1
The first beam test of a monolithic particle pixel detector in high-voltage CMOS technology
Ivan Peric, Christian Takacs, Jörg Behr, Franz M. Wagner, Peter Fischer
University of Heidelberg
This work draws on the results from an ongoing research project
commissioned by the Landesstiftung Baden-Württemberg
2
Introduction
• Monolithic pixel detectors in high-voltage CMOS technology• Main features:• Easy to implement (standard CMOS technology used), radiation hard and
fast• Allow in-pixel signal processing (CMOS)• Can be very thin (thinner than 50 μm)• Possible applications: particle tracking in the case of high occupancy and
harsh radiation environment such as in LHC (upgrade)
3
Introduction
• First test beam results• First irradiation results
FRM II
CERN (SpS)
DESY
4
<-60V
High-voltage monolithic detectors
0V
1. Idea – use high-voltage P/N junctions as sensor
2. Idea – place the (CMOS) electronics inside the N-well
Collection speed Radiation hardness
10 μm tcoll<<100ps
MAPS (as comparison)
High-voltage monolithic detectors
drift
diffusion
5
High-voltage monolithic detectors
1. Idea – use high-voltage P/N junctions as sensor
2. Idea – place the (CMOS) electronics inside the N-well
Collection speed Radiation hardness
MAPS (as comparison)
High-voltage monolithic detectors
drift
Rad. damage
Rad. damage
6
HVD types
RO chip
Binary information
Analog information
Analog information
Type ABinary readout
Type BAnalog readoutRolling shutter
addressing
Type CCapacitive
readout
Similar to 3D detectors!!!
7
HVD types
RO chip
Binary information
Analog information
Analog information
In-pixel signal processing Time measurements possible (fast
readout) Leakage current compensation (+ rad.
hardnes)
Larger pixels Larger capacitance Static current consumption
Smaller pixels Smaller capacitance No static current consumption
Time measurements not possible Leakage current added to signal (- rad. hardnes)
Any kind of in-pixel signal processing possible (hybrid detector)
Radiation tolerant layout can be easily implemented
Slightly increased noise because of capacitive transmission
Testbeam!!!
8
Type A
FFComparatorCR-RC
4-bit tune DACReadout bus
CSA
N-well
AC coupling
Bus driver
3.3 V
-60 V P-substrate
RAM
40 μm 15 μm
220 fF (50 e ENC measured)
9
Type A
0 10 20 3 0 40 50 60
48
50
52
54
56
58
60
62
Response time Thre shold: 720 e Sign al: 1110 e
Noise
DAC sett ing N
ois
e/e
100
110
120
130
140
150
160
Re
sp
ons
e tim
e/n
s
10-60 V P-substrate
Type B
Readout bus
N-well
AC coupling
3.3 V 2 V
ResetNWB
SelB
ResB
9 μm 12 μm
10 fF (90 e ENC measured)
11
Type B
12
Type C
CR-RC
CSA
N-well
AC coupling
3.3 V
-60 V P-substrate
RAM
35 μm 15 μm
100 fF (system: 80 e ENC measured) – sensor 30 e ENC!
Readout chip
13
The “Taki” chip
• 128X128 pixel-matrix – pixel size 21X21µm2
• The chip can be easily scaled to 4 or 16 times larger area
• Fast digital readout – designed for ~50 s frame readout time (164 s tested)
• 128 end-of-column single-slope ADCs with 8-bit precision
• Low power design - full chip 55mW (only analog)
• Radiation hard design
14ADC
Row
-contro
l („Sw
itcher“)
Digital output
10000001000
Pixel size: 21 X 21 m Matrix size: 2.69 X 2.69 mm (128 X 128) Possible readout time/matrix: ~50 s (400ns/row) (tested so far 1.28 s/row) ADC: 8 – Bit Analog power: 54.9mW (7.63mW/mm2) Analog power: ADC: 0.363mW/ADC (90μA+10 μA)
8 LVDS
Counter
Amplifier
Comparator
Latch
Ramp gen.
Chip structure
Pixel matrix
15
ADC
Difference Amplifier S/H Amplifier S/H AmplifierSwitches
ComparatorLogic Counter Pixels
Current source
Ramp
Difference amplifier
Bricked pixels
Guard ring
Switched capacitor amplifier Single slope ADC Asynchronous 8-bit counter
21 um
16
Comp
Amp
Amp
Amp
A1 43
5R1
C2
C3R3Input
1
1
2A2
6
C4
4
3A3
8
C5
5
6
7
C1
7
9
10
1112
C6
2
C1
R2X
R2Y
UX
UY
CX
CY
R1
6X3X
LoadBiasP
VPLoad
0
PDDKS
VInput
0
(Problem with the) difference amplifier
The amplifier oscillates under standard bias conditions
17
Noise (present and future)
Amplifier Noise : 57 eFollower noise meas : 24 e
Reset noise meas: 65 e
Reset noise theory : 42 e
DKS
Better design
Reducing of ENC from 90 e to 30 e is realistic - > all S/N ratios will be increased by factor 3
10 eBetter design
10 e
The ENC is mainly caused by the readout electronics
0 e
18
Test system
FPGA
Bias voltage “generators”
USB
Trigger connector
48 MHz
Power (FPGA)
Very simple detector test system – a single PCB Only 4 external voltages needed, high voltage is generated by batteries USB 1 communication with DAQ PC
Radioactive s.
Power (det.)HV
19
FPGA
Matrix
R
receiver
-
cmp
RAM
cnt
Th
Ampl
Col
Row
&
ID
Wr
Frame
Pedestals
Rd
Pixel detector FPGA PC
From TLURAM
S
receiver
Reset values
Del
DKS mode, frame mode- or zero suppressed cluster readout
Pedestal and reset-offset subtraction
Zero suppresion
Cluster readout
RO FIFO
RO FIFO (frame m.)
20
Test beams with EUDET telescope
Test beam DESY
Test beam CERN
EUDET telescope
DUT
DUT
21
Test beams with EUDET telescope
22
Results – MIP signal
0 1000 2000 3000 4000 5000 60000
1000
2000
3000
4000
5000
6000
7000
Num
ber
of p
ixel
sSignal [e]
1 MSP 2 MSP 3 MSP 4 MSP 5 MSP 6 MSP
60Co Spectrum
MIP spectrum (CERN SpS - 120GeV protons) MIP spectrum (60 Co)
MIP spectrum (CERN SpS - 120GeV protons)The signal increases from 1200 e (single pixel) to 2200 e (6-pixel cluster)The measured S/N ratio varies from 12.3 (single pixel) to 9.8 (6-pixel cluster)
0 5 10 15 20 25 30 35 40 45 500
100
200
300
400
500
600
700
800
seed p ix e l3 M S P
w h o le c lu s te r
S N R
coun
t
23
Results – signal
0 1 2 3 4 5 60
5
10
15
20
25
30
Sig
nal [
AD
U]
Number of pixels
55Fe peak
Comparison between 60Co and 120GeV proton spectra60Co signals higher by 10% - expected from theory due to lower particle energySeed pixel sees about 50% of the total signalThe next MSP sees only 25% of the seed pixel signalCluster size is 6 pixelsModerate charge sharing (the seed gets the most)Do we expect this? – the gaps between n-wells are large, the most of the particles hit the gaps
As comparison 55FeSeed pixel sees about 90% of the total signalCluster size is 3 pixelsNo charge sharing
0 1 2 3 4 5 60
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
2400
Sig
nal [
e]
Number of pixels
Most probable signal (60Co ) Most probable signal (120 GeV p)
24
6 pixel cluster
25
Primary and secondary signal (explanation of the measured spectra)
The drift leads to the primary signal P – this signal portion is not shared between pixels, it is collected in the pixel next to the particle hit pointThe diffusion of the electrons generated in the non-depleted bulk is the secondary signal mechanism
Direct hit Hit between the pixels (occurs quite often)
P
S1
S2S3
P
S1 S2S3
0 1 2 3 4 5 60
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
2400
Sig
nal [
e]
Number of pixels
P
S1
S2
P measured with type A det. – good agreement
26
Is there sharing of primary signal?
Is there sharing of primary signal?Such clusters could be lost after applying the seed cut…
Do we have gaps with zero E-field? (Moreover, they could be insensitive to particles)
27
Fe-55 (explanation of the measured spectra)
No charge sharing A small part of the signal is seen by the next pixel
Seen very seldom Seen very seldom
28
Gap investigation
Do we have gaps with zero E field?
If yes, there will be a certain number of clusters with two equal seedsCOG correction should be then 0.5 pixel size
29
CoG correction distribution in pixel frame of reference
0 0.0020.0040.0060.0080.010.0120.0140.016
0.0180.02
00.002
0.0040.006
0.0080.01
0.0120.014
0.0160.018
0.02
0
50
100
150
200
250
300
350
400
The COG correction distribution is not homogenous inside a pixel due to reduced charge sharing – the small CoG correction values occur more frequently
Large CoG values occur very seldom => there are no sensitive gaps with E=0 but… the gaps could be insensitive
Or the clusters with two equal seedscould be lost after applying the seed cut…
In-pixel CoG coordinate [mm]
Number of clusters
Pixel centre: (0.0105, 0.0105)
30
Efficiency
2.5 3.5 4.5 5.5 6.5 7.5 8.5 9.5 10.5 11.5T
T^ 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Efficiency
Purity
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 20 40 60 80 100 1200
10
20
30
40
50
60
effic iencyeffic iency
pixel coordinate x
pix
el
coo
rdin
ate
y
Efficiency is the answer but…Efficiency is homogenous over the matrix area and saturates at 86% for low seed/cluster thresholds
Efficiency
Seed and cluster cut [SNR]
31
Track and system geometry
Scintillator Scintillator
Telescope planes DUT
32
(Irregular events) double track event
Out of time track – not seen by DUT
In time track – seen by DUT
0 1 2 3 4 5 6 7 8 90
5000
10000
15000
20000
25000
30000
35000
num ber of tracks found per event
cou
nt
0
40000
log
(co
un
t)1
0
1
2
3
4
T [μs]T (trigger) 0 160 800
“Mimotel”
“taki”
Readout times
33
(Irregular events) empty event
In time track – not seen by Mimotel
0 1 2 3 4 5 6 7 8 90
5000
10000
15000
20000
25000
30000
35000
num ber of tracks found per event
cou
nt
0
40000
log
(co
un
t)1
0
1
2
3
4
T [μs]T (trigger) 0 160 800
Mimotel
“taki”
Readout times
34
(Irregular events) double track event seen as a single track event
Out of time track – not seen by DUT In time track – not seen by Mimotel
0 1 2 3 4 5 6 7 8 90
5000
10000
15000
20000
25000
30000
35000
num ber of tracks found per event
cou
nt
0
40000
log
(co
un
t)1
0
1
2
3
4
T [μs]T (trigger) 0 160 800
Mimotel
“taki”
Readout times
35
Efficiency (conclusions)
• Efficiency lower than 100% probably due to timing issues
– Readout of telescope and DUT are not synchronous
– DUT integration (readout) time 164 μs
– Telescope integration time = 800 μs
– Large cluster and track multiplicity in telescope
– multiple tracks in telescope due to high beam intensity and long integration time
– Small cluster multiplicity in DUT due to shorter integration time
• Some “out of time” particles hit the telescope after the trigger moment (during the readout) – the particles are not seen by the DUT due to wrong timing
• Neglecting of all multiple track events increases efficiency from 72% to 86%
• Problem: A part of scintillator outside the telescope area: some out of time tracks are seen as single tracks by telescope. If we were able to filter these out of time tracks too, we would probably measure a better efficiency
36
In-pixel measurements – back-propagation
Alignment
Back-propagation
Excellent spatial resolution of the EUDET telescope allows the investigation of DUT properties as function of the in-pixel hit pointWe performed series of such n-pixel measurementsThe fitted coordinate is back-propagated to the DUT frame of reference and DUT pixels frame of reference
37
In-pixel CoG
0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020-0.006
-0.004
-0.002
0.000
0.002
0.004
TEL: fitted hit position in y [m m ]
• CoG correction works but the slope is too small (by factor ~ 3) probably due to absence of charge sharing (primary signal) and noise (Eta-correction does not lead to better results)
• Good check of the back-propagation tool
In-pixel position
Pixel centre: 0.0105 mm
38
In-pixel efficiency
0.0000.005
0.0100.015
0.020
0.0000.005
0.0100.015
0.020
0.00.2
0.40.60.81.0
x[mm]y[mm]
efficiency
There are no insensitive regions! => There are no E=0 gaps!
In-pixel position
Pixel centre: (0.0105, 0.0105)
39
Spatial resolution
DUTshiftXEntries 3326Mean -0.0006307
RMS 0.009879
0 0.10
10
20
30
40
50
60
70
80
90DUTshiftX
Entries 3326Mean -0.6307 mSigma fit 8.64 m
Measured - fitted X position
[mm]
DUTshiftYEntries 3326Mean -6.423e-05RMS 0.00787
0 0.10
20
40
60
80
100
DUTshiftYEntries 3326Mean -6.423e-02 mSigma fit 7.28 m
Measured - fitted Y position
[mm]
• Spatial resolution
• Sigma residual X: 7.3 μm
• Sigma residual Y: 8.6 μm
• The difference is probably caused by the bricked pixel geometry – still not understood completely, simulations will be done
• The spatial resolution is not as good as in the case of standard MAPS due to absence of charge sharing in the case of primary signal
• It is not completely clear why is the resolution worse than 21 μm /sqrt(12) = 6.1 μm
• The residual is sometimes larger than the pixel pitch.
40
Seed pixel – fitted hit point mismatch
• The back-propagation (in pixel measurements) show that the fitted hit point (measured by the telescope) is sometimes outside (in the next pixel) of the seed pixel.
• This mismatch worsens the spatial resolution• The fitted hit point – seed pixel mismatch occurs more probably when fitted point is near the pixel
boundary• The mismatch seems, however, not to be caused by the electronic noise
DUT seed
x [m m ]y [m m ]
predicted seed
telescopefits
41
Seed pixel – fitted point mismatch (clusters and their pixel S/N)
2.36.1
14.0496.0
2713
6.44.2
4.2150.4
0.81.3
2.53.4
0.0111.2
1.3-2.5
3.23.8
3.4303.0
3.50.6
2.92.2
3.4552.8
1.93.8
3.22.5
1.6274
3.93.2
105.1
4.6371.6
0.22.2
-0.81.3
1.8282.7
2.13.1
6.5-0.2
4.020-0.3
1.31.0
0.92.4
4.6231.4
0.70.2
2.71.4
4.0174.8
1.80.6
3.3-0.8
0.7332.0
5.05.9
A few clusters when the fitted point is outside the seed pixels are shown – the seed pixel amplitude (S/N amplitude) is always very high – there is little chance that we have chosen the wrong seed due to electronic noise.
42
fittedreal track
Seed pixel – fitted point mismatch
The mismatch seems to be caused by the measurement-setup uncertainties, e.g. mechanical instability, multiple scattering on PCB “vias”.
1.3mm
Cu
predicted error
PCB
Via
Chip
43
Summary (test beam)
• Efficiency: 86%
• Purity: 72%
• Sigma X-residual 8.6 μm
• Sigma Y-residual 7.3 μm
• S/N ratio seed: 12.3
• S/N ratio cluster (6 pixels): 10
• There is little charge sharing – the seed pixel receives 50 % or more of the total signal
• There are no insensitive regions
• Spatial resolution worse than expected probably due to MS, to be understood
44
Irradiation
Irradiation with neutrons has been performed by Franz M. Wagner at FRM II - “Forschungsneutronenquelle Heinz-Maier-Leibnitz” http://www.frm2.tum.de/
45
Irradiation with neutrons (1014 neq) – signal (Type A)
0 1000 2000 3000 4000 5000 60000
50
100
150
200
250
300
350
Num
ber
of h
its
Signal [e]
1 MSP 2 MSP 3 MSP 4 MSP 5 MSP 6 MSP
0 1 2 3 4 5 60
200
400
600
800
1000
1200
1400
1600
1800
2000
2200
2400 60Co Irradiated chip (1014 n
eq)
Not irradiated
Sig
nal [
e]
Number of pixels in cluster
IrradiatedNot irradiated
After irradiation to 1014 neq the seed “MIP” (most probable 60Co) signal is 1000 e and the cluster signal is 1300 e – the real MIP is by about 10% lowerThe measurement has been performed at 0CLeakage current / pixel increases from 350 fA to 130 pA
46
Irradiation with neutrons (1014 neq) – noise (Type C)
0.00 25.00m 50.00m 75.00m 100.00m 125.00m 150.00m0
100
200
300
400
500
Num
ber
of h
its
Signal [V]
Irradiated chip (1014 neq
) 55Fe spectrum, -30C
0.00 25.00m 50.00m 75.00m 100.00m 125.00m 150.00m0
200
400
600
800
1000
1200not irradiated chip 55Fe spectrum, -30C
Num
ber
of h
its
Signal [V]
Irradiated to 1014 neq, - 30C, noise about 30 e
Irradiated to 1014 neq, room temperature, noise about 60 e
0.00 25.00m 50.00m 75.00m 100.00m 125.00m 150.00m0
200
400
600
800
1000
Num
ber
of h
its
Signal [V]
Not irradiated chip, 55Fe spectrum, room temperature
0.00 25.00m 50.00m 75.00m 100.00m 125.00m 150.00m0
500
1000
1500
2000
2500
3000
Num
ber
of h
its
Signal [V]
Irradiated chip (1015 neq), 55Fe spectrum, room temperature
Type C detector, 55Fe spectrum Excellent noise performance after irradiation No clustering possible with this detector
47
Summary
Type B (10 fF detector capacitance, 21 μm x 21 μm pixel size)
• Signal:
– not irradiated: 1200 e (seed) to 2200 e (cluster) (MIP)
– Irradiated to 1014 neq : 1000 e (seed) to 1300 e (cluster) (60Co)
• Noise: 90 e (not irradiated) – the high noise is the result of non-optimal design, will be reduced by new design (the chip has already been submitted)
• Type A (220 f detector capacitance, 55 μm x 55 μm pixel size)
• Signal:
– Not irradiated: 1700 e (MIP) (good agreement with type B)
– Noise 55 e at 110 ns shaping time
• Extrapolations for type A:
• Signal after 1014 : 1200 e (MIP)
• Signal after 1015 : 800 e
• Type C (100 f detector capacitance, 50 μm x 50 μm pixel size)
– Noise after 1014 neq : 60 e (longer shaping times) (room T)
– Radiation hardness of more than 2 MRad tested
• Future plans: irradiation to at least 1015 neq and 50 MRad
48
Thank you