Light Response of PPDs Study - GWW at LCWS 2011

Measuring and Modelling the Light Response of Pixelized Photon Detectors with Signicant Photon-Induced Correlated Noise
Graham W. Wilson
University of Kansas
September 29, 2011
Graham W. Wilson (University of Kansas)
LCWS11 Granada: Calorimetry/Muons
September 29, 2011
1 / 36
Introduction
Introduction
Solid-state photon detectors such as those in use by CALICE and T2K are considered attractive candidates for light detection in a future linear collider detector when compared with the traditional PMT. Some advantages B-eld tolerance Cheap Apparent photon resolving capability Low voltage/size/integrability Some disadvantages Dark noise rate per mm2 (can be 500 kHz cf 0.5 Hz). Correlated noise: cross-talk and after-pulsing Dynamic range
Graham W. Wilson (University of Kansas) LCWS11 Granada: Calorimetry/Muons September 29, 2011 2 / 36
Introduction
Photo-detector response modelling

Question Does LED calibration data (points) follow Poisson model (red)?
ch1_led: MPPC Channel 1
Counts per bin
3000
ch1_led Entries 108589 Mean 523.4 226.5 RMS
2000
1000
200
400
600
Integrated Charge (ADC counts)
800
1000
Individual photons often lead to not just 1 red pixel, but 2, 3, 4 etc. due to cross-talk and after-pulsing. Technology and device dependent.
Motivation
Noise and Time Structure: Important or Not ?

ILC/CLIC At ILC with a large inter-bunch spacing (337 ns), a relevant issue is noise within a time interval consistent with the BX under consideration. For CLIC, much tighter specs. Various types of potential detectors: AHCAL, Scint. ECAL, Muon-detector, scint. ber tracking have dierent requirements. Why this study? These studies use the 100-pixel Hamamatsu MPPC aimed at high eciency low light-level detection. I work on the D0 ber tracker and its light calibration and have explored instrumenting parts of a LC detector with similar instrumentation for time-stamping/TOF. At D0 using VLPCs we have an average threshold per ber of 1.0 photo-electrons corresponding to 1% dark probability and Poisson distributed LED distributions....
Motivation
Literature and this work

Literature There is a substantial literature from many groups about characterization measurements related to pixelized solid-state photo-multiplier response (see references). Short-comings In practice many adopted methods are either approximate, or require dedicated measurements unlikely to be available en masse in situ for LC detector channel counts, or potentially lose substantial predictive power / precision by using more parameters than warranted by the underlying eects. This work Hopefully helps to put some of the issues better in focus and provides a coherent yet relatively simple integrated experimental method to test the response modelling and the measurement of associated parameters.
Method
Experimental Goal and Design

Goal Measure the photo-detector response under essentially identical conditions using identical time intervals for both non-illuminated (dark) data and illuminated (LED) data. Currently using MPPC-S10362-11-100C sensor with 100 pixels. Design Use 100-1000 Hz clock to re LED on every other cycle. Measure total charge integrated within a 240-300 ns time window using dual-range QDC (FSR 800 pC). Use single-hit TDC in common start mode to check timing characteristics. Bias will be varied from run-to-run, but use 70.65 V corresponding to 0.72V over-voltage as default setting.
September 29, 2011
6 / 36
Method
Experimental Setup
LED intensity control V1 supply Bias and LabView amplification interface control CAEN SP5600 LED trigger pulse
HP pulse generator (V2)
VM-USB
LED Box Power Supply Jacketed clear fiber and Amplifier with MPPC1mm S10362-11100C sensor Lecroy 622 OR
PS744 Linear FI/FO QDC gate
V965 QDC
Ortec 935 CFD
Linux DAQ computer V775 TDC
CAEN V993 Dual Timer
TDC Start VM-USB
Ch1 : Pulses at 2n t Ch2 : Pulses at (2n+1) t
(LED trigger) (Dark count measurement)
September 29, 2011
7 / 36
Method
LED Circuit
Use Kapustinsky [1] type driver circuit to obtain fast time response. Currently using o-the-shelf green LED ( = 565nm).
Agilent E3630A variable supply voltage
HP 8131A pulse generator for trigger pulse
LED
September 29, 2011
8 / 36
Method
Typical LED Event 1: prompt 1 pixel event ( 40 ns)
September 29, 2011
9 / 36
Method
LED Charge Distribution (gate width = 300 ns)

Measure charge response to LED (black) and dark counts (red). Photon intensity = 0.22 detected photons/pulse.
ch1_ped: MPPC Channel 1
Counts per bin
2500
ch1_ped ch1_led Entries 100000 Mean 380.2 349.4 94.56 61.76 RMS
2000
1500
1000
500
300
400
500
600
700
800
September 29, 2011
10 / 36
Method
LED Charge Distribution - Log

Measure charge response to LED (black) and dark counts (red). Photon intensity = 0.22 detected photons/pulse.
Counts per bin
103
ch1_ped ch1_led Entries 100000 Mean 380.2 349.4 94.56 61.76 RMS
102
10
1 300 400 500 600 700 800
September 29, 2011
11 / 36
Method
LED Time Distribution

Measure time response to LED (black) and dark counts (red). Photon intensity = 0.22 detected photons/pulse.
Events per bin
2000
ch1_ped ch1_led Entries 56495 63201 Mean 229.8 1905 RMS 58.02 1158
1500
1000
500
0 150
200
Measured Time After Start (TDC counts)
250
300
350
400
Find FWHM for this green LED of 5ns. (1 TDC count = 0.3 ns)
Method
Typical LED Event 2: prompt 2 pixel event
September 29, 2011
13 / 36
Method
Typical LED Event 3: prompt 3 pixel event + delayed pulse
September 29, 2011
14 / 36
Method
Typical LED Event 4: multiple pixels + likely afterpulses
September 29, 2011
15 / 36
Method
Charge Distribution Studies
Standard conditions mostly: Room temperature (Kansas summer day ....) Integration time: 270 ns LED supply voltage: 18.25 V. Leading to 4.5 detected photons at 1 V over-voltage and nominal intrinsic gain of 2.4e6. Amplier gain: 38 dB
September 29, 2011
16 / 36
Method
Fitting Method
Assume that the measured LED distribution, T(Q), arises from the convolution of N(qdark ) and L(qlight ), where Q = qdark + qlight .
Counts per bin
15000
ch1_ped ch1_led Entries 100000 Mean 109.9 257.1 RMS 102.9 23.3
10000
5000
0 0
200
400
600
Use measured non-illuminated data (red) for N(qdark ) and adjust the parameters of the L(qlight ) model to give the best t to the LED data (black) for the additional charge arising from LED photons.
Method
Light Model: Essentially 2 parameters.

Pixel distribution Use the simple probability distribution model discussed in Vinogradov et al [2] based on a Poisson distributed random variable for the number of photons, , which result in a detected primary avalanche. Each primary avalanche may then create a secondary avalanche with a duplication probability, pD , and likewise a secondary avalanche may create a tertiary avalanche with the same probability etc, etc ... pD includes both cross-talk and after-pulsing assuming for simplicity equal charge response to both sources. Charge Measurement Model Parameters 1 pixel gain, g (ADC counts per pixel) 1 pixel intrinsic fractional gain resolution, 1 = g /g N-pixel non-linearity, , where qlight = Ng (1 + N)/(1 + )
2 2 N-pixel variance non-linearity, , where N = N(1 + N)g )/(1 + )
Method
Light Model Continued

Example p(nA = 0) = exp{ } p(nA = 1) = exp{ } (1 pD )
1 p(nA = 2) = exp{ }{ pD (1 pD ) + 2 2 (1 pD )2 }
p(nA = 3) = 2 exp{ }{ pD (1 pD ) + 2 pD (1 pD )2 + 1 3 (1 pD )3 } 6 etc .. Needless to say - currently neglect nite pixel number Characteristics Mean pixel count: /(1 pD ) Variance of pixel count: ( /(1 pD ))(1 + pD )/(1 pD )
Results
Example Fit (Run6 70.35V 38dB L)

Counts per bin
3000
ch1_led Entries100000 Mean 257.1 RMS 103
2000
1000
0 0
200
400
600
20 / 36
September 29, 2011
Results

Counts per bin
103
102
10
1 0
200
400
600
21 / 36
September 29, 2011
Results
Counts per bin
3000
2000
1000
pD (%) g 1 2 /dof
2.150(7) 5.00(25) 66.35(6) 0.113(1) -0.0022(4) 0.007(10) 936 / 794
0 0
200
400
600
September 29, 2011
22 / 36
Results

Counts per bin
3000
ch1_led Entries 108589 Mean 523.7 RMS 226.9
2000
1000
200
400
600
800
1000
1200
Fit: = 3.36 0.01, pD = 14.0 0.2%, 2 /dof=1921/1594.

Results

Counts per bin
2500
ch1_led h_fitted Entries 1600 Entries 100000 Mean 275.3 Mean 269.4 RMS 90.95 RMS 93.42
2000
1500
1000
500
100
200
300
400
500
600
Fit: = 3.36 0.01, pD = 13.0 0.3%, 2 /dof=774/794.

Results

Counts per bin
2000
1500
1000
500
100
200
300
Fit: = 1.268 0.006, pD = 0, 2 /dof=510/394.

Results
Example Fit (Run11 70.65V 38dB 0.3L)

Counts per bin
4000
3000
2000
1000
200
400
600
Fit: = 0.961 0.004, pD = 15.0 0.3%, 2 /dof=1424/1194.

Results
Example Fit (Run12 70.65V 38dB 1.6L)

Counts per bin
ch1_led Entries 100000 Mean 746 RMS 293.9
1500
1000
500
500
1000
1500
Fit: = 5.22 0.02, pD = 13.4 0.2%, 2 /dof=2327/1994.

Results
Bias Scan
Bias Scan 200
Gain (ADC counts per fired pixel)
Slope = 153.6 2.5 (ADC counts/V) V BD = 69.93 0.01 (V) 150 2/ndf = 8.6 / 4
100
50
70
70.2
70.4
70.6
70.8
Bias Voltage (V)
71
September 29, 2011
28 / 36
Results
Measured Photon Number

Measure the Poisson parameter of the detected primary LED photons, . (gives relative photon detection eciency vs V ).
Bias Scan
LED photon number
2 Linear term = 5.41 0.11 (photons/V) 1 NL term = -0.97 0.13 (photons/V^2) 2/ndf = 0 0.2 0.4 10.3 / 4 0.6 0.8 1
Over-Voltage (V)
Results
Duplicate Probability
Bias Scan
Pulse duplication probability (%)
30
Quadratic term = 27.54 0.80 (V^-2) 2/ndf = 4.5 / 5
20
10
0 0.2
0.4
0.6
0.8
Over-Voltage (V)
September 29, 2011 30 / 36
Results
Photon Resolving Quality

Gain Dispersion Parameter
0.25
Pixel Number Resolution

0.2
0.15
0.1
0.05
0 0
0.2
0.4
0.6
0.8
1.2
1.4
Over-Voltage (V)
Results
Dark Count Probability
Can measure charge exceeding arbitrary threshold in charge integration window (has not been the focus of this work so far) using measured non-illuminated ADC spectrum.
September 29, 2011
32 / 36
Results
Future Renements
Adjust LED intensity at each bias voltage to keep number of detected photons approximately constant. Should be a better method for measuring the photon-induced noise dependence on over-voltage. Incorporate after-pulsing time structure together with pulse shape time constant and resolve cross-talk and after-pulsing importance. Improve/better characterize charge measurement systematics amplier/ADC linearity. Improvements in the tting, statistical method and convolution.
September 29, 2011
33 / 36
Conclusions
Conclusions
Summary Find that measured pixel number distributions can be well modelled using a simple light-detection model including duplicate avalanches. Important for understanding particle detection eciency close to threshold. Experimental technique allows simultaneous determination of gain, duplicate probability, relative PDE, dark rate and pixel resolving capability for this particular sensor. Conclusions New PPDs have signicant noise which is sometimes correlated and needs to be taken into account when evaluating detector/sensor options, predicting performance and measuring performance. Correlated noise makes the standard procedure of setting a threshold at 1.5 photo-electrons not as useful as might be expected.
References
J.S. Kapustinsky et al. A Fast Timing Light Pulser for Scintillation Detectors NIM A, 241 (1985) 612. S. Vinogradov et al. Probability Distribution and Noise Factor of Solid State Photomultiplier Signals with Cross-Talk and Afterpulsing 2009 IEEE Nuclear Science Symposium, N25-111. Y. Du and F. Reti`re, e After-pulsing and cross-talk in multi-pixel photon counters NIM A, 596 (2008) 396. A. Vacheret et al., Characterization and simulation of the response of MPPCs to low light levels NIM A, 656 (2011) 69.
September 29, 2011
35 / 36
References
H. Oide et al., Studies on multiplication eect of noises on PPD, and a proposal of a new structure to improve performance NIM A, 613 (2010) 23.
September 29, 2011
36 / 36
Backup Slides
September 29, 2011
37 / 36

Light Response of PPDs Study - GWW at LCWS 2011

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Light Response of PPDs Study - GWW at LCWS 2011

Uploaded by

Copyright:

Available Formats

Measuring and Modelling the Light Response of Pixelized Photon Detectors with Signicant Photon-Induced Correlated Noise

September 29, 2011

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

Photo-detector response modelling

Counts per bin

ch1_led Entries 108589 Mean 523.4 226.5 RMS

Integrated Charge (ADC counts)

Noise and Time Structure: Important or Not ?

Literature and this work

Experimental Goal and Design

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

PS744 Linear FI/FO QDC gate

Ortec 935 CFD

Linux DAQ computer V775 TDC

CAEN V993 Dual Timer

TDC Start VM-USB

Ch1 : Pulses at 2n t Ch2 : Pulses at (2n+1) t

(LED trigger) (Dark count measurement)

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

HP 8131A pulse generator for trigger pulse

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

Typical LED Event 1: prompt 1 pixel event ( 40 ns)

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

LED Charge Distribution (gate width = 300 ns)

Counts per bin

Integrated Charge (ADC counts)

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

LED Charge Distribution - Log

Counts per bin

1 300 400 500 600 700 800

Integrated Charge (ADC counts)

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

LED Time Distribution

Events per bin

Measured Time After Start (TDC counts)

Typical LED Event 2: prompt 2 pixel event

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

Typical LED Event 3: prompt 3 pixel event + delayed pulse

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

Typical LED Event 4: multiple pixels + likely afterpulses

Graham W. Wilson (University of Kansas)

LCWS11 Granada: Calorimetry/Muons

September 29, 2011

Charge Distribution Studies