Introduction to
Synchrotron Radiation
Detectors
Heinz Graafsma
Photon Science Detector Group
DESY-Hamburg; Germany
heinz.graafsma@desy.de
The Detector Challenge:
Las e rs à
é l e ctron s l i bre s
Free-Electron Lasers
Brillance
d2/0.1%B .F.)
PETRA-3
1023
Li m i te de di ffracti on
1022
ESRF (2000)
brilliance
1021
1020
ES RF (fu tu r)
3è m e
ge n e rati on
ES RF (2000)
ESRF (1994)
1019
Synchrotron Sources
ES RF (1994)
1018
1017
16
10
15
10
2è m e ge n e rati on
Second generation
1014
1013
Rayon n e m e n t
s yn ch rotron
1è re gé n é rati on
First generation
1012
1011
1010
X-rayTu be s à
rayon s X
tubes
109
108
107
106
1900
1900
1920
1940
1960
Années
1980
1960
2000
1980
2000
2
The Detector Challenge:
3
XFEL Beamline Layout
4
The Detector Challenge:
• Spectroscopy (determine energy of the X-rays):
– meV – 1 keV resolution
– time resolved (100 psec) – static
• Imaging (determine intensity distribution)
– Micro-meter – millimeter resolution
– Tomographic
– Time resolved
• Scattering (determine intensity as function
momentum transfer = angle)
– Small angel – protein crystallography
– Diffuse – Bragg
– Crystals - liquids
5
What are the basic principles ?
1. In order to detect you have to transfer
energy from the particle to the detector
2. X-ray light is quantized (photons)
3. A photon is either fully absorbed or not
at all (no track like for MIPs)
4. The energy absorbed is transferred
into an electrical signal and then into a
number (digitized).
6
Signal Generation  Needs transfer of Energy
Any form of elementary excitation can be used to detect the
radiation signal:
Ionization (gas, liquids, solids)
Excitation of optical states (scintillators)
Excitation of lattice vibrations (phonons)
Breakup of Cooper pairs in superconductors
Typical excitation energies:
Ionization in semiconductors: 1 – 5 eV
Scintillation:
appr. 20 eV
Phonons:
meV
Breakup of Cooper pairs:
meV
7
Band structure (3)
Ge
Eg = 0.7 eV
Si
Eg = 1.1 eV
Indirect band gap
GaAs
Eg = 1.4 eV
Direct band gap
8
What would you like to know about your X-rays?
1.
2.
3.
4.
Intensity or flux (photons/sec)
Energy (wavelength)
Position (or mostly angles)
Arrival time (time resolved
experiments)
5. Polarization
9
4 modes of detection
1.
2.
3.
4.
Current (=flux) mode operation
Integration mode operation
Photon counting mode operation
Energy dispersive mode operation
10
Current mode operation
X-ray
detector
I
Integrating mode operation
X-ray
detector
C
V(t)
11
Photon counting mode
X-ray
detector
C
R V(t)
Upper threshold
Lower threshold
12
Energy dispersive mode
X-ray
detector
C
R V(t)
Height  total charge = energy of the photon
13
Some general detector parameters
• QE = quantum efficiency = fraction of incoming photons
detected (<1.0). You want this to be as high as possible.
• DQE = detective quantum efficiency =
signal noiseout  1.0
signal noise in
You can never increase signal, nor decrease noise! So
signal to noise will always degrade in the detector. (NB:
signal to noise is the most important parameter when
you measure something!)
• Gain = relation between your signal strength (V, A, ADU)
and the number of photons.
14
Some more parameters for 2D systems
• Point Spread Function (PSF) (Line spread
function (LSF) or spatial resolution):
A very small beam (smaller than the pixel size)
will produce a spot with a certain size and
shape. Very important are the FWHM; and the
tails of the PSF.
This is experimentally difficult  use sharp
edge and LSF
Note: pixel size is not spatial resolution! (but
should be close to it in an optimal design).
15
Some more parameters for 2D systems
• Modulation Transfer Function (MTF):
How is a spatially modulated signal (line pattern)
recorded (transferred) by the detector?
Modulation  contrast 
Max  Min
Max  Min
This depends on the frequency.
Is directly related to the LSF and the DQE
16
100 %
0%
17
Some more parameters for 2D systems
• Modulation Transfer Function (MTF) Example
100  0
 1.0
100  0
Ideal:
contrast 
Effect of noise:
150  50
contrast 
 0.5
150  50
Effect of PSF:
contrast 
75  25
 0.5
75  25
18
FEL Sources vs. Storage Rings
• Pulse length: 103 shorter (100 fsec vs 100
psec)
• Emmittance: 102 horizontal, 3 vertical
lower
• Intensity per pulse: 3x102 higher (1012 ph)
• Monochromaticity: 10 better
 Peak brilliance: 109 higher
19
FEL Challenge: Different Science
x109
• Completely new
science
• Fast science 100 fsec
• “Single shot” science
Consequences for the detector:
(H.Graafsma; Jinst, 4, P12011, (2009))
1. Single shot-science: 1012 ph in 100 fsec
 (complete) ionization of sample;
followed by coulomb explosion.
• Fortunately scattering is faster: “diffract-anddestroy”. (<50 fsec) (Nature 406, 752, (2000)).
• Crystal diffraction is “self-gating” (Nature Photonics, 6,
35, (2012)).
21
Single shot imaging…
K. J. Gaffney and H. N. Chapman,
Science 8 June 2007
Consequences for the detector:
1. Single shot-science: 1012 ph in 100 fsec 
(complete) ionization of sample; followed by
coulomb explosion.
• Fortunately scattering is faster: “diffract-and-destroy”.
(<50 fsec) (Nature 406, 752, (2000)).
• Crystal diffraction is “self-gating” (Nature Photonics, 6, 35,
(2012)).
2. Central hole in detector & no beamstop: 1012 ph
@ 12 keV  1K rise in mm3 Cu  3000 K per
bunch train + huge background
23
Radiation doses: worst case (?)
> 5000 h User-operation per year
> Undulator shared between 2 experiments  2500 hrs/exp./year
> Date taking 50% (rest alignment etc.)  1250 hrs/year
> Each branch can take ½ of the load: 15000 pulses/sec
 6.75 1010 pulses/year
> Certain experiments expect 5 x104 photons per pixel (200 mm) per
pulse. Small angle and liquid scattering always same place on detector
 3.4 1015 photons/year = 1016 ph/3 years
(@ 12 keV  silicon surface dose of 1 Giga Gray!!!)
Angular coverage / detector size:
> 12 keV = 0.1 nm in order to study features (d) to atomic resolution.
> Bragg’s law (2dsin(q) = l)  2q = 60 degrees  120 degrees total
2q
Liquid scattering: momentum transfer 10 A-1  200 degrees  back scattering
Angular resolution 2 examples:
Coherent Diffractive Imaging (CDI):
> 0.1 nm spatial features: dmin
> 100 nm samples (e.g. virus): D
Nyquist >2000 sampling points (pixels) 0.5 mrad
D2q = dmin x asin(l/2dmin) / 2D
X-ray Photon Correlation Spectroscopy:
Speckle size: Qs = l/D, D is sample or beam size
Compromise between sample heating (large beam) and speckle size
(small beam)
25 mm beam at l=0.1 nm  4 mrad speckles (80 mm at 20 m)
E-XFEL Challenge: Time structure = difference with “others”
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).
100 ms
100 ms
27 000 bunches/s
But with
4.5 MHz rep rate
600 ms
99.4 ms
X-ray photons
<100 fs
220 ns
FEL
process
XFEL Detector requirements
4.5 MHz
The XFEL solutions:
Hybrid Pixel Array Detectors
Hybrid Pixel Array Detector (HPAD)
Diode Detection Layer
• Fully depleted, high resistivity
• Direct x-ray conversion
• Silicon, GaAs, CdTe, etc.
Connecting Bumps
• Solder or indium
• 1 per pixel
CMOS Layer
• Signal processing
• Signal storage & output
Gives enormous flexibility!
X-rays
Hybrid Pixel Detectors
Pixelated
Particle
Sensor
Particle / X-ray
Amplifier &
Readout Chip
CMOS
Qsignal
Power
Indium Solder
Bumpbonds
Clock Inputs
Connection
wire pads
Power
Inputs
Outputs
Data Outputs
Particle / X-ray  Signal Charge  Electr. Amplifier  Readout  Digital Data
31
The new generation: Medipix et al.
Sensor
Sensor
Substrate
Sensor Substrate
Substrate
Al
InGaAs
InGaAs
UBM
UBM
UBM
Insulator
Insulator
Au
Au
bump
Au
Au
UBM
UBM
Al
Al
CMOS ROIC
32
Why are HPADs so popular ?
• Custom design of functionality: you design
your readout chip specific for your
application (unlike CCDs).
• Direct detection  good spatial resolution
• Massive parallel detection  high flux
• But: development takes long and is
expensive.
33
The Adaptive Gain Integrating Pixel Detector
The AGIPD consortium:
PSI/SLS -Villingen: chip design; interconnect and module assembly
Universität Bonn: chip design
Universität Hamburg: radiation damage tests, “charge explosion” studies; and sensor design
DESY: chip design, interface and control electronics, mechanics, cooling; overall coordination
Some Facts
6 years development
~ 20 people
Some Milestones
First 16x16 pixels prototype
End 2010
Definition of final design
Summer 2011
Production, assembly and test
>2013
The Adaptive Gain Integrating Pixel Detector
High dynamic range:
Dynamically gain switching system
Extremely fast readout (200ns):
Leakage comp.
1,4
Normal Charge
1,2
V
sensitive
amplifier
Vthr
Discr.
1,6
Trim
DAC
ADCmax
Output Voltage [V]
C1
Analogue pipeline storage
Analogue
encoding
C2
Control logic
1,8
C3
1,0
0,8
0,6
0,4
0,2
Cf=100fF
Cf=1500fF
Cf=4800fF
0,0
0
5000
10000
Number of 12.4 KeV - Photons
15000
AGIPD03 Gains
0 Gy
7,00E+03
Not corrected for variation of
charge injection circuit
6,50E+03
6,00E+03
Amplitude [LSB]
5,50E+03
5,00E+03
>Gain Ratios:
4,50E+03
>H/M= 23.0 (23.1; 23.0; 22.8; 23.0)
4,00E+03
>M/L= 4.37 (4.67; 4.23; 4.10; 4.46)
3,50E+03
3,00E+03
2,50E+03
0,00E+000
1,00E+003
2,00E+003
Analogue Pix 0
Analogue Pix 1
Analogue Pix 2
Analogue Pix 3
Gain Pix 0
Gain Pix 1
Gain Pix 2
Gain Pix 3
2,87E+3 + 3,26E+1x
2,80E+3 + 2,77E+1x
2,75E+3 + 2,52E+1x
2,77E+3 + 2,94E+1x
3,61E+3 + 1,42E+0x
3,96E+3 + 3,03E-1x
3,54E+3 + 1,20E+0x
3,79E+3 + 2,84E-1x
3,49E+3 + 1,11E+0x
3,70E+3 + 2,70E-1x
3,47E+3 + 1,28E+0x
3,76E+3 + 2,87E-1x
3,00E+003
4,00E+003
5,00E+003
Charge [Clk]
6,00E+003
7,00E+003
8,00E+003
9,00E+003
1,00E+004
AGIPD – Analogue Memory &
Radiation Hardness
>Droop (loss of signal)
•Time
R O Bu s
•Radiation dose
Analogue Mem
CDS
+
THR
SW
CTRL
DAC
Analogue Mem
RO Amp
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).
100 ms
100 ms
600 ms
99.4 ms
Write:
within 220 nsec
Store and read:
for 100 msec
AGIPD - Analogue Memory
100 msec “loss free” Charge Storage in
Analogue Pipeline
• Switch design is the challenge
• Thick oxide & MIM caps in IBM
process are OK
AGIPD analogue memory:
• DGNCAP (thick oxide n-FET
in n-well) caps
• Minimise voltage drop
across T1
– Floating n-well
– Special precautions for
radiation hardness needed
AGIPD03 Memory Leakage
Droop (av. reduction of dynamic range) [%]
20
0
-20
Transistor type –
connection of upper n-well –
potential of guard ring
-40
-60
-80
-100
-120
1,00E-004
RVT-IN-GND (0Gy)
RVT-VDD-GND (0Gy)
RVT-IN-GND (100kGy)
RVT-VDD-GND (100kGy)
RVT-IN-GND (1MGy)
RVT-VDD-GND (1MGy)
LP-IN-GND (0Gy)
LP-VDD-GND (0Gy)
LP-IN-GND (100kGy)
LP-VDD-GND (100kGy)
LP-IN-GND (1MGy)
LP-VDD-GND (1MGy)
LP-VDD-VDD (0Gy)
RVT-VDD-VDD (0Gy)
LP-VDD-VDD (100kGy)
RVT-VDD-VDD (100kGy)
LP-VDD-VDD (1MGy)
RVT-VDD-VDD (1MGy)
1,00E-003
1,00E-002
Storage time [s]
1,00E-001
1,00E+000
AGIPD ASIC
ASIC per pixel
HV
Pixel matrix
Analog Mem
CDS
+
-
SW
CTRL DAC
Analog Mem
RO Amp
THR
Calibration circuitry
Adaptive gain amplifier
352 analog memory cells
Chip
ASIC
periphery output
driver
RO bus (per column)
Sensor
…
…
Mux
Imaging with AGIPD 0.2 prototype
The Adaptive Gain Integrating Pixel Detector
Basic parameters
64 x 64
•1 Megapixel detector (1k  1k)
pixels
•200mm  200mm pixels
•Flat detector
•Sensor: Silicon 128 x 512 pixel tiles
•Single shot 2D-imaging
•4.5 MHz frame rate
•2  104 photons dynamic range
•Adaptive gain switching
•Single photon sensitivity at 12keV
•Noise 300e
•Storage depth 350 images
•Analogue readout between bunch-trains
bump bond
chip
Connector to interface
sensor
wire bond
HDI
Base plate
~ 2mm
~220 mm
1k x 1k (2k x 2k)
Calibration challenges:
> 106 x 3 gains; with > 10 points per gain curve: O(107)
> 106 x 350 storage cells > 10 points per droop curve: O(109)
> How to store the calibration data and how to correct data?
> How often do we need to recalibrate
> On-chip calibration sources
> Cross calibration with physics (photons, alpha, …)
> How long does this all take?
Some reflections on the future
• Active Sensors (DSSC)
45
DSSC - DEPMOS Sensor with Signal Compression
> DEPFET per pixel
> Very low noise (good for soft X-rays)
> non linear gain (good for dynamic range)
> per pixel ADC
> digital storage pipeline
> Hexagonal pixels
200mm pitch
> MPI-HLL, Munich
> Universität Heidelberg
• combines DEPFET
> Universität Siegen
• with small area
drift detector
(scaleable)
> Politecnico di Milano
> Università di Bergamo
> DESY, Hamburg
46
DEPMOS Sensor with Signal Compression
DEPFET:
Electrons are collected in a storage well
⇒Influence current from source to drain
gate
drain
source
Storage well
Fully depleted silicon
e-
Output voltage as function of charge
injected charge
injected charge
DSSC
47
Some reflections on the future
• Active Sensors (DSSC)
• Built-in intelligence per pixel (AGIPD)
48
Some reflections on the future
• Active Sensors (DSSC)
• Built-in intelligence per pixel (AGIPD)
• Communication pixels (Medipix-3)
49
Medipix3 – charge summing concept
The winner takes
all
• The
incoming
Charge
processed is
quantum
is every
assigned
summed in
4 as
a
single
hit on an
pixel
cluster
event-by-event basis
55m
50
Some reflections on the future
•
•
•
•
Active Sensors (DSSC)
Built-in intelligence per pixel (AGIPD)
Communication pixels (Medipix-3)
More functionality per area/pixel: 3D-ASIC
technology (Helmholtz Cube)
51
Hybridization

Cut the sensor as close as possible

Use thinned readout chips

Stay within the exact n-fold pixel pitch
52
XFS Module Specification: PSI/SLS
Operate 2x4 (8) Chips per Module. ~78 x 39 mm2
53
PILATUS @ SLS
Sensor
Wire bonds
Read-out chips
Base plate
Al support
Module Control Board MCB
Cable
Courtesy: Ch. Brönnimann, PSI SLS Detector Group
54
55
Current State-of-the-art
The “Helmholtz-Cube”
Vertically Integrated Detector Technology
Replace standard sensor with:
3D and edgeless sensors, as well
as High-Z
Replace standard bumpbonds with new
interconnect techniques
Replace standard ASIC and wire
bonds with thinned ASICs and
TSV as well as ball-grid arrays
Replace standard
ASIC with 3D-ASICs
Develop new high speed IO’s
58
3D-ASIC technology
59
60
The “Helmholtz-Cube”
Vertically Integrated Detector Technology
S
Sensor
Analogue ASIC
Digital ASIC
I/O ASIC
Electrical + Cooling + Support
Power-in
Optical I/O
62
Summary Detectors
• Signal-to-noise ratio most fundamental
parameter in measurements.
• A detector is always a compromise (ex.
speed vs. noise). Application determines
what you compromise.
• Never take a detector as a “perfect black
box”, be aware of limitations.
• Understanding your detector is part of
understanding your science.
63