High-Voltage Pixel Sensors for ATLAS Upgrade

Ivan Peric, 9th Hiroshima Symposium, Hiroshima, 2013 1

High-Voltage Pixel Sensors for ATLAS Upgrade

Ivan Pericfor HVCMOS collaboration

University of Heidelberg, Germany

Ivan Peric, 9th Hiroshima Symposium, Hiroshima, 2013

HVCMOS detectors

2

• HV CMOS detectors (particle detectors in standard HV-CMOS technologies) are depleted active pixel detectors

• Main charge collection mechanism is drift (certain signal part is collected by diffusion as well)• Implemented in commercial CMOS (HV) technologies (350nm and 180nm)

PMOS NMOS

p-substrate

Depletion zone

Potential energy (e-)

deep n-well

Drift

Diffusion


HVCMOS detectors

3

• Collection electrode is a deep-n-well in a p-substrate• Pixel electronics is embedded in the n-well (PMOS: directly, NMOS in a P-well)• Can be implemented in many commercial technologies (we tried also 65nm UMC CMOS);

however the possibility to bias the n-well with a relatively high voltage is important• Best properties offer HV CMOS technologies – the n-well is deep enough so that reverse

voltages of up to ~120V can be used (no punch through between p-well and substrate)

p-substrate

Depletion zone


deep n-well

Drift

Diffusion


HVCMOS detectors

4

• Example for AMS: 20/10 cm (350/180nm CMOS) substrate resistance -> acceptor density ~ 1015 cm-3

• Depleted layer thickness estimation from the technology datasheet (area capacitance) for 60V bias (120 max): 10µm (350nm), 7µm (180nm)

• Typical measured MIP signal for a 50 µm x 50 µm pixel in AMS 0.35 µm (60V bias): 1800e (we estimate about 800e from depleted region and about 1000e by diffusion)

p-substrate

Depletion zone


deep n-well

Drift

Diffusion


HVCMOS detectors

5

• Example for AMS: 20/10 cm (350/180nm CMOS) substrate resistance -> acceptor density ~ 1015 cm-3

• Depleted layer thickness estimation from the technology datasheet (area capacitance) for 60V bias (120 max): 10µm (350nm), 7µm (180nm)

• Typical measured MIP signal for a 50 µm x 50 µm pixel in AMS 0.35 µm (60V bias): 1800e (we estimate about 800e from depleted region and about 1000e by diffusion)

Seed:drift+

diffusiondiffusion

diffusion

diffusion

diffusion

diffusion

diffusion


Development in 350nm Technology

6

• Two development periods: 1) general development and 2) applications • In 1) we used AMS 0.35µm technology• Several prototypes have been designed• Three detector types: • A) Monolithic detector with intelligent CMOS pixels• Pixel electronic is rather complex – CMOS based charge sensitive amplifier, usually discriminator,

threshold tune…• B) Monolithic detector with 4-PMOS-transistor pixel and rolling shutter RO• C) Capacitively coupled hybrid detectors• Good results, >98% efficiency in test-beam, high radiation tolerance


Development in 350nm Technology

7

• Two development periods: 1) general development and 2) applications • In 1) we used AMS 0.35µm technology• Several prototypes have been designed• Three detector types: • A) Monolithic detector with intelligent CMOS pixels• Pixel electronic is rather complex – CMOS based charge sensitive amplifier, usually discriminator,

threshold tune…• B) Monolithic detector with 4-PMOS-transistor pixel and rolling shutter RO• C) Capacitively coupled hybrid detectors• Good results, >98% efficiency in test-beam, high radiation tolerance


Test-Beam Results

Simple (4T) integrating pixels with pulsed reset androlling shutter RO21x21 µm pixel size

Seed pixel SNR 27, seed signal 1200e, cluster 2000e

Spatial resolution 3-3.8µm

Efficiency vs. the in-pixel position of the fitted hit.Efficiency at TB: ~98% (probably due to a rolling shutter effect)


Development for Mu3e Detector

9

• The first applications of HVCMOS detectors will be the Mu3e experiment at PSI and the luminosity monitor for Panda experiment (GSI)

• 180nm HVCMOS technology chosen due to lower power consumption• Low particle energy, thin detector required => monolithic pixel detector, thinned to 50µm• Pixels contain only CSAs, every pixel connected to its readout cell, placed at the chip periphery,

by an individual wire• The concept is feasible for large pixels (80µm x 80µm)• Advantages: minimal pixel capacitance, optimal SNR, separation of digital and analog circuits• Disadvantage: inactive periphery (about 5%)• Collaboration: Heidelberg PI and ZITI, PSI, ETH und University Zürich, University Geneva

Readout cells


Development for Mu3e Detector

10

• The first applications of HVCMOS detectors will be the Mu3e experiment at PSI and the luminosity monitor for Panda experiment (GSI)

• 180nm HVCMOS technology chosen due to lower power consumption• Low particle energy, thin detector required => monolithic pixel detector, thinned to 50µm• Pixels contain only CSAs, every pixel connected to its readout cell, placed at the chip periphery,

by an individual wire• The concept is feasible for large pixels (80µm x 80µm)• Advantages: minimal pixel capacitance, optimal SNR, separation of digital and analog circuits• Disadvantage: inactive periphery (about 5%)• Collaboration: Heidelberg PI and ZITI, PSI, ETH und University Zürich, University Geneva

Readout cells0 500 1000 1500 2000

0,0

0,2

0,4

0,6

0,8

1,0mu 956.0343esig 64.28302e

Sig

nal f

ract

ion

Input referred threshold [e]

VN 20, VNFoll 2


Development for ATLAS

11

• Also the use in HL LHC ATLAS upgrade is investigated• Concept: The use of active HVCMOS sensors as replacement for the standard strip- and pixel-

sensors and the use of existing (or slightly modified) readout ASICs• Group of pixels connected to one readout channel, address information is coded as signal

amplitude• Realization: one pixel contains: CSA, comparator, threshold tune circuit and the address

generator• Address signals of the grouped-pixels are summed and connected to the input of the RO-channel• Collaboration: CPPM, CERN, Universities of Geneva, Bonn, Göttingen, Glasgow, Liverpool,

Heidelberg, LBNL,…

ROC

A


Pixel Readout

12

• Pixel readout: three pixels connected to one readout channel of the ATLAS FEI-chip (FEI4)• Capacitive sensor-to-chip signal transmission, no need for bump bonds• Advantages: smaller pixels, different pixel geometries can be combined with one ASIC (e.g. for

the end caps), little material, fast readout, good resolution for large incident angles

Glue

Pixel readout chip (FE-chip)

Pixel CMOS sensor33x 125 μm

Summing lineTransmittingplate

Pixel electronics based on CSA

Bump-bond padCoupling

capacitance


Strip Readout

13

• Strip readout: larger number of pixels (e.g. 100) grouped into segmented strips, readout with an amplitude sensitive strip-readout chip (multichannel chip)

• Advantages: Pixel detector (nxn pixels) is readout with a relatively small number of analog channels (~n) – in contrast to rolling shutter readout, time resolution is high

• Less material than in the case of the hybrid pixel detector and a similar time resolution. • If summing scheme can cope with two simultaneous hits, the concept can work at relatively high

occupancies (e.g. 8 particles / cm2 / 25ns )

Summing line

Every pixel generates unique current


Experimental Results

14

• Results of the project• A small detector prototype chip “CCPD” has been designed• CCPD can be readout with both a strip- and a pixel-readout chip• Stand-alone readout is also possible• Two chip iterations• 1) optimized for small noise• 2) optimized for radiation tolerance

2

3

1

2

3

1

CCPD Pixels



15

• Results of the project• A small detector prototype chip “CCPD” has been designed• CCPD can be readout with both a strip- and a pixel-readout chip• Stand-alone readout is also possible• Two chip iterations• 1) optimized for small noise• 2) optimized for radiation tolerance



16

• Three testing programs:• 1) Test in standalone modus: a) lab tests with electric signals (using charge injection circuit) and

b) measurements with radioactive sources. Goals: functionality tests, measurements of noise, threshold dispersion, and the MIP signal amplitude

• 2) Irradiations• 3) Tests with pixel readout chip (it works - three addresses can be distinguished, first testbeam

measurement done, time stamp distribution ok => good time resolution)• 4) Tests with strip readout chip (still to be done)

2

3

1

2

3

1

CCPD Pixels

Signal amplitudes measured by FEI4


Standalone Tests

17

• Test in the standalone mode:• Pixel addresses connected to a monitor line that can be accessed from outside via single IO pad• Several CSA outputs can be measured directly – allows spectral measurements

Monitor line

Analog-multiplexer

pixels


MIP Signal

18

• Several CSA outputs can be measured directly – allows spectral measurements• Measured Sr-90 MPW signal at rather low 30V bias voltage (maximal 120V) ~1350e (we estimate

400e from depleted region at 30V – diffusion part 950e)• Estimated MIP signal for 60V bias: 1500e

Sr90:1400 e

Fe55:1660 e


Noise and Threshold Dispersion

19

• Threshold and injection scans – noise, threshold dispersion• Results for CCPD2 optimized for radiation hardness (not for low noise)• Average pixel noise ~ 75e (large spread)• Threshold tuning: dispersion ~ 25e• Estimated MIP signal at 60V: 1500e

40 60 80 100 1200

5

10

15

20

25

30

Pix

el c

ount

Noise [e]

Simple pixels

800 820 840 860 880 900 920 940 960 980 10000

20

40

60

80

100

120

Pix

el c

ount


Mean: 891eSigma: 24e

Noise distribution Threshold dispersion



20

• Average pixel noise ~ 75e (large spread)• Threshold tuning: dispersion ~ 25e• Estimated MIP signal at 60V: 1500e• Required:• 6 x SD(Noise) + 6 x SD(Threshold) = Smallest signal• 6 x SD(Noise) + 6 x SD(Threshold) = 600e• Question: what is the smallest signal for a MPW of 1500e? (probably ~ 1500/2 = 750 e)

20

Smallest signal

Smallest signal ~ 6(SD(Noise) + SD(Threshold))

Noise Threshold dispersion

MP

W

Landau distribution

Base line

Mean Th

1500e~750e



21

• Average pixel noise ~ 75e (large spread)• Threshold tuning: dispersion ~ 25e• Estimated MIP signal at 60V: 1500e• Required:• 6 x SD(Noise) + 6 x SD(Threshold) = Smallest signal• 6 x SD(Noise) + 6 x SD(Threshold) = 600e• Question: what is the smallest signal for a MPW of 1500e? (probably ~ 1500/2 = 750 e)• In theory ok, but we still need to improve threshold tuning, so far we achieved a mean value of

~800e, 400e is required

21

Smallest signal

Smallest signal ~ 6(SD(Noise) + SD(Threshold))

Noise Threshold dispersion

MP

W

Base line

Mean Th

1500e~750e

800 820 840 860 880 900 920 940 960 980 10000

20

40

60

80

100

120

Pix

el c

ount


Mean: 891eSigma: 24e

Landau distribution


Irradiations – older Results

22

• Irradiation studies:• Two damage mechanisms: nonionizing and ionizing• Results are generally promising, but we still do not have the results from a test-beam

measurement with irradiated devices• Older results (AMS 0.35µm technology)• X-ray irradiation up to 60 Mrad (rad-hard device layout – enclosed transistors, chip on during

irradiation) – increased noise and leakage current observed - after annealing and cooling they return to normal noise


Irradiations – older Results

23

• Older results:• Proton irradiation to 1015 neq/cm2 (standard device layout, chip off during irradiation) – increased

leakage and noise – the MIP signal does not decrease significantly – diffusion still works? • Neutron irradiation to 1014 neq/cm2 (rolling shutter chip) – increased leakage and noise – diffusion

part of the signal is decreased

0 1 2 3 4 5 60

200400600800

10001200140016001800200022002400

60Co Irradiated chip (1014 neq

) Not irradiated

Sig

nal [

e]

Number of pixels in cluster

Irradiated

Not irradiated

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80.0

0.2

0.4

0.6

0.8

1.0

~num

ber o

f sig

nals

signal amplitude [V]

22Na - 0V bias (0.075V or 1250e) 22Na - 30V bias (0.18V or 3125e) 22Na - 60V bias (0.22V or 3750e) 55Fe - 60V bias (100mV or 1660e) RMS Noise (2.4mV or 40e)

Temperature: - 10CIrradiated with protons to 1015n

eq

Proton irradiation Neutron irradiation


Irradiations – CCPD1

24

• 1) Two sets of detectors have been irradiated to 435 Mrad and 80 Mrad with protons at the PS (CERN) (chips on during irradiation)

• 2) X-ray irradiation to 50 Mrad (chips on during irradiation)• 3) Neutron irradiation to 1016 neq/cm2 (chips off during irradiation, only nonionizing damage)• Influence of ionizing radiation higher than expected. Despite of that, Sr-90 spectrum can be

measured after 80Mrad (proton irradiation)

CCPD1 irradiated to 80 Mrad with protonsSr-90 spectrum

CCPD1 at 380 Mrad (81015 neq) proton-irradiationBeam signals



25

• Chips were affected by x-ray irradiation (ionizing) strongly - large amplifier gain drop• The chip irradiated to 435 Mrad works (responds to test signals), but particle signals can not be

distinguished from noise after about 380 Mrad (gain drop too high – high threshold, large leakage, activation, cooling not possible)

CCPD1 irradiated to with protonsCount rate

CCPD1 irradiated with x-raysAmplifier gain loss

Initial rate

Wrong settings

Better settings

60V bias

30V bias

Better settingsAnother beam position



26

• The detector irradiated with neutrons (1016 neq/cm2) works (capacitively readout by FEI4), particles can be clearly detected at the room temperature, testbeam measurement has been done and will give us the rough estimation about the efficiency (setup is not optimized – e.g. no threshold tuning done)


CCPD2: X-Ray Irradiation to 862 Mrad

27

• Several weak points in design have been identified that cause CCPD1 to be susceptible to ionizing radiation (symptoms are: gain drop and base line shift)

• The weak pints have been fixed in CCPD2 (at expense of a slightly higher noise)• CCPD2 implements three pixel types, fully rad hard, partially rad hard and a simple pixel that

uses positive feedback and has a CMOS comparator• A detector has been irradiated to 862 Mrad with x-rays. (chips on during the irradiation, 2 hours of

annealing at 70C after each 100Mrad)• Result for one partially rad hard pixel: input referred noise before irradiation 25mV (90 e)• Input referred noise after irradiation 40mV (150 e) at room temperature• We observe that amplifiers work with reduced bias current (2µA instead of 5µA) – probably only

partially rad hard pixels are affected – bias NMOS diode can be affected by oxide charge

862 Mrad90e 150e



28

• The noise increase can be addressed to• 1) Gain drop (by factor of two for the pixel)• 2) Bias current drop (2µA instead of 5µA per amplifier) (under this condition we would have only

48 mA preamp current consumption per cm2 detector area)• 3) HV leakage current



29

• Sr-90 spectra have been recorded before and after irradiation - no sign of signal loss at sensor• 1V Injection (5000 e): 430 mV

• Sr-90 spectrum

2500 e



30

• Several effects are still not understood• The cause of the gain drop• Several possibilities:• Observed drop in the amplifier bias current• Possible decrease of the feedback resistance, due to ionizing damage in the feedback transistor

=> shaping time decrease• Notice that the fully rad hard pixels are not significantly affected

10 days annealing

Fully rad-hard pixels Partially rad-hard pixels

0 200 400 600 800 1000 12000

20

40

60

80

100

120

140

160

180

200

220

240

260 Pixel1 Pixel2 Pixel3 Pixel4

Am

plitu

de [m

V]

Dose [Mrad]

Re-optimization of the settingsDose 862 Mrad

0 200 400 600 800 1000 12000

50

100

150

200

250

300

350

400


Am

plitu

de [m

V]

Dose [Mrad]

2h 70C annealing

Re-optimization of the settingsDose 862 Mrad


-50


31

• Several effects are still not understood• The origin of HV leakage current• The current gets higher for lower (!) n-well voltage• Parasitic PMOS? trapping of electrons in SiO2?• Injection of holes into n-well and their flow to p-substrate?• Tunneling of trapped holes from SiO2 to substrate?

0 200 400 600 800 10000,0

0,5

1,0

1,5

2,0

2,5

3,0

3,5

Leak

age

curr

ent p

er p

ixel

[nA

]

Dose (Mrad)

10 days annealing

2h 70C annealing

0V

1.8V

-60V

-60V

e-e-

Possible cause of leakage current?



32

• The radiation hardening measures done for CCPD2 seem to be successful

CCPD1 irradiated with x-raysAmplifier gain loss

0 200 400 600 800 1000 12000

20

40

60

80

100

120

140

160

180

200

220

240


Am

plitu

de [m

V]

Dose [Mrad]

CCPD2 irradiated with x-raysAmplifier gain lossRad hard pixels

862 Mrad

862 Mrad

862 Mrad


Segmented Strip Measurements

33

• Setup



34

• Strip measurement circuit

Amp

Monitor

Chip

Oscilloscope

Th1

Fe55

Absorber



35


Amp

Monitor

Chip

Oscilloscope

Th1

0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,80

100

200

300

400

500

600

700

800

Cou

nts

Measured voltage [V]

Analog addresses

Fe55

Absorber



36


Amp

Monitor

Chip

Oscilloscope

Th1

0,0 0,2 0,4 0,6 0,8 1,0 1,2 1,4 1,6 1,80

100

200

300

400

500

600

700

800

Cou

nts

Measured voltage [V]

Analog addresses

10

2

4

6

8

10

Pixel Column

Pix

el R

ow

0

137,5

275,0

412,5

550,0

687,5

825,0

962,5

1100

Fe55

Absorber


Conclusion

37

• We are investigating the use of HVCMOS detector for HLLHC ATLAS upgrade• Test detectors CCPD1 (rad soft design) and CCPD2 (rad hard design) work• MIP signal (1500 e), noise (75e) and threshold dispersion (25 e) values are good enough for

efficient detection, however threshold tuning still have to be improved • CCPD1 has been irradiated with x-rays, protons and neutrons, it is affected by ionizing radiation

stronger than expected, however operation up to ~80Mrad is possible• CCPD2 has been irradiated to 860Mrad with x-rays, it works, noise doubled at room T• The noise increase can be mitigated by cooling and design optimization• Irradiations of CCPD2 with neutrons and protons are planned• Operation after 1015-16 neq/cm2 could be possible if the diffusion signal is not entirely lost after

these fluencies• Plans for the next small test-detector• Optimization of noise by increasing feedback resistance, bias current, etc.• Optimization of pixel geometry• Design and production of a larger test-detector (e.g. 1 cm2) planned for 2014


Properties of HVCMOS Detectors

38

• Good properties:• Fast charge collection (field ~ 6-8.5V/µm, collection time ~ 100ps)• High radiation tolerance• Thinning is possible (active region several 10µm at the surface)• Relatively cheap due to the use of a commercial process (1.5 kEUR / 8inch wafer)• Disadvantages:• Small depleted region, relatively small primary- (drift collected) signal, pixel capacitance ~100fF

for larger pixels• We expect that the drift-collected signal does not decrease with irradiation, the question is how

much of the diffusion part remains• SNR can be improved using the charge sensitive amplifier at the cost of increased power• Main challenges: achieve good detection efficiency and low time walk for a given power budget• Simulation example for 30µm x 125µm pixel: a good SNR and a time walk of about 10ns can be

achieved at the power consumption of about 100mW/cm2

• Some limitations arise from the fact that the electronic is placed inside the collecting electrode• Additional capacitance, crosstalk• Solution: the use of simplified pixel electronics


Thank you!


Backup Slides


Pixel electronics (1)

A

D

CCPD bus

Strip bus

4-bit DAC

(CR filter)

Programmable currentG

G

In<0:3>RW

SFOut

Cap. Injection

Amplifier

Filter

Comparator Output stage

CCPD electrode

BL

Th

41

Circular devices

Circular devices


Simple Pixel

AGnd

Vdd

1.81.8

ThP

0

Positive FB

VNSF

BL

(CR filter)

BLR

ThR

4 3 2

StripIn StripOut

HB

Sel

SelABufABuf

OutBLOutSF

OutAmp

OutDisc

OutDisc2

OutDisc3

Discriminator


876

11109

Standard pixel

43

543

EnR

InR(3:0)

EnL

InL(3:0)E

n(5:0)

210

sost

EnL

EnR

EnL

EnR

st so

EnR

EnL

EnL

EnRL0 R0 L1 R1 L2 R2 Str Ld(0:2) dc ao

PL PR

Col(0:2) Col(3:5)

row0(R

),row1(L)

row2(R

),row3(L)

En(11:6)

En(5:0)

ampout

ampoutL0 R0

ao0 ao1

EnL/R=1 - enables CCPD, disables hitbus/strip

PL

dc

str

monitor


876

11109

Simple Pixel

44

543

EnR

InR(3:0)

EnL

InL(3:0)E

n(5:0)

210EnL

EnR

EnL

EnREnR

EnL

EnL

EnRh0 h1 h2 nu S1 S0 Str Ld(0:2) dc ao

PL PRPL

dc

Col(0:2) Col(3:5)

row0(R

),row1(L)

row2(R

),row3(L)

En(11:6)

En(5:0)

monitor

ampout

ampouth0

ao1

EnL/R=1 – enables hitbus; strip and CCPD are always on

str

PR

dch1 h2

h0 h1 h2 sl sr

0

24

3

51

ao0S(1:0)S(3:2)


876

11109

Simple Pixel

45

53

EnR

InR(3:0)

EnL

InL(3:0)E

n(5:0)

210EnL

EnR

EnL

EnREnR

EnL

EnL


PL PRPL

dc

Col(0:2) Col(3:5)

row0(R

),row1(L)

row2(R

),row3(L)

En(11:6)

En(5:0)

monitor

ampout

ampout

ao1


str

PR

dc

h0 h1 h2 sl sr

ao0S(1:0)S(3:2)

out

in

in

out

in

out

out

in

sl

in outhit

4

out

in

in

out

0

2

2

4

4

6


876

11109

Simple Pixel

46

53

EnR

InR(3:0)

EnL

InL(3:0)E

n(5:0)

210EnL

EnR

EnL

EnREnR

EnL

EnL


PL PRPL

dc

Col(0:2) Col(3:5)

row0(R

),row1(L)

row2(R

),row3(L)

En(11:6)

En(5:0)

monitor

ampout

ampout

ao1


str

PR

dc

h0 h1 h2 sl sr

ao0S(1:0)S(3:2)

out

in

in

out

in

out

out

in

sr

in outhit

4

out

in

in

out

0

2

2

4

4

6

Documents

High-Voltage Pixel Sensors for ATLAS Upgrade