

– but many developments of sensors, readout, cooling,… are also expected


BACKUPS


Collimation phase 2

Linac4 + IR upgrade phase 1

New injectors + IR upgrade

phase 2

Early operation


Tracker related R&D Projects

41

Proposal title Contact Date Status

Letter of intent for Research and Development for CMS tracker in SLHC era R Demina 14.9.06 Approved

Study of suitability of magnetic Czochralski silicon for the SLHC CMS strip tracker P Luukka, J Härkönen, R Demina, L Spiegel

31.10.07 Approved

R&D on Novel Powering Schemes for the SLHC CMS Tracker L Feld 3.10.07 Approved

Proposal for possible replacement of Inner Pixel Layers with aims for an SLHC upgrade A Bean 31.10.07 Approved

R&D in preparation for an upgrade of CMS for the Super-LHC by UK groupsWP1: Simulation studies/ WP2: Readout development/ WP3: Trigger developments G Hall 31.10.07 Approved

The Versatile Link Common Project F Vasey, J Troska 11.07 Approved

3D detectors for inner pixel layers D Bortoletto, S Kwan 12.07 Received

Proposal for US CMS Pixel Mechanics R&D at Purdue and Fermilab in FY08 D Bortoletto, S Kwan 12.07 Received

R&D for Thin Single-Sided Sensors with HPK M Mannelli 7.2.08 Received

An R&D project to develop materials, technologies and simulations for silicon sensor modules at intermediate to large radii of a new CMS tracker for SLHC F Hartmann, D Eckstein 6.3.08 Approved

Development of pixel and micro-strip sensors on radiation tolerant substrates for the tracker upgrade at SLHC M de Palma 9.4.08 Received

Power distribution studies S Kwan 15.6.08 Approved

Cooling R&D for the Upgraded Tracker D Abbaneo 21.07.08 Approved

First Level Trigger based on Tracking F Palla 4.11.08 Received

ATLAS workshop Nov 2008Geoff Hall


Calorimeter Algorithms

e/photon

tau jet

• Electron/photon– Large deposition of energy

in small region, well separated from neighbour

• tau jet– Isolated narrow energy

deposition– simulations identify likely

patterns to accept or veto

Jet triggers

• The rates for electromagnetic triggers as a function of threshold for various types of event (left: jet events, middle: H->4e, right H->) at a luminosity of 2x1033 cm-2.s-1[Physics TDR Vol 1]. Scaling to 1035 cm-2.s-1, it is clear that typical single isolated electron rates would be far higher than tolerable for practical thresholds.


Electron triggers

• Rejection vs efficiency for electron identification obtained from the HLT pixel matching and low (left) and high (right) luminosity [Physics TDR Vol 1]. The upper curve shows the performance with the existing pixel detector.


45

estimated link power contribution

LHC unsparsified analog 0.36 Gb/s (effective)

2 / analog fibre

60 mW 230 W

SLHC digital APV no sparsification 2.5 Gb/s 32 / GBT ~ 2W 490 W

SLHC digital APV with sparsification 2.5 Gb/s 256 / GBT ~ 2W 60 W

SLHC binary unsparsified 2.5 Gb/s 128 / GBT ~ 2W 120 W

linkspeed

# of 128 chan.chips/link

powerper link

link power/sensor chan.

LHC unsparsified analog230 W / sensor channel: ~ 10% of overall channel budgetneed to do better at SLHC (e.g. 10% of 0.5 mW = 50 W)

SLHC digital APV without sparsification not viablelink power contribution too high (no. of channels will increase at SLHC)

SLHC digital APV with sparsification appears bestbut can only be achieved with extra buffering between FE chips and link

more chips to develop, some additional power

SLHC binary unsparsified next besthas strong system advantages

no. of chips / link depends on estimations of data volume – some details in backup slides


cms tracker upgrade thanks to many cms tracker collaborators, past and present, too numerous to...

Documents