Transcript Pulse measurements - High Energy Physics at Wayne State

Pixilated Photon
Detectors
and possible uses at
ILC and SLHC
WSU, 23 Oct 09
Rubinov “at” fnal.gov
23 Nov 09
Rubinov, WSU HEP seminar
1
-- T. Nakaya (Kyoto) @ Pixel08 --
Intro to SiPM
 Q: What is an SiPM?
 A: SiPM (Silicon Photo Multiplier)
MRS-APD (Metal Resistive Semiconductor APD)
SPM (Silicon Photo Multiplier)
MPGM APD (Multi Pixel Geiger-mode APD)
AMPD (Avalanche Micro-pixel Photo Diode)
SSPM (Solid State Photo Multiplier)
GM-APD (Geiger Mode APD)
SPAD (Singe Photon Avalanche Diode)
MPPC (Multi Pixel Photon Counter)
From
Yamamoto
Pixelated Photon Detector
23 Nov 09
Rubinov, WSU HEP seminar
2
2
Photodiodes, Avalanche, Geiger
Mode
From A Para
(Fermilab)
Photodiodes:
• p-n junction , reverse bias
• Electron-hole pair generated by an incoming
photon drifts to the edges of the depleted region
• I(t) = QE * q * dN/dt(t)
• Absolute calibration
• No gain
• Suitable for large signals
23 Nov 09
Rubinov, WSU HEP seminar
3
3
Photodiodes, Avalanche, Geiger
Mode
From A Para
(Fermilab)
Avalanche Photodiodes:
• Photodiodes operating at higher bias voltage
• Higher voltage  stronger electric field  higher energy of drifting
carriers  impact ionization  Gain
• (Im)Balance between the number of carriers leaving the depletion
region and the number generated carriers per unit time: dNleave/dt >
dNgenerated/dt
•Stochastic process: signal quenches when the ‘last’ electron/hole
fails to ionize.
• Large fluctuations of the multiplication process  Gain fluctuations
Excess noise factor (beyond-Poisson fluctuations)
23 Nov 09
Rubinov, WSU HEP seminar
4
4
Photodiodes, Avalanche, Geiger
From A Para
Mode
(Fermilab)
Geiger Mode Avalanche Photodiodes:
• Avalanche Photodiodes operated at the
elevated bias voltage.
• Larger field  carriers gain kinetic energy
faster  shorter mean free path
• Breakdown voltage: nothing really breaks
down, but dNleave/dt = dNgenerated/dt (on
average) at this voltage
• Some electrons can generate selfsustaining avalanche (current limited
eventually by the series resistance)
• Probability of the avalanche generation
increases with bias voltage (electric field)
• Operation mode: one photon (sometimes)
~1e6 electron avalanche
23 Nov 09
Rubinov, WSU HEP seminar
5
5
First prototype of MAPD (MRS APD 1989)
First metall-resistor-semiconductor
A.Gasanov, V.Golovin, Z.Sadygov
APD (MRS APD) structure was designed
(russian patent #1702881, from 10/11/1989)
Anfimov Nikolay, Dubna, JINR
PDE of MRS APD is just few %
23 Nov 09
Low-light intensity spectrum
of MRS APD (A. Akindinov
et. al, NIM387 (1997) 231
Rubinov, WSU HEP seminar
6
by
23 Nov 09
Rubinov, WSU HEP seminar
7
7
23 Nov 09
Rubinov, WSU HEP seminar
8
Q=CD*(Vbias-Vbd)
23 Nov 09
Rubinov, WSU HEP seminar
9
9
Intro to SiPMs
 Analogy
 When comparing PMTs to SiPMs, SiPMs
enthusiast usually list advantages of SiPMs
BORING and PREDICTABLE
 I list advantages of conventional PMTs on next
slide from a tube company
23 Nov 09
Rubinov, WSU HEP seminar
10
PMT vs SiPM
Adapted from IEEE & Eric Barbour
Tubes: Advantages
1. Characteristics highly independent of temperature.
2. Wider dynamic range, due to higher operating voltages.
3. Very low dark current.
SiPM: Disadvantages
1. Device parameters vary considerably with temperature,
complicating biasing.
2. May need cooling, because lower operating temperature
may be required.
23 Nov 09
Rubinov, WSU HEP seminar
11
Analogy?
I think we have seen
transition from vacuum
tubes to solid state
before.
Transistors are not tiny
vacuum tubes and
SiPMs are not tiny
PMTs
I think that the reason we have transistors instead of tubes boils down to this:
$
23 Nov 09
Rubinov, WSU HEP seminar
12
Is the SiPM the perfect LLL
sensor?
• Die eierlegende Woll-Milch-Sau (german)
(approximate english translation: all-in-one device suitable
for every purpose)
R. Mirzoyan
There will be different devices
optimized for different applications
23 Nov 09
Rubinov, WSU HEP seminar
13
SiPM Animal
 Research in SiPMs is very active, in many
different directions – I’m not going to do a
survey
 Extended blue sensitivity (Cherenkov light, dual
readout calorimetry)
 Increased PDE ( muon detectors)
 Reduced crosstalk (improved noise factor)
 Improved timing (PET)
 Large area (Cherenkov)
 Increased dynamic range (Calorimeters)
 LOWER COST (everyone)
23 Nov 09
Rubinov, WSU HEP seminar
14
Areas of interest for LHC
 Issues of special interest to SLHC
(more detail on CMS specifics later)
 Radiation hardness
 Dynamic range/Linearity
 Stability (radiation, temperature and time)
 CERN has a strong, active community
working on all these issues
23 Nov 09
Rubinov, WSU HEP seminar
15
Areas of interest for LC
 Issues of special interest to LC
(more detail on SiD specifics later)
 Blue sensitivity
 Cost/unit area
 Optical coupling to detector
 Calice is a strong collaboration doing
fantastic work on these areas
23 Nov 09
Rubinov, WSU HEP seminar
16
Understanding
SiPM operation
 Here I'm going to focus on 2 issues
 DC measurements
 Vb determination
 Rquench determination
 Cross talk measurement
 Pulse measurements
 Afterpulsing measurements
23 Nov 09
Rubinov, WSU HEP seminar
17
DC Measurements
 Static characteristics - IV curves at fixed
temperatures:
 Keithley 2400 sourcemeter
 Temperature controlled chamber
 Labview data acquisition program
 Forward bias  series (quenching) resistance
 Reverse bias  breakdown voltage, integral
behaviour of the detector s a function of the operating
temperature
23 Nov 09
Rubinov, WSU HEP seminar
1818
Forward Bias Scan
Limited by quenching
resistor
dI/dV = 1/R
Exponential
growth with V
Resistance decreases
with temperature
(polysilicone)
23 Nov 09
Rubinov, WSU HEP seminar
1919
Quenching Resistance
Summary for MPPCs
Detector
type
Quenching
Resistor
@ 25 oC, k
dR/dT
k/oC
1/R dR/dT
25 
200
2.23
0.011
50 
105
1.08
0.010
100 
85
0.91
0.011
From A Para
(Fermilab)
23 Nov 09
Rubinov, WSU HEP seminar
2020
Reverse Bias Scan
Quenching
resistance
Temperature
100 pA
1 V above breakdown
I~5x10-7A
Gain ~ 4x106
‘Photodiode’ current
level ~ 10-13 A
How relevant is the
current below the
breakdown voltage?
Breakdown
From A Para
(Fermilab)
23 Nov 09
Rubinov, WSU HEP seminar
2121
Vbd(T). Preliminary analysis
33
y = 0.0002x 2 + 0.0482x + 29.168
FermilabIRST #30
August 29th 2008
Diego Cauz
University & INFN of Udine
32
Vbd (V)
31
30
29
28
27
-50
23 Nov 09
-40
-30
-20
-10
0
T (C)
Rubinov, WSU HEP seminar
10
20
30
40
50
22
Cross Talk Measurement
Single avalanche rate
Single + 1 cross talk
Single +2 cross talk
Ratios of
rates give
relative
probabilities
of 1,2,3
extra pixels
firing due to
cross-talk
Single +3 cross talk
23 Nov 09
Rubinov, WSU HEP seminar
23
Cross Talk Rates as a Function of Bias
Voltage
• Cross talk probability increases with the bias
voltage
• Cross talk probability is bigger for larger
size pixels
But… The cross talk is mediated by infrared
photons produced in the avalanche, hence is
ought to be proportional to the gain. And
different size pixel detectors have different
gain !
23 Nov 09
Rubinov, WSU HEP seminar
24
Cross Talk Probability as a Function of
Gain
• At the same gain the cross-talk probability is much larger for smaller
size pixels
• At the operating point the Hamamatsu detectors have very small
cross talk (~few %)
23 Nov 09
Rubinov, WSU HEP seminar
25
Pulse measurements
 MPPC-11-050C#37 at 71.1deg F operating at 69.81 (recommended V
is 70.02 at 25C)
 Current reading is 0.044uA
 1pe is about 13.25mV
23 Nov 09
Rubinov, WSU HEP seminar
26
A little bit about after pulses
Observed signal grows with the bias voltage.
This growth has several components:
• increase of the gain
• increase of afterpulsing.
The latter is a much bigger effect. So what??
Afterpulses provide a kind of additional gain.
True, but this contribution fluctuates 
degrades the charge measurement resolution
(excess noise factor).
Relative width of the observed pulse height
spectrum slightly decreases with bias voltage
for 10 nsec gate (presumably a reflection of
the increased number of detected photons),
but it increases for longer gates.
Bottom plot shows a contribution to resolution
from fluctuations of the afterpulses
contribution in different gates.
23 Nov 09
Rubinov, WSU HEP seminar
27
Detector Recovery / Afterpulsing
 Pulse arrival
distribution: clear
afterpulsing for
about~ 1 sec
 At least two
components:
 1=39 nsec
 2=202 nsec
 These components
probably correspond
to traps with different
lifetimes
23 Nov 09
Rubinov, WSU HEP seminar
2828
F. Retiere @ NDIP08
Photo-Electrons
S10262-11-050C
short~15ns
long~85ns
Dark-noise rate
Time after the first pulse (ns)
23 Nov 09
Time after the first pulse (ns)
Rubinov, WSU HEP seminar
29
29

After subtracting the effects of cross-talk +
after pulse, the dark noise is found to be linear
to V.
F. Retiere @ NDIP08
23 Nov 09
Rubinov, WSU HEP seminar
30
30
SiPM pulse shape
 Actually, there is some subtle issues in
measuring pulse shape
23 Nov 09
Rubinov, WSU HEP seminar
31
 The idea is to model the avalanche as a fast, brief (almost) short across a
capacitor (Cdet) which is then recharged through a resistor (Rq)
 this is one micro pixel, so 1 pe by definition
 Also include parasitic capacitance across this resistor (Crq)
 Also model the rest of the device by a collection of Cdetp, Rqp, Crqp
 the parallel stuff is important, it gives that characteristic “kink”
This kink
is this
plus this
There are 4 values of Crq from 1 to 10 fF. So Crq is important for “spike” but not “tail”
Crq= 10fF, 5fF, 2.5fF, 1fF
 and this is what we are left with...
 So the size of the “spike” makes a huge difference to the shape
of what is observed- including the integral
 But, the slow component is not so affected
 This fig has 8 plots: before and after the filter for each value of Crq
 But its even worse than that...
 The details of the assumed filter make a big difference as well
 I picked this very gentle, 6db stop band filter to prevent this...
For this run, I dropped the
Crq=10fF curve
These are 5fF, 2.5fF and
1fF curves
recall that Cdet is 3fF for
this MPPC 025u
 ... how about we lower the HiFreq cutoff and concentrate on the
shape of the falling edge. Lets say cut at 100MegHz
 So that corresponds to digitizing at 200MSPS
Simulation vs reality
70mV
60mV
50mV
40mV
30mV
20mV
10mV
0V
-10mV
38
Using SiPMs
Until you have spread your wings, you will have
no idea how far you can walk
despair.com
23 Nov 09
Rubinov, WSU HEP seminar
39
CMS
 Two approaches
 Straight replacement of the HPD
 Coupling individual fibers to individual SiPMs:
Electrical Decoder Unit
23 Nov 09
Rubinov, WSU HEP seminar
40
23 Nov 09
Rubinov, WSU HEP seminar
41
CMS
23 Nov 09
Rubinov, WSU HEP seminar
42
Linearity number of cells is the issue
23 Nov 09
Rubinov, WSU HEP seminar
43
Radiation is an issue
23 Nov 09
44
EDU
 The EDU
 100% compatible with existing mechanics/optics
23 Nov 09
Rubinov, WSU HEP seminar
45
CMS
Either of these could use fantastic new
devices from Zecotek
23 Nov 09
Rubinov, WSU HEP seminar
46
Anfimov Nikolay, Dubna, JINR
MAPDs main characteristics
MAPD-1 with surface
pixels (p-type substrate)
556 pixels*mm-2
MAPD-3N with deep
microwells (n-type substrate)
15 000 pixels*mm-2
47
ILC- SiD
 For SiD there are two possible uses of SiPM
 HCAL : 3x3 cm cells directly coupled to SiPMs
 Tail catcher/Muon system with scintilator strips
and WLS fibers coupled to SiPMs
23 Nov 09
Rubinov, WSU HEP seminar
48
Scint HCAL for SiD
The key issue here is coupling of the scintillator to SiPM
Northern Illinois University has some very clever and pioneering work on this
(basic idea is to put a dimple in the center of the cell)
We have made an Integrated Readout Layer board for tests of these cells
23 Nov 09
Rubinov, WSU HEP seminar
49
SiD muon system
 For SiD muon system there are 3 main issues
1. Cost
2. Cost
3. Cost
23 Nov 09
Rubinov, WSU HEP seminar
50
Our setup
Detail of optical coupling and
adopter board
using Keithley 2400 for bias
(not shown)
23 Nov 09
Rubinov, WSU HEP seminar
51
MTest 2008
Beam from Nov10 to 16
Minerva test of TOF counters
Added one bar with SiPM for
testing (Ham, IRST)
Using NIM based 6ch amp built
at Fermilab for this work
Using optical coupling designed
at Notre Dame
Using 120 GeV proton beam
(1in x 1in spot)
Very preliminary results below
23 Nov 09
Rubinov, WSU HEP seminar
52
Single PE signals
 Scope traces
5mv/div
using LED
23 Nov 09
Rubinov, WSU HEP seminar
53
Ham-100 during beam spill
Notice the Y scale is 100mv/div!
23 Nov 09
Rubinov, WSU HEP seminar
54
IRST SiPM with 1.8m sint in 120Gev Beam at 34V, I=1.1uA
Notice the Y scale is 100mv/div!
23 Nov 09
Rubinov, WSU HEP seminar
55
Summary of test beam
 If you have enough photons, SiPMs will make
PERFECT muon detectors.
 So the questions are:
 Size of scintillation strip and WLS fiber diameter
(cost)
 Length of strip and WLS fiber (cost)
 Area of the SiPM (coupling the fiber to the SiPM)
(cost)
 Electronics to readout the SiPM – does not drive
the cost
23 Nov 09
Rubinov, WSU HEP seminar
56
Conclusion
We are on a cusp of a revolution in
Low Light Level photo detectors.
The only questions is are we going to be
manning the barricades
23 Nov 09
Rubinov, WSU HEP seminar
57
The Future
 I have seen the future of SiPM readout
 Readout electronics will be integrated into the SiPM!
because
 SiPM is an inherently digital device
 We ALWAYS convert the signal from the SiPM to digital
 So why do we have an analog step in between?!?
2pe
1pe
0pe
2pe
ADC
1pe
0pe
58
The Future

Ingredients required for integrated readout
1. SiPM is CMOS compatible
RMD makes SiPMs through Mosis
2. Will work for in HEP applications
Pixel architectures have demonstrated
readout of arrays like this
3. Cost effective
(in volume)
59
So why
DIGITAL-ANALOG-DIGITAL?
 Because this requires an ASIC
 The people who make SiPMs do not know what we
want
 The people who know what we want do not make
SiPMs (yet)
Application Specific IC has to have a specific application
Because it gives us the most flexibility
60
Back from the future
 Our current strategy is to maximize flexibility
 which is the opposite of what we eventually want
61
Next step(s): 4ch board






Still very generic, but now think infrastructure
Best available commercial components without
heroic efforts (~1ns resolution, ~400 pe range)
Integrated with SiPM specific features
(bias generator, current readback, temp sensor)
Optimized for medium ch count (dozen(s) SiPMs)
Flexible: using 50ohm input, generic daughter
board connection to support faster readout/more
memory
Large FPGA to allow DSP and TDC features
62
Next step(s)

Still very generic, but now think infrastructure
63

Still very generic, but now think infrastructure
CW bias generator
12bit,
250MSP
2 stages of diff amps
S ADCs
simple USB interface
daughter brd
for faster
interface
bias offset/ch
hi res current readback/ch
largish
FPGA
64

Still very generic, but now think infrastructure
CW bias generator
bias offset/ch
hi res current readback/ch
65
Near future
 Move from more generic to more specific
 Develop a simple ASIC
 Optimize for 100s of SiPMs
66