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