Embed Size (px)
Citation preview
Heinz Graafsma ([email protected])
DESY-Hamburg; Germany
University of Mid Sweden
X-ray, Electron and Ion-imager
developments at DESY-Hamburg
Page 2
Layout
>Storage Rings and Free-Electron Lasers
>Detectors for Storage Rings
>Free-Electron Laser Challenge and Solutions
>New Sources and Detectors
>The Data Challenge
Page 3
Storage Rings (SR)
and
Free-Electron Lasers (FEL)
Page 4
Photon Sources in Hamburg
FLASH
PETRA III
EuropeanXFEL
Page 5
Storage Rings
brill
ianc
e
PETRA-3
Second generation
First generation
X-ray tubes
ESRF (2000)
ESRF (1994)
Free-Electron lasers
RotatingAnodes
Third generation
1900 1960 1980 2000
1900 1920 1940 1960 1980 2000
ESRF (1994)
2ème generation
1ère génération
Tubes àrayons X
Années
ESRF (futur)
Limite de diffraction
ESRF (2000)3èmegeneration
Lasers àélectrons libres
Rayonnementsynchrotron
1020
1018
1016
1014
1012
1010
108
102110221023
1019
1017
1015
1013
1011
109
107
106
Brillance(photons/s/mm 2/mrad2/0.1%B.F.)
X-ray Source development: a continuing success story
Page 6
Photon Science: from fundamental to applied
Study of extremely charged ions Structure of viruses
Authentication of paintings
Page 7
Photon Science: many different techniques
ERSFGrenoble, France
Page 8
Storage Rings and FELs: two very different sources
Both SRs and FELs use undulators
to generate the radiation
> Storage Rings (circular accelerator; recirculating):
§equilibrium source ==> large electron bunches / low electron density
• è electrons emit photons incoherently: I a ne è low intensity per pulse§many stored electron bunches è quasi continuous source (nsec)
> Free-Electron Lasers (linear accelerator; single pass):
§non equilibrium source è small electron bunches / high electron density
• è electrons emit photons coherently: I a ne2 è high intensity per pulse§ few electron bunches è pulsed source (10’s msec – 100’s nsec)
Page 9
Detectors
for
Storage Rings
Hybrid Pixel Array Detector (HPAD)
Diode Detection Layer• Fully depleted, high resistivity
• Direct x-ray conversion• Si, GaAs, CdTe, etc.
Connecting Bumps• Solder or indium
• 1 per pixel
CMOS Layer• Signal processing
• Signal storage & output
Gives enormous flexibility!
X-rays
Page 11
Storage Ring: Photon Counting Detectors
> Used the Medipix-3 readout chip as a basis for system development: Large Area Medipix Based Detector Array, LAMBDA
> Modular system è very versatile; single chip to multi mega-pixels
> Successfully deployed at PETRA III @ DESY
> Large interest from other Synchrotron Sources== > create spinoff: X-Spectrum GmbH
courtesy of X-Spectrum GmbH
LAMBDA 10M for SAXS in-vacuum for electrons and ions
Page 12
DESY development of a Timepix4 based system
Timepix3 (2013) Timepix4 (2021)Technology IBM 130 nm – 8 metal TSMC 65 nm – 10 metalPixel size 55 x 55 µm 55 x 55 µmPixel arrangement 3-side buttable
256 x 2564-side buttable (TSV)
512 x 448Sensitive area 1.98 cm2 6.94 cm2
Rea
dout
mod
es
Data driven(tracking)
Mode ToT and TOAEvent packet 48-bit 64-bitMax rate < 43 Mhits/cm2/s 357.6 Mhits/cm2/sPix rate equiv. 1.3 kHz/pix average 10.8 kHz/pix average
FrameBased(imaging)
Mode Count: 10 bit + iToT Count: 8 or 16 bit CRWFrame Zero suppressed (with pix addr) Full frame (no pix addr)Max countrate
82 Ghits/cm2/s ~ 800 Ghits/cm2/s
Max frame rate
N/A (worst case: 0.8ms readout) 80 kHz CRW
TOT energy resolution < 2 keV < 1 keVTime resolution 1.56 ns ~ 200 psReadout bandwidth ≤ 5.12 Gbps (8 x 640 Mbps) ≤163.8 Gbps (16 x 10.2 Gbps)
Target minimum threshold < 500 e- < 500 e-
3.5 x TPX3
8 x TPX3
10 x TPX3
2 x TPX38 x TPX3
32 x TPX3
40 x LAMBDA
Page 13
Ion detection (Controlled Molecule Imaging, Küpper)
Electric field
Molecule in prepared state
Detector
Thanks to Melby Johny, Sebastian Trippel, Hubertus Bromberger, Jochen Küpper
Page 14
Ion detection (Controlled Molecule Imaging, Kupper)
Electric field
Laser destroys molecule
Detector
Trajectory
Time of flight (m/q ratio)
Thanks to Melby Johny, Sebastian Trippel, Hubertus Bromberger, Jochen Küpper
Page 15
Coulomb explosion imaging of Pyrrole
~2 ns (std dev) time resolution
Page 16
Correlation experiments – XPCS, XCCA
> Measure per-pixel autocorrelation of X-ray signal
> Event rate at PETRA-III: ~2e7 hits/s in Maxipix detector
> 2e9 hits/s at PETRA-IV should be possible with multi-chip Timepix4
Colloidsample
Coherent X-rays
Speckle pattern on detector
10 ns timescale
Proof-of-principle test with Timepix3 at P10(most pixels masked to reduce data rate)
10-6 10-5 10-4 10-3 10-2 10-1 100 1010.98
1
1.02
1.04
1.06
1.08
1.1
1.12
1.14
Taus [s]
G2
func
tion
Thanks to Fabian Westermeier, Michael Sprung
Page 17
TimePix-4 System Development
> First iteration of chip already produced by CERN§ Pixel-level performance confirmed§ Some bugs with high-speed readout –
revised version coming in Spring 2021 (Michael any update?)
> Currently developing single-chip system§ Readout based on FPGA evaluation
board§ Should be available for first experiments
late 2021
Page 18
Free-Electron Lasers
Page 19
The Free-Electron Laser revolution
• Fast science 100 fsec
• “Single shot” science
• High Rep-rate science
• Completely new science
x109
Page 20
Injector at DESY campus
Linear Accelerator1.9 km17.5 GeV
Undulator/Photon Tunnels
Experiment Hallin Schenefeld
Inauguration 1. September 2017
The European XFEL (a small addition to the ILC)
Page 21
The European XFEL: a Burst-Mode machine
Page 22
4.5 MHz
Thedetectorchallenge
Page 23
The Adaptive Gain Integrating Pixel DetectorDESY, PSI, University Hamburg and University Bonn
Page 24
The Adaptive Gain Integrating Pixel Detector
C1
C2
C3
Discr.
Con
trol
logi
c
TrimDAC
Vthr @ VADCmax
Ana
logu
e en
codi
ng
Normal Charge sensitive amplifier
High dynamic range:Dynamically gain switching system
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
Out
put V
olta
ge [V
]
Cf=100fF Cf=1500fF Cf=4800fF
Sensor
Page 25
Looking at the direct beam at PETRA
photons/secondLogarithmic color scale
Gain switching experimentally provenØ 104 photons / pulseØ Single photon sensitivity
220 nsec rep-rate è in-pixel storage of images
Page 26
AGIPD: Analog Storage Memory
Sensor Electronics per pixel
ASIC periphery
Chipoutputdriver
Mux
HV
+-
THR
DACSWCTRL
Analog Mem
Analog Mem
CDS
RO Amp
Calibration circuitry
Adaptive gain amplifier
352 analog memory cells
……R
O b
us (p
er c
olum
n)
Pixel matrix
Page 27
AGIPD Pixel Electronics: > 80% frame storage area
• 200x200micron2 pixels• 352storage cells +vetopossibilities.
• Minumum signal ~300e- =0.1photon of12.4keV
• Maximumsignal ~33106 e- =104 photons of12.4keV
• 4.5MHzframe rate• 64x64pixels perASIC• 2x8ASICspermodule (128x512pixels,no dead area)
• 4modules perquadrant
200
mic
ron
Page 28
AGIPD 1Mpix Systems: SPB and MID Beamlines at European XFEL
Independently movable
quadrants
External electronics
Image plane out of vessel
Modules of 2x8 chips
Arbitrary hole in
size & shape
• SPB & MID User operation
Page 29
In first run solved lysozyme and the unknown structure of B-lactamase CTX-M14
2Fo-Fcmap(at1.5σ)overthecorrespondingpartofthemodelofß-lactamaseCTX-M14frommultidrugresistantKlebsiellapneumoniae.
Detailedviewofthebindingoftheß-lactamaseinhibitoravibactamtoSer70.Theinhibitorisboundcovalentlytotheprotein(hemiacetal).
ResultsfromXFEL2012_2,initialrefinement:P3221UnitCell:41.8441.84233.289090120Rwork/Rfree:0.184/0.201AverageBISO:29.3RMSDbonds(Å):0.007RMSDangles(°):0.851
Page 30
More recently: super hot liquid water
Felix Lehmkuhler et al.
„Proceedings of the National Academy of Sciences“ (PNAS), 2020; DOI: 10.1073/pnas.2003337117
Page 31
The DEPFET Sensor with Signal CompressionDESY, EuXFEL, MPG-HLL, PNSensors, PolyTech Milano, INFN, Universities Heidelberg, Bergamo,
Courtesy: Matteo Porro
Page 32
Output voltage as function of charge
injected charge
DEPFET Sensor with Signal Compression DEPFET: Electrons are collected in a storage well
⇒Influence current from source to drain
source draingate
Fully depleted silicone-
Storage well
injected charge
DSSC
Non-Linear DEPFET
Page 33
§ 1 Megapixel camera 4.5 MHz peak frame rateØ 4 quadrants (512 x 512) Ø 16 ladders (512 x 128)Ø 32 monolithic sensors 128x256Ø 256 Readout ASICS 64 x 64 Hexagonal
§ Sensors:Ø MiniSDD arrays 1st camera: limited DRØ DEPFET arrays 2nd camera
§ Readout conceptØ In pixel analog to digital conversionØ In pixel digital storage (800 frames) with the
possibility to overwrite non-valid frames (VETO)Ø Output average data rate: 134.4 Gbit/s
DSSC Detector System Overview
256 x 128 256 x 128
detector ladder
Page 34
DSSC 1 Megapixel Camera (with mini-SDD iso non-linear DEPFETs)
§ In May 2019, DSSC has beeninstalled and commissioned at theSpectroscopy and CoherentScattering (SCS) Instrument
§ 1-Megapixel diffraction image of pinholes with 707 eV photons
Page 35
DEPFET sensors for the second camera
§ In the second DSSC camera the MiniSDD sensors will be replaced by DEPFET arrays
§ DEPFETs are active pixels that provide a non linear responseØ Low noise for single photon
detectionØ High dynamic range
§ The DEPFET arrays are compatible with the existing DSSC system
§ A maximum signal > 104 ph / pixel is achievable for E ≥ 800 eV, assuming an 8-bit ADC and single photon detection capability
Measurements on DEPFET pixel prototypes
Page 36
Future Sources
and
Detectors
Page 37
Requirements for future sources
CW-XFEL (>2028)PETRA-IV (~2028)
• Diffraction-Limited Storage Ring• Approx continuous X-ray beams• x 100 increase in X-ray brilliance• Measurements from atomic to
macroscopic scale, 10µs resolution
• Free electron laser• Extremely intense X-ray pulses• 100 kHz to 1MHz continuous
bunch rate (source)• “Flash photography” on atomic
scales
Page 38
Experiments at PETRA-IV
More than100-fold increase in coherent flux
Page 39
Example – coherence-based imaging (e.g. ptychography)
x 100 acquisition speedè high frame rates
Single photonsè low noise
Direct beamè large DR
Large, seamlessdetectorè 4 side buttable
Page 40
§ ≥ 132 kHz continuous frame rate (PETRA revolution frequency)
§ ≤ 100 µm pixel size
§ Multi-megapixel (modular design)
§ Minimal dead area
§ Single photon sensitivity at 6 keV
§ ≥104 ph/pixel/image dynamic range (109 ph/pixel/s)
§Noise below Poisson statistics
§ Compatible with different sensors
§Standard Si sensors
§High-Z (e.g. CdTe) for hard X-ray detection
§Amplifying sensors for soft X-ray detection
High-speed imager – target specifications
Page 41
Integrating pixel detectors
> Existing integrating pixel detectors provide a strong starting point
AGIPD (DESY / PSI) Jungfrau (PSI)
3520 images per second352-image bursts at 5 MHz200 µm pixel size
2200 images per second16-image bursts at 1 MHz possible75 µm pixel size
Page 42
Achieving >100 kHz frame rate
>Overhaul of entire readout chain needed!
A-D converter
Ultrascale+FPGA serializer
Firefly multi-link fiber
New chip
New Readout board
Fast (5-10G)digital outputs Firefly multi-link fiber
Page 43
High Frame Rate Integrating ASIC design?
> AGIPD provides solid basis for analog pixel
§ High sensitivity and dynamic range through gain switching
> Digital side: fast digitization
§ used in DSSC detector for EuXFEL
§ presented in design papers by Italian PIXFEL consortium
§ Shared ADC between multiple pixels?
> Digitial side: high-speed off-chip drivers
§ 10 Gbit/s off-chip drivers should be possible in 65 nm CMOS!
§ Implemented in Timepix4 (65 nm CMOS, CERN, NIKHEF)
Inv. ampDyn. gain
switch
Adaptive gain preamp
Correlated double sampling
Hold
In
Analogpixel
Shared digital
ADC (e.g. 10 bit SAR, 1MHz)
10 Gigabit off-chip drivers
Memory
Page 44
Chip-level speed improvements
A-D converter
Ultrascale+FPGA serializer
Firefly multi-link fiber
New chip
Readout board
Fast (5-10G)digital outputs Firefly multi-link fiber
>On-chip analog to digital conversion
§Distributed ADCs throughout pixel matrix
> High-speed off-chip data transmission
§ Up to 10 Gigabit per signal pair (TBC)
Page 45
Readout board improvements
A-D converter
Ultrascale+FPGA serializer
Firefly multi-link fiber
New chip
New Readout board
Fast (5-10G)digital outputs Firefly multi-link fiber
96 x 5 Gigabit or48 x 10 Gigabit 24 x 25 Gigabit optical lanes
3 x 100 Gbit per fiber
> Improvements in FPGAs and optical links allow sufficiently fast readout tomeet specs
> Possible module design (using currently-advertised components)
§1000 x 250 pixels of 100 µm (100 mm x 25 mm)
§14 bit depth x 132,000 fps -> 462 Gbit/s
Page 46
Dynamic Range: Adaptive Gain (AGIPD & JUNGFRAU)
C+ -
+ -
+ -
Inv. amp
VoutQin
Dyn. gain
switch
~50 ph
~ 1200 ph
~ 10000 ph
>Maximum signal limited by capacitor size in pixel
>Charge removal circuitry could increase dynamic range
Page 47
Increasing dynamic range
>Charge removal schemes already developed by other groups
§Charge bucket removal – Cornell MMPAD
• M W Tate et al 2013, J. Phys.: Conf. Ser. 425 062004§Charge (Current) splitting – Femilab FASPAX
• T. Zimmerman and S. Junnarkar 2019, NIMA 923 p64
>Current strategy:
§First version of chip without charge removal (104 ph/px/image = 109 ph/px/s)
§ Investigate and prototype extended DR schemes for second version
Page 48
Reducing dead area using through-silicon vias (TSV)
Silicon pixel sensor
ASIC
Circuit board
ASIC
HD connectorMech. Mech.
Guard ring
Wire bonds
Dead area
Edgeless silicon pixel sensor
ASIC
Circuit board
ASIC Throughsiliconvias
HD connectorMech. Mech.
Page 49
Experience with TSV development
>Worked with Fraunhofer IZM on TSV development
§Using Medipix3 chip (TSV compatible but not fully optimised)
§Single chip TSV prototypes with sensors working
SEM image of viasSingle Medipix3 chip with TSVs and sensor
X-ray image takenwith chip
Page 50
The Data Challenge
Page 51
The Data rate and volume challenge
>Possible module design:
§1000 x 250 pixels of 100 µm (100 mm x 25 mm)
§14 bit depth x 132,000 fps -> 462 Gbit/s
>4 module (1 megapixel) -> 1848 Gbit/s, i.e. 231 Gbyte/s
Page 52
Real-time data processing and reduction
>Raw images must be converted to photons – preferably before saving
§Photon images can be compressed more effectively
§Long acquisitions require summing many shorter images
Raw AGIPD data
Corrected image
Page 53
Need for real-time image conversion
>Provide feedback during experiments
>Detector dynamic range is reliant on high frame rate
§For long exposures, sum multiple images together
>Photon images compress better than raw data
76 68 77 81
72 66 71 83
65 76 154 65
71 77 68 80
0 0 0 0
0 0 0 0
0 0 8 0
0 0 0 0
Raw data (ADUs) Photon image
Page 54
Real-time data processing, selection and reduction
> Plan to develop layer of hardware for image conversion and selection
>Different options will be investigated
§MicroTCA crate with FPGA cards
§PCs with accelerators (GPU, FPGA)
> Flexibility is an important factor (e.g. high-level synthesis of firmware)
>Work supported by collaboration projects with SLAC (Hir3x), China (CHILFEL) and within Helmholtz (Innovation Pool)
Page 55
Summary
> ”General purpose” readout chips (MediPix-3, TimePix-4,..) can find applications in many experiments at Storage Ring facilities
>Custom build readout chips (AGIPD, JUNGFRAU,..) needed for Free-Electron Laser facilities
> System aspects often more important than front-end chip performance
> Future Storage Rings (PETRA IV) and Free-Electron Lasers (CW-XFEL) more similar from detector point of view
>High-Speed Imager project well underway, feasibility of basic components studied and confirmed. Specific designs started.
>Data rate and Data volume need an online approach: data correction, data selection (trigger and veto), data compression.
>Usefulness of Machine Learning and Artificial Intelligence under study.
Page 56
Photon-Science Detector-Systems Group
