SPAD Pixel Detectors with HighSPAD Pixel Detectors with HighTime ResolutionTime Resolution
Edoardo CharbonEdoardo CharbonTU DelftTU Delft
2
Photons
3
Not Only IntensityNot Only Intensity……
•• CountingCounting•• Time-of-arrivalTime-of-arrival•• CorrelationCorrelation
4
Correlating PhotonsCorrelating PhotonsPolaritons in GaAs
microcavity (λ=770nm)
Balili, Science 316, 1007 (2007)
Photon states
=gg(2)(2)(0)(0)
gg(1)(1)(0)(0)
ThermalThermal CoherentCoherentIncoherentIncoherent
221111
111100
≠
Green Hg line from Hg-Ardischarge lamp (λ=546nm)
Young’ interference fringes
5
Stellar Stellar Hanbury-Brown Hanbury-Brown and and TwissTwissInterferometerInterferometer
6
4x4 SPAD arrayCMOS chip
10Hz dark count rate 120dB dynamic range70ps resolution25% detection prob.
Modern g(2) ImagerModern g(2) Imager
•• On-chip electronics for digital outputs On-chip electronics for digital outputs•• Off-chip processing (e.g. with digital oscilloscope) Off-chip processing (e.g. with digital oscilloscope)•• 4x4 array: 120 HBT coincidence experiments running simultaneously 4x4 array: 120 HBT coincidence experiments running simultaneously
7
Time-resolved BioimagingTime-resolved Bioimaging
• Super-resolution Microscopy– Stimulated Emission Depletion (STED)– Single Plane Illumination Microscopy (SPIM)– Scanning Photoionization Microscopy (SPIM)
• Molecular Imaging– Fluorescence Lifetime Imaging Microscopy (FLIM)– Förster Resonant Energy Transfer (FRET)– Fluorescence Correlation Spectroscopy (FCS)
• Nuclear Medicine– Positron Emission Tomography (PET)– PET & Magnetic Resonance Imaging (MRI)– Single-photon Emission Computer Tomography (SPECT)
8
OutlineOutline
•• Single-Photon DetectionSingle-Photon Detection•• From Pixel to ImagerFrom Pixel to Imager•• Scaling Up ApplicationsScaling Up Applications•• The Next Big ChallengesThe Next Big Challenges
9
Single-Photon DetectionSingle-Photon Detection
10
Single/few-photon DetectorsSingle/few-photon Detectors
•• Charge coupled devices (Charge coupled devices (CCDsCCDs))•• Electron Multiplying Electron Multiplying CCDs CCDs ((EMCCDsEMCCDs))•• Streak CamerasStreak Cameras•• Photomultiplier Tubes (Photomultiplier Tubes (PMTsPMTs))•• Multi/micro-channel plates (Multi/micro-channel plates (MCPsMCPs))
Silicon Avalanche Photodiodes Silicon Avalanche Photodiodes ((SiAPDsSiAPDs))Single-Photon Avalanche DiodesSingle-Photon Avalanche Diodes ( (SPADsSPADs))
11
Multiplication in SiliconMultiplication in Silicon
•• ReviewReview::Photon to electron - Secondary electron - MultiplicationPhoton to electron - Secondary electron - MultiplicationMultiplication in depletion region by Multiplication in depletion region by impact ionizationimpact ionization
p+
n-
n+
V
+
-depletion region
Reverse bias
12
Linear (or Proportional) ModeLinear (or Proportional) Mode
High variability of gain From bias
n
p+V
IA
V
V
-IA
ConventionalAvalanche
opticalgain<G>
Vbd
1
Ve + Vbd
13
Geiger Mode (SPAD)Geiger Mode (SPAD)
V
V
-IA
Conventional AvalancheGeiger
opticalgain<G>
Vbd
1Ve + Vbd
n
p+V
IA
Virtually infinite gain
14
•• Reach-through APD (RAPD)Reach-through APD (RAPD)–– VerticalVertical structure, thick device, high voltages structure, thick device, high voltages
Early SPAD Early SPAD Si Si IntegrationIntegration
n+p
π
p+
Multiplication
AbsorptionMcIntyre et al.
n-substrate
p-epi
p+
n+
Cova et al.Multiplication
•• Patterned double epitaxial APD (DJ-SPAD)Patterned double epitaxial APD (DJ-SPAD)–– PlanarPlanar structure, thin device, rel. low voltages structure, thin device, rel. low voltages
15
Planar ProcessesPlanar Processes
p substratep substraten-welln-well
p+p+ p-p-
Electric Field Electric Field ξξ
Multiplication regionMultiplication region
•• p- guard ring for electric field reduction in edgesp- guard ring for electric field reduction in edges•• Prevention of premature edge breakdownPrevention of premature edge breakdown•• Creation of zone with constant electric fieldCreation of zone with constant electric field
16
Quenching the AvalancheQuenching the Avalanche
Passive quenching: Operation cycle:
t
VVbdbd
VVopop’’
V
V photonarrival
avalanchequenching
SPADrecharge
VVopop’’
RRqq
VIIAA
Dead time
DEAD TIME
17
Controlling Dead TimeControlling Dead Time
t
VVbdbd
VVopop’’
V
V photonarrival
avalanchequenching
SPADrecharge
Dead time
DEAD TIME
18Niclass, Thesis 2008
Double Threshold Active QuenchingDouble Threshold Active Quenching
19
Salient Specs in Salient Specs in SPADsSPADs
•• Dead timeDead time•• AfterpulsingAfterpulsing•• Dark countsDark counts•• Photon detection probabilityPhoton detection probability (PDP) (PDP)•• Timing resolutionTiming resolution
…… and in SPAD imagers and in SPAD imagers•• Cross-talkCross-talk•• PDP UniformityPDP Uniformity
20
Dark Counts: Dark Count RateDark Counts: Dark Count Rate
•• State-of-the-art State-of-the-art SPADs SPADs in dedicated technology:in dedicated technology:0.1~1Hz/0.1~1Hz/µµmm22
•• State-of-the-art CMOS State-of-the-art CMOS SPADsSPADs::1~10Hz/1~10Hz/µµmm22
1 15x15 50x50
1Hz250Hz
3kHz
Mechanisms:
–Band-to-band tunneling generation–Trap-assisted thermal generation–Trap/tunneling assisted generation
21
Band-to-band TunnelingBand-to-band Tunneling
Ineffective guard ring:Tunneling due to high doping
Effective guard ring:Low-probability tunneling
22
Guard Ring EfficacyGuard Ring Efficacy
•• Ineffective guard ringIneffective guard ring•• Thus, high DCRThus, high DCR
•• Uniform multiplication zoneUniform multiplication zone•• Good prevention of prematureGood prevention of premature
edge breakdownedge breakdown
Niclass, Charbon, et al., JSTQE’07Gersbach, Charbon, et al., ESSDERC’08
23
Dark Count RateDark Count RateN
iclass et al. 2006
24
Photon Detection ProbabilityPhoton Detection Probability
Gersbach, C
harbon, et al. SS
Sensors 2009
25
Timing ResolutionTiming Resolution
PMT: 28psCMOS SPAD: 47ps
[Becker & Hickl]
26
From Pixel To ImagerFrom Pixel To Imager
27
DIGITAL DOMAINDIGITAL DOMAIN
SPAD in CMOSSPAD in CMOS
RRQQ
VVOPOP
VIIAA
Passive quenching technique
OUTOUT
digitalpulse
VDDVDDVVOPOP’’
TTQQ
28
ChallengeChallenge
•• Photocharges Photocharges cannot be accumulated like incannot be accumulated like inCCDsCCDs
•• Photon pulses arrive when photons impingePhoton pulses arrive when photons impinge
How to capture photon counting?How to capture photon counting?
How to capture photon arrivals?How to capture photon arrivals?
…… in parallel, on thousands of in parallel, on thousands ofpixels!pixels!
29
Imaging: Three ArchitecturesImaging: Three Architectures
1.1. Random Access ReadoutRandom Access Readout2.2. Event-driven ReadoutEvent-driven Readout3.3. Fully-parallel ProcessingFully-parallel Processing
30
1. Random Access Readout1. Random Access Readout
•• ProsPros–– SimpleSimple
•• ConsCons–– Highly inefficientHighly inefficient–– Low frame rateLow frame rate–– Enormous number of photons lost!Enormous number of photons lost!
31
Random Access ReadoutRandom Access Readout
Logic GatesLogic Gates
Guard RingGuard Ring
AnodeAnode
Niclass, Charbon, et al. JSSC 05
32
First First MassiveMassive SPAD Pixel Array SPAD Pixel Array
Niclass, Charbon, Niclass, Charbon, ISSCCISSCC 05 05
33
Photon Counting UniformityPhoton Counting Uniformity
•• Uniform counting at low, medium and high illuminationUniform counting at low, medium and high illumination
34
Spatio-Temporal Spatio-Temporal UniformityUniformity
Ti:Sapphire femtosecond laserλ= 470nmTAC resolution = 4.88ps
0 5 10 15 20 25 30
50
55
60
65
70
75
80
FW
HM
/ps
column number
35
Cross-talkCross-talk•• Electrical cross-talk reduced by potential barrierElectrical cross-talk reduced by potential barrier•• Optical cross-talk alleviated by reduced number of carriers in avalancheOptical cross-talk alleviated by reduced number of carriers in avalanche
Niclass, C
harbon, et al. JSS
C 2005
36
4µs 10µs 25µs 100µs 1ms
Ultra-high Dynamic RangeUltra-high Dynamic Range
37
2. Event-Driven Readout2. Event-Driven Readout
•• ProsPros–– Ideal with low photon countsIdeal with low photon counts
•• ConsCons–– First photon of column detectedFirst photon of column detected–– Large dead timeLarge dead time
SPAD SPAD SPAD SPAD
ID IDTDC
COLUMN
38
LASP ArchitectureLASP Architecture
Niclass, Favi, Kluter, Gersbach, Charbon, Niclass, Favi, Kluter, Gersbach, Charbon, ISSCC2008, JSSCISSCC2008, JSSC 2008 2008
39
LASP:LASP:First Fully Integrated SensorFirst Fully Integrated Sensor
128x128 SPAD array32 parallel TDCs
R = 70-500nsTP = 97ps
6.4Gb/s I/Os
32 Event-driven MU
Xes
Niclass, Favi, Kluter, Gersbach, Charbon, Niclass, Favi, Kluter, Gersbach, Charbon, ISSCCISSCC 2008, 2008, JSSCJSSC 2008 2008
40
TCSPC TestTCSPC Test
41
3D Imaging: Time-of-flight Cam3D Imaging: Time-of-flight Cam
pulsedlight source
TOF measurement
3D image reconstruction d
d = (c/2) TOFd = (c/2) TOF
Single-photonsensor
target
Time-of-flight
42
Three Dimensional ImagingThree Dimensional Imaging
Accuracy:• 1mm
Frame rate:• 1Hz
Digital output
43
3. Fully Parallel Processing3. Fully Parallel Processing
Control/DataControl/DataLinesLines
Supply/Bias LinesSupply/Bias Lines
Pixel-levelprocessing
44
Pros and ConsPros and Cons
•• ProsPros–– Full parallelismFull parallelism–– No photons are lost within detection cycleNo photons are lost within detection cycle
•• ConsCons–– Readout bandwidthReadout bandwidth–– Substrate/supply noiseSubstrate/supply noise
45
MEGAFRAME:MEGAFRAME:Massive Integration in DSMMassive Integration in DSM
Y-D
ecod
er
32x16 Array
32x16 Array
Serializer
Serializer
I/O pads
I/O pads SPAD TDC
RegisterControls
Fine interpolator Coarse
interpolator
PLL clock
START
STOP
Principle of TDC
Implemented on 130nm CMOS
46
Pixel SchematicPixel SchematicThermometer
coder
16 element delay line
6b ripple counter
Quenching
SPAD
Cal
10b memory
Global clock280MHzFrequency doubler
Global STOPFF
START
Vdd
Delay element
Column data bus
Gersbach, Charbon, et al., ESSCIRC 2009
Pitch: 50umMax. Resolution: 119psBandwidth: 1MS/sAccuracy: 1.2LSB (INL)Timing jitter: 128ps (FWHM)Timing uniformity: < 2LSB
47
Pixel LayoutPixel Layout
Over 500 transistorsIn 50 x 50 µm2
48
TDC PerformanceTDC Performance
INL
DNL
49
TDC UniformityTDC Uniformity
50
Dark Count RateDark Count Rate
Median DCR: 100Hz
51
Timing JitterTiming Jitter
52
Gersbach, Maruyama, Labonne, Richarson, Walker,Grant, Henderson, Borghetti, Stoppa, and Charbon,ESSCIRC 2009
32x32 pixel array
I2C
PLL
1.6mm
• 1MS/s-pixel• 100ps resolution• 100ns range• 1.2LSB precision• 2LSB uniformity
The MEGAFRAME32 ChipThe MEGAFRAME32 Chip
53
Scaling Up ApplicationsScaling Up Applications
Less than 32 SPADsLess than 32 SPADs
55
Chemiluminescence ReactorChemiluminescence Reactor
Gersbach, Maruyama, Sawada, Charbon, µTAS’06
56
Chemiluminescence ReactorChemiluminescence Reactor
57
Chemiluminescence ReactorChemiluminescence Reactor
58
Integrated Integrated MicroreactorMicroreactor
SU8 channel reactor
IgG Reservoir
ECL Reservoir
Fluidic channel
In Situ Optical Detection
E. Charbon and Y. Maruyama,Springer, 2010
SPAD Array
32 to 1024 Pixels32 to 1024 Pixels
60
Two-photon FLIM SetupTwo-photon FLIM Setup
DichroicBeam Splitter
Mode-lockedTi:Sapphire
Laser (740~920nm)
TDC
Detector
Filter (λ=488nm)
Attenuator
Fluorescent sampleOn x/y table Histogram processing
61
Triple-exponential DecayTriple-exponential Decay
Fluorophore:Oregon Green Bapta-1
Gersbach, Charbon, et al.,Optics Letters, 2009
62
Wide-field One-photon FLIMWide-field One-photon FLIM
Rahmadi Trimananda
63
The SampleSource: W
est Geo
rgia
Micro
scopic C
enterBisaccate Pine Pollen (Magnification: 3200x)
64
Wide-field One-photon FLIMWide-field One-photon FLIM
254ms
Marek Gersbach
1024 to 20,000 Pixels1024 to 20,000 Pixels
The Megaframe-128 Chip
C. Veerappan, J. Richardson, R. Walker, D.-U. Li, M. W. Fishburn, Y. Maruyama, D. Stoppa, F. Borghetti, M. Gersbach, R.K. Henderson, E. Charbon, ISSCC2011
The Megaframe-128 Chip
The Megaframe-128 Chip
50um pitch12.3mm
11.0mm
Imager Block Diagram
Pixel ArchitectureM
att W. Fishburn
Photon CountingM
att W. Fishburn
Photon Time-of-ArrivalM
att W. Fishburn
TDC Characterization
55ps resolution, 55ns range
INL DNL
System-level Timing UncertaintyBlue laser Red laser
Cumulative Noise
Optical Burst Detection UniformityC
hockalingam V
eerappan
MEGAFRAME Summary
• Format: 160x128 pixels• Timing resolution: 55ps• Impulse resp. fun.: 140ps• DCR (median): 50Hz• R/O speed: 250kfps• Size: 11.0 x 12.3 mm2
78
MultisensorMultisensorChipChip
•• Pitch: 25Pitch: 25µµmm•• Single shot time Single shot time resres.:.:
230ps230ps
•• Readout speed:Readout speed:40~2441fps40~2441fps
•• PRNU:PRNU:3.5%3.5%
Y. Maruyama and E. Charbon, TransducersTransducers, 2011
7979
Multisensor Multisensor PrinciplePrinciple•• AnalysisAnalysis
–– ElectrochemicalElectrochemical–– OpticalOptical–– CombinationCombination
opticalexcitation
electrochemical analysis
optical analysis
Labeled and label-less DNA probes
80
DNA from Blood and UrineDNA from Blood and Urine
Yuki Maruyama
81
Point-of-care CyclePoint-of-care Cycle
Yuki Maruyama
Single-photon detection in
Medical Applications
83
Positron Emission Tomography(PET)
Most commonly used: Fludeoxyglucose (18F)
e+e+e-e-
AnnihilationAnnihilation
γγ
γγ
scintillatorscintillator
PMTPMTCoincidenceCoincidence
1m1m
84
Source: S
un
Source: S
un
Positron Emission Tomography(PET)
Cancerous Ganglion
85
The The SPADnet SPADnet ProjectProjectObjective:Fully digital, scalable photonic component capableof detecting single and multi‐photon bursts, theirtime‐of‐arrival and intensity
COMMUNICATIONDETECTOR
DATA BUS
86
The InnovationThe Innovation
•• SPAD sensors with massively parallel chip-levelSPAD sensors with massively parallel chip-leveltime detectiontime detection
•• Large format with through-silicon-via basedLarge format with through-silicon-via basedpackagingpackaging
•• Advanced optical couplingAdvanced optical coupling•• Network between sensors with high-speedNetwork between sensors with high-speed
message-passingmessage-passing•• Digital coincidence by hierarchical messageDigital coincidence by hierarchical message
snoopingsnooping•• Novel image reconstruction exploiting spatialNovel image reconstruction exploiting spatial
informationinformation
87
The ImpactThe Impact
•• Cheaper, simpler, scalable, robust Cheaper, simpler, scalable, robust PETsPETs•• Higher levels of reliabilityHigher levels of reliability•• Higher speed and flexibility in dataHigher speed and flexibility in data
processing for imagingprocessing for imaging•• Full compatibility with MRI and otherFull compatibility with MRI and other
imaging techniquesimaging techniques•• Use of existing and new radiotracers withUse of existing and new radiotracers with
low lifetime and high specificity will below lifetime and high specificity will befeasiblefeasible
88
The Next Big ChallengesThe Next Big Challenges
89
MooreMoore’’s Law for Single-photons Law for Single-photon
2003 2006
1 kpixel
32 pixel
10 kpixel
100 kpixel
1 Mpixel0.8 CMOS 0.35 CMOS
2009
90nmCMOS
160x128
112x4
64x48
128x2
128x128
2012
512x256
1M
130nmCMOS
90
Fill FactorFill Factor
Guard rings, design rules, on-pixelGuard rings, design rules, on-pixelprocessingprocessing
FF = 1%
0.8µm CMOS
FF = 9%
0.35µm CMOS
FF = 25%
0.13µm CMOS
59µm 25µm 15µm
10µm?
FF = 35%
90nm CMOS
91
How Far Are We from 1Mpx?How Far Are We from 1Mpx?
•• Current minimum pitch: 15Current minimum pitch: 15µµm (0.13m (0.13µµm)m)•• 1024x1024 pixels: 16x16mm1024x1024 pixels: 16x16mm22
•• Assuming a minimum pitch of 10Assuming a minimum pitch of 10µµm (90m (90nnm)m)•• 1024x1024 pixels: 11x11mm1024x1024 pixels: 11x11mm22
Richardson et al., IIS
W 2009
15µµmm
92
•• 2P FLIM (P. French, Imperial College, London)2P FLIM (P. French, Imperial College, London)
•• Fluorescence imaging in 9.4T MRI (with Prof. Fluorescence imaging in 9.4T MRI (with Prof. RudinRudin,,ETH)ETH)
•• SPIM*-FCS (with Prof. SPIM*-FCS (with Prof. LangowskiLangowski, Heidelberg), Heidelberg)
•• TIRF DNA probing (with COSMIC, Edinburgh)TIRF DNA probing (with COSMIC, Edinburgh)
•• NIRI (with Dr. Wolf, USZ)NIRI (with Dr. Wolf, USZ)
Bioimaging ProjectsBioimaging Projects
*Selective/Single Plane Illumination Microscopy*Selective/Single Plane Illumination Microscopy
93
Intra-operative ß+ probe (Intra-operative ß+ probe (CTI-ForimtechCTI-Forimtech))
Concept:Concept: wireless, disposable probe wireless, disposable probe–– Detect and localize small tumours or Detect and localize small tumours or metastatic metastatic lymph nodes intra-lymph nodes intra-
operativelyoperatively–– Guide biopsy probe to tumourGuide biopsy probe to tumour–– Delineate tumour borders or invasion Delineate tumour borders or invasion duringduring operations operations–– Search and localize tumourSearch and localize tumour
residuals at the end of theresiduals at the end of thesurgical interventionsurgical intervention
Target:Target: melanoma, pelvic melanoma, pelvic tumourstumours,,mediastinoscopymediastinoscopy
Other Medical ProjectsOther Medical Projects
94
Other Medical Projects (Cont.)Other Medical Projects (Cont.)
Intra-operative sensors (Intra-operative sensors (EndoTOFPET-USEndoTOFPET-US, FP7 project), FP7 project)
Concept:Concept: asymmetric PET with TOF asymmetric PET with TOF–– External External SiPM SiPM plateplate–– Endoscopic Endoscopic plateplate–– Ultra-sound guidanceUltra-sound guidance
Target:Target: prostate, pancreas prostate, pancreas
Rectal/intestinal Endoscope with Miniature detector array
External SiPMplate forcoincidence
Tumour
95
ConclusionsConclusions
•• Single-photon imagers are here to staySingle-photon imagers are here to stay•• New and old apps enabledNew and old apps enabled•• Next challengesNext challenges
–– More miniaturizationMore miniaturization–– More parallelizationMore parallelization–– More flexibilityMore flexibility–– Novel imaging paradigmsNovel imaging paradigms
96
Acknowledgements
http://cas.et.tudelft.nlhttp://cas.et.tudelft.nl