PhD Thesis defense Michal Szelezniak ULP, Strasbourg 25 February 2008

Development of pixel detectors with integrated signal processing for the

Vertex Detector in the STAR experiment at the RHIC collider

PhD Thesis defenseMichal SzelezniakULP, Strasbourg25 February 2008

Michal Szelezniak - PhD thesis defense - 25 February 2008

22

Outline

The new vertex detector for the STAR experiment Development of Monolithic Active Pixel Sensors

(MAPS) at IPHC MAPS prototype for PIXEL detector 3-sensor telescope system with prototype readout

for PIXEL detector Future development plans Summary and Conclusions

Development of pixel detectors with integrated signal processing for the Vertex Detector in the

STAR experiment at the RHIC collider


33

STAR experiment

(a) (b) (c) (d)

End view of tracks registered by the STAR TPC in a heavy-ion collision

STAR was constructed to study Quark-Gluon Plasma created in heavy-ion collisions at Relativistic Heavy Ion Collider (RHIC)

a) Lorentz contracted ions before the collisionb) Hard interactions between partons of incoming nucleic) New, high-density state of matter (QGP?)d) Hadronization and freezout

Location of the new vertex detector


44

Penetrating probes (created early in a collision) are sensitive to the evolution of the medium

– Particles with very high transverse momentum– Heavy particles containing charm or bottom quarks

To study next:– Charm flow to test thermalization of light quarks at RHIC– Charm energy loss to test pQCD in a hot and dense medium at RHIC

QGP in heavy-ion collisions

(from HFT proposal)

The D0 signal, after topological cuts, is shown by the solid black circles.

The original spectrum, before software cuts, is shown by the line of open circles.


55

HFT: new vertex detector for STAR

– Goal: increasing pointing resolution from the outside in

– TPC pointing resolution at the SSD is ~ 1 mm– SSD pointing at the IST is ~ 300 µm – IST pointing at the PIXEL is ~ 250 µm – PIXEL pointing at the VTX is ~ 30 µm

To measure heavy flavor production it is necessary to measure charm and bottom hadrons through direct topological reconstruction

New Vertex Detector is needed!

D0 (cū)

Heavy Flavor Tracker

~100 µm

Secondary vertex

PIXEL at 2.5 and 8 cm

Primary vertex

IST at 14 cm

SSD at 23 cm

PIXEL: spatial resolution < 10 μmradiation length ~ 0.3 %

VXD3 0.4%, ALICE pixel detector ~1%


66

PIXEL DetectorPIXEL characteristics: Two layers at 2.5 & 8 cm radius

– 10+30 ladders– 10 sensors/ladder

Nearly 164 M pixels 0.28 % radiation length/layer Air cooled

Quick extraction and sensor replacement Monolithic Active Pixel Sensors

– Thinned to 50 μm thickness– 30 μm x 30 μm pixels– 640 x 640 pixel array– Integration time <200 μs at L=8×1027

– Power disspation <100 mW/cm2

Ladder with 10 MAPS sensors MAPS

RDObuffers/drivers

4-layer kapton cable with aluminium traces


77


(MAPS) at IPHC– Simulations and tests of in-pixel voltage amplifiers,– Tests of advanced pixel structures with in-pixel memories– Tests and study of AC coupling for in-pixel amplifiers– Tests and study of MAPS operated in current mode

(PhotoFET)

MAPS prototype for PIXEL detector 3-sensor telescope system with prototype readout for

PIXEL detector Future development plans Summary and Conclusions




88

Monolithic Active Pixel Sensors

Properties:

Standard commercial CMOS technology

Sensor and signal processing integrated in the same silicon wafer

Signal created in low-doped epitaxial layer (typically ~10-15 μm)

Charge collection mainly through thermal diffusion (~100 ns), reflective boundaries at p-well and substrate

Charge sensing in n-well/p-epi junction

100% fill-factor High granularity Low power dissipation Substantial radiation tolerance Thinning available as standard

post-processing Only NMOS transistors inside pixels

MAPS pixel cross-section (not to scale)

Thin active volume → MIP signal limited to <1000 electrons

Thermal diffusion → cluster size of ~10 pixels (20-30 μm pitch)

sensitivity to charge of a few tens of electrons ← noise at the level of 10 e-


99

MAPS vs. other technologies

High granularity (several μm pitch) Small material budget Fast readout Radiation tolerance

8” wafer with MAPS prototypes

MAPS Hybrid Pixel Sensors CCD

+ - +

+ - +

+ ++ -

+ ++ -

Hybrid Pixel Sensors:

detector bump bonded to readout chip

CCD:

integrated detector and readout, external processing

MAPS:

integrated detector/readout/processing


1010

Simple pixel architectures

GND

VDD VDD

select

output

outputin equilibrium

time

chargecollection

a) b)

chargecollectingdiode

reset

GND

VDD VDD

select

output

reset

output

time

chargecollection


VDD

Continuous reverse bias (self-biased)

Classical diode with reset

Reset noise, offset

No reset noise, no offset

read

read


1111

Pixel sensor architectures

Typical sensor readout– Raster scan – Charge integration time = array

readout time– Multiplexed sub-arrays to decrease

integration time

Column parallel readout architecture– All columns readout in parallel and then

multiplexed to one output– Charge integration time = column

readout time

On-chip signal processing requires high S/N – signal amplification is needed

Analog readout – simpler architecture but ultimately slower readout

Digital readout – offers increased speed but requires on-chip discriminators or ADCs


1212

Example of a simple in-pixel amplifier Amplifier in cascode configuration (only NMOS transistors)

VDDA

GND

Vin

GND

VDDA

Vout

Vcascode

power_on

Typical gain: 4-6

Switches for switched-power operation

Cascode transistor to reduce the Miller effect that is present in a common-source configuration: Cin = Cgs + Cgd(1+G)

Lower input capacitance higher charge-to-voltage conversion factor

Typical biasing voltage: ~0.7 V 1mg

2

1

m

m

g

gG

2mg

Typical power consumption (3.3 V)P=20 μW

(0.35 μm CMOS process)


1313

Optimization of pixel design

vbiasbiasingdiode


vbias

gain gain

a b

out out

Typical connection AC-coupling

Compact layout implementation of AC coupling

Improves CCE (5%) Degrades ENC (25%)

DC coupling gives better ENC performance


1414

Investigated in-pixel amplifiers

E.g. memory discharge time:

MOSFET capacitor 7μm x 7μm (200 fF)

5s/div and 200 mV/div

Pixel with 2 internal memories

VDDA

GND

Vin

GND

VDDA

Vout

Vcascode

power_on

VDDA

power_on

VDDA

GND

Vin

GND

VDDA

Vout

Vcascode

power_on

Design gain = 8

Measured gain < 4.5

ENC = 20 e-

Design gain = 9

Measured gain < 5

ENC = 18 e-

Basic Design gain = 5

Measured gain = 4

ENC = 12 e-

Promising structure for on-chip CDS processing


1515

Noisy prototype (ENC 50-60 e-) due to large noise bandwidth

Coupling of digital signals to memory nodes during sensor operation prevented the use of the integrated CDS

MAPS operated in current mode

PhotoFET cell – collected charge modulates current in the PMOS transistor

Early prototypes: single cell ENC ~ 5e-

Tested in pixel array configuration Two in-pixel current memory cells

Signal distribution from one pixel


1616

CDS in current modeTwo CDS performing circuits validated (in discrete implementation)

–Capacitance arithmetic (integrator + amplifier)–Subtraction on an operational amplifier (two integrators + amplifier)

PhotoFET – interesting concept and promising results

BUT

Not ready to provide a reliable solution for a vertex detector

Simpler subtraction – faster operation

More amplifiers – higher power consumption

More compact architecture

Lower power consumption


1717

Increased tolerance to ionizing radiation

standard diode layoutstandard diode layout

thin-oxide diode layoutthin-oxide diode layout

Shot Noise Contribution @ 30°C Shot Noise Contribution @ 30°C

and @4 ms integration timeand @4 ms integration timeENCENCshotshot = 39 electrons = 39 electrons

ENCENCshotshot = 12 electrons = 12 electrons

n+n+p+ n+p-well

depleted region

p++ substrate

passivation

oxide

p-epi

n-well

FOXFOXFOX

n+n+p+p-well

depleted region

p++ substrate

passivationoxide

p-epi

n-well

FOX FOX n+

gnd gnd

n+


1818


(MAPS) at IPHC MAPS prototype for PIXEL detector

– Tests and study of performance as a function of ionizing radiation dose

– Tests and study of sensor’s susceptibility to latch up

3-sensor telescope system with prototype readout for PIXEL detector

Future development plans Summary and Conclusions




1919

On-chip data processing and complementary RDO

2011

Install final detector

2010

Install 3-module demonstrator (based on Phase1)

First prototypes in hand and tested

Correlated Double Sampling (CDS)= subtraction of two consecutive signal samplesreduces low frequency noiseextracts signal accumulated during integration time

Data sparsification reduction of the amount of data transferred, typically through zero-suppression

Few years back it was planned to built a demonstrator detector based on sensors with 4 ms integration time.

Pixel

Sensors CDS

ADC Data

sparsification

readout

to DAQ

analogsignals

Phase-1 sensors 640 μs integration time

Complementary detector readout

MimoSTAR sensors 4 ms integration time

Ultimate sensors < 200 μs integration time

analog

digital digital signals

Disc.

CDS


2020

MAPS Prototype for STAR

MimoSTAR2:

*Joint Test Action Group (JTAG) is the IEEE 1149.1 standard entitled Standard Test Access Port and Boundary-Scan Architecture

Analog readout

Radiation tolerant diode design

JTAG* controlled configuration


2121

MimoSTAR2 performance – ionizing radiation 60Co

Significant improvement in resistance to ionizing radiation

Satisfies initial PIXEL detector requirements

55Fe signal collected in central pixels Degradation of noise performance

Peak corresponds to the full charge collection (1640 e-)


2222

MimoSTAR2 performance – latch upSetup at the Tandem Van der Graff accelerator facility at BNL

No latch ups observed up to energies equivalent to 6000 MIPs

Parasitic thyristor


2323

MimoSTAR2 performance – beam tests

particletrack

DeviceUnderTest

reference planes(strip detectors)

reference planes(strip detectors)

scintilatorscintilator

Standard setup for tests with minimum ionizing particles

(5 GeV e-

@ DESY)

detection efficiency > 99.8 % when S/N >12

Analysis by Auguste Besson, IPHC

STD 0.8 ms

STD 4.0 ms

RAD 0.8 ms

RAD 4.0 ms

STD 0.8 ms

STD 4.0 ms

RAD 0.8 ms

RAD 4.0 ms


2424



for PIXEL detector – Construction and tests of the telescope head– FPAG and software programming for JTAG communication– Study of efficiency of the proposed hit finding algorithm– Laboratory calibrations, ALS test and sensors alignment,

tests in the STAR environment

Future development plans Summary and Conclusions




2525

Motivation for the 3-sensor telescope

The telescope is a small prototype and contains all elements easily scalable to meet the requirements of the PIXEL

Test functionality of a prototype MIMOSTAR2 detector in the environment at STAR 2006-2007:

– Charged particle environment near the interaction region in STAR.– The noise environment in the area in which we expect to put the final

PIXEL.– Performance of the MIMOSTAR2 sensors.– Performance of our hit finding algorithm.– Performance of our hardware / firmware as a system.– Functionality of our tested interfaces to the other STAR subsystems.


2626

Implementation of the 3-sensor telescope

MIMOSTAR

2

MIMOSTAR

2

MIMOSTAR

2

Motherboard

Analog signalsClock & controlJTAGLU prot. Power

Analog signalsClock & controlCluster FIFOHot Pixel MapMemory Access(for full frame)Trigger infoPower

Stratix

Daughtercard

Trigger, Clockfrom MWPC

Powerfrom MWPC

JTAGx3 for MIMOSTARx1 for daughtercard

Latch upmonitor and reset

powerDDL to Linux PC

serial / ip connection

JTAG

Trigger, ClockCluster FIFOBusy to trigger

PC(WIN)

control conectionto PC in DAQ room

STRATIX

DAUGHTER CARD

RORC SIU

MimoStar2 chips on kapton cables

MOTHERBOARD

Acquisition Server (Linux)

Control PC (Win)


2727

Zero suppression through on-the-fly hit finding

8-bit post-CDSdata50 MHz datastream.

18pixel addresscounter

Cluster Finding Saving Address Only

Cluster sensor operateson these 9 pixels

Enable

ToEventBuilder

columnn

columnn-1

columnn+1

row nrow n-1 row n+1

highthresh.

shift register length = 1 column

Hits are recognized when:1. signal in the central pixel exceeds high threshold2. and any one of the neighboring 8 pixels exceeds

low threshold.

Efficiency and accidental rates are comparable to the traditional ADC sum method.

Functionally equivalent to a raster scan

Checks 9 pixel window at each clock cycle

Only pixel addresses are saved


2828

Cluster Finder Efficiency

Sum method Two Threshold FPGA method

Cut on the central pixel goes from 14 to 8 ADC counts (left to right) every 1 ADC = 7.1 e-

Detection efficiency >99% and accidental hit rate <10-4 achievable for a range of settings

Expected close to 3 orders of magnitude data rate reduction for a 4 ms PIXEL detector


2929

MimoSTAR2 Telescope test at the ALS1.2 GeV electrons at the ALS Booster Test Facility

Due to not decoupled DAC pads on the sensor, our noise level was double the value achieved under normal conditions.

Decoupled 11-15 e-

Not decoupled 30-35 e-

@ 30º C

MPV = 49 (Standard) and 43 (Radtol) ADC counts at ~230 electrons

Sensors aligned based on straight tracks reconstructed in all 3 planes

Scan of threshold levels to calibrate the system for the next stage of tests in the STAR environment

• High cut 25 ADC• Low cut 14 ADC


3030

The interraction point is ~2 m away

MimoSTAR2 Telescope test at STAR

Telescope head 145 cm from interaction point 5 cm below beam pipe.

Magnet Pole Tip

Electronics BoxBeam Pipesignals originating at the collision point Background tracks

parallel to the beam

(magnified)

theoretical projection of the beam diamond

Increased width from multiple Coulomb scattering in the beam pipe

No environmentally induced noise observed Operation in magnetic field of 0.5 T Average RHIC luminosity 8×1026 cm-2s-1

On average 25 clusters per cm2 per frame (1.7 ms)

Operation of the complete system was validated

Analysis by Xiangming Sun, LBL

View of TPC end cap

(Run 200 GeV Au-Au)


3131







3232

What will a pixel for the PIXEL look like?

The simplest pixel Sequential pixel readout

In-pixel amplifier In-pixel CDS Column parallel readout On-chip discriminators

MAPS developed for STAR started with a very simple pixel architecture

Currently, the most promising architecture developed by IPHC and CEA-Saclay

There is always room for improvements

… and we still have a little bit of time

Meets PIXEL requirements

Mimosa 16


3333

Final detector system

2011


2010


Under development +

Currently in the testing phase

Pixel

SensorsCDS Disc.

Data

sparsification

readout

to DAQ

analogsignals

Phase-1 sensors – 640 μs integration time

Ultimate sensors – <200 μs integration time

digitalsignals

Pixel


3434







3535

Summary and Conclusions

MAPS development is keeping pace with requirements for STAR– Development of pixels for on chip CDS processing

(in-pixel amplifiers, on chip CDS, alternative current mode)

MimoSTAR2 prototype was a necessary precursor to the final STAR PIXEL sensor

– Validation of the technology based on the first prototypes– Development and testing of the PIXEL detector readout system

The existing sensor architecture with column parallel readout should satisfy PIXEL detector requirements

IPHC-LBL development plan leads us to achieving the design goals in the next few years (2010 – detector demonstrator, 2011 final installation)

PIXEL detector is going to be the first vertex detector built with MAPS technology – significant impact on the HEP field


3636

Thank you for your attention


3737

Backup Slides


3838

Introduction to the STAR experiment

Penetrating probes (created early in a collision) are sensitive to the evolution of the medium– Particles with very high transverse momentum– Heavy particles containing charm or bottom quarks

Some of the observed physics:

To study next:– Production of heavy quarks– Elliptic flow of heavy quarks

x

zFlow

Suppression of the side-away jets

source source


3939

Penetrating probes (created early in a collision) are sensitive to the evolution of the medium

– Particles with very high transverse momentum– Heavy particles containing charm or bottom quarks

To study next:– Charm flow to test thermalization of light quarks at RHIC– Charm energy loss to test pQCD in a hot and dense medium at RHIC

Selected result: spectra of heavy quarks

QGP in heavy-ion collisions

The corresponding heavy flavor decayed electron spectra are shown as black curves.

Single electron/positron spectra from semileptonic decays are not sufficient.

S. Batsouli et al. Phys. Lett. B557, 26 (2003)


4040

D0 reconstruction

(from HFT proposal)

The D0 signal, after topological cuts, is shown by the solid black circles.

The original spectrum, before software cuts, is shown by the line of open circles.


4141

STAR pointing resolution

Pointing resolution of the TPC alone

Pointing resolution at the vertex by the TPC+SSD+IST+PIXEL detectors


4242

PIXEL development plan

Original plan (2006)

New plan (2007)

06 2011


binary readout

640 μs integration time

08 2007

Wafers of full-reticule MimoSTAR4

08 2008

Install 4ms detector (based on MimoSTAR4)

analog readout

4 ms integration time

06 2011


binary readout

On-chip zero suppression


03 2008

Submit Phase1 for fabrication

08 2010


binary readout


binary readout



4343

MimoSTAR2 Telescope test at the ALS

Merged cluster data – typically 2-3 hits per cluster. Increased noise in sensors results in reduced performance.

Electronic noise background


4444

PIXEL Data Rates for a 4ms detector

Rate @ R1 (2.5 cm) = 52.9 / cm2

Rate @ R2 (8 cm) = 7.3 / cm2 (at L = 1027 cm-2s-1) Average event size = 168 kB * Data Rate = 168 MB/s at 1 kHz * On average 2.5 pixels per cluster

MIMOSTARSensors

50.7 GB/s

ADCsADCs

ADCs

AnalogSignals

CDS

38 GB/sDAQ EVENTBUILDER

114 MB/secHit

Finder+ address

63 GB/s 42 GB/s 168 MB/s

*Bit rate without any overhead


4545

PIXEL ladder


4646

Telescope results

RDO system with on-the-fly data sparsification implemented and functional for Mimostar2 sensors.

Prototype system fully functional and characterized.

Fully functioning interfaces between the prototype system and STAR detector infrastructure.

Completed measurements of detector environment at STAR.


4747

Fast, column-parallel architecture

VREF1 PWR_ON

MOSCAP

RESET

VREF2 VDD

PWR_ON

VR1

VR2

READ

CALIB

ISF

PIXEL

COLUMN CIRCUITRY

OFFSET COMPENSATED COMPARATOR

(COLUMN LEVEL CDS)

SOURCEFOLLOWER

latch

Q

Q_

READ

READ

+

+

+

+

+ +

-

- -

-

LATCH

CALIB

READ

PWR_ON

RESET

READ

CALIB

LATCH

CDS at column level (reduces Fixed Pattern Noise below temporal noise)

122 , inrefCsfrefCALIB VVVVVV

)( 122

122

2

ininsfref

inrefsfin

sfCinREAD

VVVV

VVVV

VVVV

VREAD,CALIB

VCVin1,2

12122

2_ 1 offRREADoffREADS VVVAV

A

AV

READSoffoffRCALIBout VVVVVAAV _21112

1212 RRREADCALIBout VVVVAAV

VS_READ

A1 Voff1 A2, Voff2

Developed in IPHC - DAPNIA collaboration


4848

Next generation of prototypes

Radiation tolerant diode design

Column parallel readout with on-chip discriminators

Binary readout

JTAG controlled configuration

On-chip zero suppression (currently at prototyping stage)


4949

Summary and Conclusions An architecture of the MAPS sensor that should comply with the final

PIXEL detector requirements exists and provides very promising initial results

The on-going development of pixel architectures and in particular in-pixel amplifiers has a potential of further improving the established performance

Readout architecture for the PIXEL detector has been prototyped and validated

– Reading out sensors with binary output will require adjustments w.r.t. the existing solution (fast LVDS readout)

– Detector dead-time is primarily limited by the number of externally allocated readout buffers

The next mile-stone for MAPS and PIXEL development will integrate the new full-size (640×640 pixels) sensor prototype (Phase-1 under development), prototype mechanical support and new readout system for fast binary sensor readout


5050

The new vertex detector for the STAR experiment– Introduction to the STAR experiment– HFT: new vertex detector for STAR– PIXEL detector

Development of Monolithic Active Pixel Sensors (MAPS) at IPHC

MAPS prototype for PIXEL detector 3-sensor telescope system with prototype readout




Documents

PhD Thesis defense Michal Szelezniak ULP, Strasbourg 25 February 2008