the agipd system for the european xfel. · 2018. 1. 31. · heinz graafsma | page 6 av. rate: 27khz...
TRANSCRIPT
The AGIPD system for the European XFEL.
Heinz Graafsma;
for the AGIPD consortium
DESY-Hamburg; Germany
&
University of Mid-Sweden
Heinz Graafsma | Page 2
Particle / X-ray Signal Charge Electr. Amplifier Readout Digital Data
Pixelated
Particle
Sensor
Amplifier &
Readout Chip
CMOS
Indium Solder
Bumpbonds Data Outputs
Power
Clock Inputs
Connection
wire pads
Power
Inputs
Outputs
Particle / X-ray
Qsignal
Hybrid Pixel Array Detectors (HPADs)
Heinz Graafsma | Page 3
Hybrid Pixel Array Detectors (HPADs)
Particle / X-ray
Qsignal
Pixelated
Particle
Sensor
Amplifier &
Readout Chip
CMOS
Connection
wire pads
Power
Inputs
Outputs
Heinz Graafsma | Page 6
av. Rate: 27kHz XFEL
120Hz LCLS
60Hz SCSS
600ms
99.4 ms
100 ms 100 ms
220 ns
FEL process
X-ray photons
<100 fs
Electron bunch trains; up to 2700 bunches in 600 msec, repeated 10 times per second. Producing 100 fsec X-ray pulses (up to 27 000 bunches per second).
27 000 bunches/s with
4.5 MHz repitition rate
European XFEL: Time Structure Challenge
Heinz Graafsma | Page 7
K. J. Gaffney and H. N. Chapman, Science 8 June 2007
Single shot imaging
Heinz Graafsma | Page 8
4.5 MHz
What are the challenges ?
Heinz Graafsma | Page 9
How to meet the challenge ?
Three dedicated Projects:
• Depmos Sensor with Signal Compression
Non-linear gain, digital storage
• Adaptive Gain Integrating Pixel Detector
Automatic adaptive gain, analogue storage
• Large Pixel Detector
Three parallel gains, analogue storage
The Adaptive Gain Integrating Pixel Detector
(AGIPD)
Heinz Graafsma | Page 11
C1
Leakage comp.
C2
C3
Discr.
Contr
ol lo
gic
Trim DAC
Vthr VADCmax
Analo
gue
encodin
g
Normal Charge
sensitive amplifier
High dynamic range:
Dynamically gain switching system
Extremely fast readout (200ns):
Analogue pipeline storage
0,0
0,2
0,4
0,6
0,8
1,0
1,2
1,4
1,6
1,8
0 5000 10000 15000Number of 12.4 KeV - Photons
Ou
tpu
t V
olta
ge
[V
]
Cf=100fF Cf=1500fF Cf=4800fF
Adaptive Gain principle
Heinz Graafsma | Page 12
Sensor Electronics per pixel
ASIC
periphery Chip output driver
Mux
HV
+ -
THR
DAC SW CTRL
Analog Mem
Analog Mem
CDS
RO Amp
Calibration circuitry
Adaptive gain amplifier
352 analog memory cells
… …
Read O
ut bus
Pixel matrix
AGIPD readout principle
Heinz Graafsma | Page 13
AGIPD 1.0 Pixel Electronics
• 200 x 200 micron2 pixels • 352 storage cells + veto
possibilities. • Minumum signal ~ 300 e- =
0.1 photon of 12.4keV • Maximum signal ~ 33 x 106 e- =
104 photons of 12.4keV • 4.5 MHz frame rate • 64 x 64 pixels per ASIC • 2 x 8 ASICs per module
(128x512 pixels, no dead area) • 4 modules per quadrant • Full-sized Chip under test since
~ 1 year.
Heinz Graafsma | Page 14
ASIC: characterization
HIGH gain range (x12.4 keV): 50
MED gain range
(x12.4 keV): 1200
LOW gain range
(x12.4 keV):
5000 (+5000)
Vth,comp
Heinz Graafsma | Page 15
ASIC: characterization
ENC = 265e- RMS
5 % Shot-to-Shot
fluctuation
Heinz Graafsma | Page 16
Beam direction
(coming from sample) AGIPD system
Chip tester box
AGIPD at the P10 Beamline
Heinz Graafsma | Page 17
Gain switching experimentally proven
104 photons / pulse
Single photon sensitivity
4.5 MHz frame rate
photons/second/pixel 10 10 10 10 10 10
Looking at the direct beam
Heinz Graafsma | Page 18
Scientific quality data obtained
• Complete system proven to work
• Calibration proven to be adequate
Maxipix
AGIPD
photons/second/pixel
10 10 10 10 10 10
Small angle scattering experiments
Heinz Graafsma | Page 19
Digital part:
FPGA daughter board
on carrier board
Detector head:
512x128 pixel sensor
bump bonded to
8x2 ASICs
Analogue part:
64 ADC
channels on a two
board system
Vacuum board with a flexible
connection
Detector structure
Si-Sensor with
(200μm)2 pixels
ASIC with charge
integrating readout and
analogue frame storage
behind each pixel
Data transfer on one of 4
10G compatible standard
ethernet links via
TCP/UDP
AGIPD 1M consists of 16 detector modules:
Heinz Graafsma | Page 20
The Real thing
Heinz Graafsma | Page 21
Single bunch imaging – a challenge to find processes fast enough
Experimental setup
● Drilled equidistant holes into a DVD
● DVD covered with zinc paint to
increase absorption
● Mounted DVD on a fast electric motor
● Measurement of hole repetition frequency
with diode and oscilloscope:
1.208kHz
Experiments: AGIPD module @APS
Heinz Graafsma | Page 22
Calculation for burst imaging
Vdisc, AGIPD =
29.51m/s
Vdisc, Laser =
29.83m/s
• APS bunch spacing: t = 154ns
• Number of pixels crossed during
burst of 352 images: ~ 8
• Pixel size: 200µm
Result from laser measurement
Single bunch imaging is possible even at a repetition rate of 6.5MHz!!
Experiments: AGIPD module @APS
Heinz Graafsma | Page 23
Mechanical concept (without vacuum chamber)
Semi-rigid PCBs allow 2-dim. move
Heinz Graafsma | Page 24
The motion stage
Collision switches End switches
Tasks
The accuracy of the positioning
of the motion stage and its
collision- and end switches
should be better than 10% of
the pixel size (<20µm).
The motion stage should run
stable over „many“ repetitions.
The impact of vacuum forces on
the accuracy and the outgasing
behaviour of the stage should
be tested.
Heinz Graafsma | Page 25
Experiment and DAQ
Dummy weight of 26Kg for
simulation of one quadrant
Dial gauge resolution is 0.1µm
with an uncertainty of 2σ =
0.39µm
Encoder uncertainty = 0.2µm
Python scripts for
automatisation of repetition
experiments.
Readout of:
dial gauge position
Encoder position
Motor position
Thermal effect not critical!!
Heinz Graafsma | Page 26
Results
End switches
N = 75 𝒙 = 𝟏. 𝟒µ𝒎 ± 𝟎. 𝟐µ𝒎
%Pix = 0.7%
N = 75 𝒙 = 𝟏. 𝟖µ𝒎± 𝟎. 𝟒µ𝒎
%Pix = 0.9%
Ambient tests
N < 15,000 𝝈 < ±𝟐. 𝟑µ𝒎
%Pix = 1.2%
Vacuum tests
N < 2,000 𝝈 < ±𝟐. 𝟎µ𝒎
%Pix = 1.0%
𝑃0=1.0 ∙ 10−6𝑚𝑏𝑎𝑟
𝑃𝑜𝑝=3.0 ∙ 10−6𝑚𝑏𝑎𝑟
No significant impact
of vacuum force
Hanging weights/
4 Quadrant test
N < 1,000 𝝈 < ±𝟏. 𝟗µ𝒎
%Pix = 1.0%
Motion stage proved to be stable over more
than 25,000 repetitions with mounted weights
Heinz Graafsma | Page 27
Planning
• Delivery 1st AGIPD 1M (old ASICs): summer 2015
• Resubmission AGIPD ASIC: summer 2015
• Upgrade 1st AGIPD 1M (new ASICs): beginning 2016
• Delivery 2nd AGIPD 1M (MID): mid 2016
AGIPD 2.0 options
Heinz Graafsma | Page 29
High-Z pixel detectors
> Aim: Increase efficiency at 50 keV by factor of 10
Replace silicon sensor in LAMBDA with high-Z semiconductor
Combine high QE with hard X-rays, high frame rate, high signal-to-noise
> Investigating different materials in collaboration with other institutes
and industry
Cadmium telluride
Gallium arsenide
Germanium
Heinz Graafsma | Page 30
Material Status Raw image quality Other pros/cons
Cadmium
telluride
Established
technology
Highest QE >
50keV
Time/dose varying
inhomogeneity
Gallium
arsenide
New
production
method, quick
progress
Flatfield correction
greatly improves
images
Single source
(Russia)
Germanium Newer
development,
still need to
engineer
system for
beamline
High uniformity
Cooling to -100 oC
required
Heinz Graafsma | Page 31
GaAs sensor tests at P02
> GaAs detector produced through
German-Russian partnership
768 x 512 layout, 55um pixels
80% efficiency at 42 keV
> Measurements of standards at
P02.2 show useful diffraction
patterns at 1000 Hz frame rate
GaAs LAMBDA Maximum =70 photons
Heinz Graafsma | Page 32
WP-75 AGIPD 15th DAC Report The European XFEL management would appreciate if the AGIPD collaboration would start to investigate the potential of an AGIPD technology based detector with smaller pixels (approx. 50 μm) and prepare an Expression of Interest (EOI), if the collaboration is interested in such a development.
Heinz Graafsma | Page 33
Three possible options identified
3 possible options identified
•IBM 130nm 200um pitch
•XFEL speed (4.5MHz) •>300e [265e] •352 images
•IBM 130nm tech. based on AGIPD
•proven tech, architecture, rad-tolerance •minimal re-develop. time needed •100um pitch option •XFEL speed (4.5MHz) •noise comparable to AGIPD1.0
•TSMC 65nm tech. novel architecture
•new tech & architecture, rad-tol. to be evaluated •re-development time: 4-to-6 years •50-100um pitch options •XFEL speed (4.5MHz) •noise comparable to AGIPD1.0
•UMC 110nm tech. based on Jungfrau •proven architecture, but rad-tol. to be evaluated •~1 year on top of std. Jungfrau time needed •75um pitch • < than XFEL speed (1-2MHz) •noise lower than AGIPD1.0 (120e~180e)
issue common to all 3 solutions: pixel area reduction memory depth reduction
Heinz Graafsma | Page 34
Summary Options
Summary
techn
olo
gy
pixel p
itch
[um
]
com
patib
le w
ithE-X
FEL fram
e rate?
estim.
mem
. dep
th
estim. n
oise
est. dyn
range
[12
ke ph
]
com
patib
le w
ith p
resent
back-en
d?
ASIC
d
evelop
men
t
advan
tages
Risks
IBM 130nm
100um
yes 4.5MHz
64-80 analog
265e~ 375e
5-10k yes ~1 year 1MPW + 1 Engineering run + contingency
tested technology & circuit, extensive info on tech rad-response
power consumption possibly re-simulation/re-layout of some components needed
UMC 110nm
75 um
no 1~2MHz
32-48 analog
120e~ 180e
10k no ~1 year on top of standard Jungfrau
Low cost tech., std. Jungfrau prototypes avail.
rad-hard response to be tested module geometrical dimension different storage cell calibrations
TSMC 65nm
100um
yes 4.5MHz
300-400 (digital)
200e~ 300e
~7k no 3-6years 3-4MPW + 1 Engineering run + contingency
digital output possibly integrated w adaptive gain (4gains+ low-res ADC)
new design, new technology, digitization, readout, limited info on tech rad-response
" (analog cells)
50 um
proba-bly yes
26-56 200e~ 300e
1.5k no 3-6years 3-4MPW + 1 Engineering run + contingency
digital output new tech, new design, power consumption, limited info on tech rad-response
Heinz Graafsma | Page 35
Summary
• AGIPD 1.0 working according to specifications; few
fixes will be implemented in AGIPD 1.1 (end 2015)
• First AGIPD 1M (SPB) scheduled to be delivered end
2015
• Second AGIPD 1M (MID) scheduled to be delivered
mid 2016
• Version with smaller pixels possible.
• AGIPD consortium is looking forward to early science
experiments.