Fast Timing for Collider Detectors · 2018. 11. 20. · ' H R J H L M 6 P3 0 HKKL K PN Hz Fast Timing for Collider Detectors - CERN Academic Training Program 10. Larger, thinner crystals

Fast Timing for Collider Detectors

Chris Tully (Princeton University)

CERN Academic Training Lectures (2/3)

11 May 2017

Outline

• Detector technologies with fast timing capabilities

• Readout methods for fast timing layers and calorimeters

Fast Timing for Collider Detectors - CERN Academic Training Program 2

Time-of-Flight Position Emission Tomography (TOF-PET)


LOR = Line-of-Response

LYSO crystals (thick)

SiPM photodetectors

Standard conversion 1ps 300 microns

Silicon Photo-Multiplier (SiPM)Silicon photomultipliers (SiPMs)

Yu. Musienko, 2016 IEEE-NSS/MIC, Strasbourg 21

Structure and principles of operation (briefly)

Al electrode

Rquench

n+/p junctions

p-Si substrate

SiO2+Si3N4p-epi layer

300µ

2-4µ

(EDIT-2011, CERN)

Vbias> VBD

GM-APD

Rq

substrate

Al electrode

Vout

Q Q

Qtot = 2Q

• SiPM is an array of small cells (SPADs) connected in parallel on a common substrate and operated in Geiger mode

• Each cell has its own quenching resistor (from 100kΩ to several MΩ)

• Common bias is applied to all cells (~10-20% over breakdown voltage)

• Cells fire independently

• The output signal is a sum of signals produced by individual cells

For small light pulses (Ng<<Npixels) SiPM works as an analog photon detector

The very first metall-resitor-smiconductor APD (MRS APD) proposed in 1989 by A. Gasanov, V. Golovin,

Z. Sadygov, N. Yusipov (Russian patent #1702831, from 10/11/1989 ). APDs up to 5x5 mm2 were

produced by MELZ factory (Moscow).


Joint effort with FBK on UHD technology

FBK UHD TechnologyReduction of all the feature sizes• Contacts• Resitor• …

Reduction of the active-to-border distance (L)Circular active area (smaller cells)• No corners (with lower field)• Hexagonal cells arranged in

honeycomb configurationLower Rq

• For even faster recharge

L < 1um

4

July 7 , 2016 A. Heering, CMS HCAL collaboration

FBK-irst SiPM for HB

5

UHD1 technology: small cellsJoint R&D effort from ND and FBK in the last 3 years

Detection Chain


q

SiPMCrystal electronics

g

Dt

tkth pe = Dt

Conversion depth

+ tk’ ph

Scintillation

process

+ ttransit

Transit time

jitter

+ tSPTR

Single photon

time spread

+ tTDC

TDC

conversion time

Random deletion 1Absorption

Self-absorption

Random deletion 2SiPM PDE

Unwanted pulses 2DCR

Unwanted pulses 1DCR, cross talk

Afterpulses

P. Lecoq

TOFPET MIP Timing Layer


LSO:Ce,0.4%Ca 2x2x5mm3, meltmount coupled to 3x3mm2 NUV SiPM from FBK, 55%PDE

S.

Gundacker

et

al.,

CERN

2x2mm2 section

511 KeV 150GeV muons

5MeV deposited

Time Jitter

Paolo Meridiani

TIME RESOLUTION VS SI AMPLITUDE

15

Resolution scales with 1/A (noise

contribution) + a constant term

Constant term compatible with

expected reference MCP precision

Time resolution expected to

improve with signal amplitude:


voltage noise

band of signal

timing jitter arising

from voltage noise

timing jitter is

much smaller

for faster

rise-time

V

time

Detecting a MIP with SiPM Readout

8/17/16HCPSS 2016 Calorimetry Lecture 1 8

Landau Peak

at ~3000 fC

1 fC ~ 6250 e−

• (low light) MIP signal with ~60 p.e. in plastic scintilltor (CMS HCAL HE)

Detecting a MIP with SiPM Readout

8/17/16HCPSS 2016 Calorimetry Lecture 1 9

Landau Peak

at ~3000 fC

1 fC ~ 6250 e−

Pedestal electronic noise RMS ~3 fC

MIP S/N ~ 1000 ?

Not really, single pe firing

dominates instrument noise

MIP S/N ~ 150

N~√DCR (Dark Count Rate)

and the DCR increases with radiation dose)

Single “pe” peak

(single pixel gain)

~50 fC each

50 * 6250 / (p)e−

Gain=3*105

(3000 fC/50 fC)/5 MeV

3pe/MeV

MIP S/N in the first ~200ps

Simulation tool for sensors optimization

Test beam data of unirradiated devices

show good agreement with simulation

● Geant4 for tracking of charged particles and

ray tracing of optical photons until detection

● Signal from LYSO combined with SiPM

response (shaping time, PDE, SPTR, etc.)

● Dark count rate from SiPM added to signal


Larger, thinner crystals Optimizition for ~25ps

11Fast Timing for Collider Detectors - CERN Academic Training Program

Crystal Pulse Reconstruction


TDC (or waveform digitizer)

of Fast Comparator Output

Lots of signal,

AC coupling possible

Acts like capacitive

divider for S and N

(N~√DCR)

Electronics sees low

input capacitance if a

shunt capacitor is used

Dark Count Rate (DCR) drops with Temperature


Radiation hardness of barrel sensors } LYSO:Ce crystals thoroughly tested

} Negligible light loss [ RIAC = 3 m-1 at 1x1015 cm2 and 100 kGy ]

} Induced radio-luminiscence marginal at the barrel fluence

} SiPMs: increase of dark current and dark rate

} Acceptable in the barrel: small-pixels SiPMs (production ready FBK/HPK)

} Total power consumption from 291 k (5x5 mm2) SiPMs:

} ~7 kW (~12 kW) at -29 oC (-23 oC)

16

A.Heering et al.

THIS TECHNNOLOGY IS NOT VIABLE IN THE ENDCAPS

End-of-life S/N is an optimization of crystal size, SiPM PDE and DCR growth from irradiation (at low temp)


Silicon Sensors with Gain

• Favorable technology in the push for higher radiation hardness, as is

needed at high eta and within calorimeters

• Important parameters of silicon: 100micron/ns (when drift velocity

saturated at ~30kV/mm E-field) and 73 e-h pair per micron for MIP

•MIP timing of ~30ps requires high S/N and uniform charge collection

– largely driven by E-field geometry and Landau fluctuations


Different Gain/E-Field Geometries are under study (RD50): Reach-Through and Deep-Depleted


Lindsey Gray, FNAL

Silicon Timing: Deep-Depleted APDs

๏ Deep depleted APD read out through capacitatively coupled mesh

• Silicon is biased, image charge read out

• Gain layer and drift region overlap

• Mesh serves to stabilize E-field shape over large area for good performance over whole device

• Operates at high gain / high voltage

๏ 20 ps resolution achieved on 8x8mm2 non-irradiated device

๏ No conclusive results yet for irradiated devices

5

verypreliminarylookattimingondetectoredge

18

edge structure of these

High field Si complicated.

It has been difficult to evaluate

w. laser model

first look at edge behavior

very encouraging!

take small difference of

edge behavior from bulk

with grain of salt

timing algorithm preliminary

small pulse height distortion

note on “appropriate pixel size”

timing detectors• for most of CMS rapidity optimal timing pixel->0.5->1 cm2 • requires attention to field uniformity/metallization and CD impact • much recent HyperFast Silicon progress in both areas

• new packaging w. Bert Harrop(Princeton) • latest Hi-BW transimpedence amp (see SNW & M.Newcomer

ACES2014)

CD=0 pF

CD=22pF

(ringing removed

w. new LVPS)

HFS- mapping Landau Distribution vs. muon data impact position

Low-Gain Avalanche Detectors (LGAD) gain O(10) – 3 suppliers (CNM, FBK, HPK)

Deep-Depleted Avalanche Photo-Diodes

(DD-APD) gain O(500) – 1 supplier (RMD)

E-Field Geometries are very different


Uniform E-Field


Wide implant Narrow implant

a) b)

0 200 400 600 800 x [mm]

0 200 400 600 800 x [mm]

40

30

20

10

y [

mm

]

40

30

20

10

y [

mm

]

0.8

0.6

0.4

0.2

Ew

[1/m

m]

Ew

[1

/mm

]

0.8

0.6

0.4

0.2

LGAD achieves uniform E-field with a wide implant

DD-APD achieves uniform E-field with a mesh

Placed on Top

Surface

Transient Current Technique


LGAD signal looks like a

current that flows across a

capacitor for the time it

takes the deposited charge

to traverse the thickness of

the device

takes roughly 1.4ns to

traverse 140 microns

LGAD now prefers 50

micron thickness

DD-APD is coming from a

~40 micron avalanche

region and is narrower

Time [ns]

Electronics Readout Schemes


Sensor Pre-amplifier Time measuring circuit

S

tr

Vth

Comparator

Cd Iin

Vth

Time [ns]

Current Amplifier (BBA)

Charge Sensitive Amplifier (CSA) Time

Cu

rre

nt

Am

pli

tud

e [

mV

]

0 1 2 3 4 5 6 7

50 40 30 20 10

H. Sadrozinski, A. Seiden, N. Cartiglia “4-Dimensional Tracking with Ultra-Fast Silicon Detectors“

Landau fluctuations in LGAD geometry

40 Nic

olo

Ca

rtig

lia, IN

FN

, To

rin

o -

ETL

re

vie

w 2

/03/1

7

Non uniform charge deposition along the track

! Set the comparator threshold as low as you can

! Use thin sensors

300 micron thick

50 micron thick

Vth [mV]

This is a physical limit to time resolution: beat it with thin detectors and low comparator threshold.


LGAD timing resolution with 50 micron thickness

11 Nic

olo

Ca

rtig

lia, IN

FN

, To

rin

o -

ETL

re

vie

w 2

/03/1

7

The correct gain value

Our measurements show that the ETL target time resolution of ~ 30 ps can

be reached with gain ~ 20-30.

The time resolution is determined by charge non-uniformity

The working point will be determined by the interplay with the electronics

Jitter term: scales

with gain (dV/dt)

Charge non

uniformity: ~

constant with gain

H. Sadrozinski, TREDI 2017


LGAD sensor fill factor – wafer-level process

25 Nic

olo

Ca

rtig

lia,

INFN

, To

rin

o -

ETL

re

vie

w 2

/03

/17

Sensor R&D

Sensor fill factor: what is the minimum

distance between pads?

High fields require

special terminations at

the edge of each pad

Currently ~ 30 micron

! Goal: ~ 20 micron


LGAD gain layer doping sensitivity high

27 Nic

olo

Ca

rtig

lia,

INFN

, To

rin

o -

ETL

re

vie

w 2

/03/1

7

Gain vs gain layer doping

Unfortunately, the gain is very sensitive to the doping level

Small decrease

in doping: 10%

Large decrease in

gain: 80%


Radiation dose effects on p-doping

29 Nic

olo

Ca

rtig

lia,

INFN

, To

rin

o -

ETL

re

vie

w 2

/03

/17

Radiation issue: Initial acceptor removal

This term indicates the “removal” of the initially-present p-doping.

For UFSD this is particularly problematic as it removes the gain layer

Irradiation ! Defects ! Boron becomes interstitial

B

The boron doping is still there, only it has been moved into a different

position and it does not contribute to the doping profile, it is inactive

B

B


Gallium doping and Carbonated Boron

30 Nic

olo

Ca

rtig

lia, IN

FN

, To

rin

o -

ETL

re

vie

w 2

/03

/17

Initial acceptor removal: mitigation

Gallium doping: Irradiation ! defects ! Gallium has lower diffusivity

Carbonated Boron: Irradiation ! defects ! Carbon fills interstitial states

C C

C

C

C

C

C

C

C

C

C

C

B

Ga

Ga

Ga

Ga

Ga

Ga

Ga

Ga

Ga

C C

C

C

C

C


Maintaining gain at high fluence


0

20,000

40,000

60,000

80,000

100,000

0 200 400 600 800 1000

Co

llect

edC

har

ge[e

]

Bias[V]

Collectedchargeasafunctionofbiasvoltageina50-micronthicksensorfordifferentfluences

Fluence:5e14neq/cm2,Activedoping=63%



Unirradiated

Chargesneeded

forgoodtimingmeasurement

Increasingfluence

LGAD maximum gain with neutron fluence

15 Nic

olo

Ca

rtig

lia, IN

FN

, To

rin

o -

ETL

re

vie

w 2

/03/1

7

Time resolution for irradiated sensors

No difference in behavior before and after irradiation: the time resolution

scales with gain. ! Keep the gain high

H. Sadrozinski, TREDI 2017

W5 2e15 n/cm2

W3, W7 6e14 n/cm2

~ 42 ps

W5 pre-rad

J. Lange, TREDI 2017

J. Lange, TREDI 2017

No unexpected features, the

signals are still large and the

leakage current does not prevent

to reach good time resolution.

Good Gaussian

behavior after

irradiation


Summary – Lecture 2

• Fast timing is possible with both light and charge collection

technologies

• Dimensions matter a lot! Fast timing in colliders is possible because

we have the technical capability to tile sq. meters of surface with

small, thin tiles (1-100 mm2) and have electronics with an analog

bandwidth and low noise thresholds on a high S/N MIP

• Radiation fluence is a fierce foe – as usual. More on this tomorrow.


Backup


Calorimeter timing measurements


MIP Timing Layer

3

Optical Transit Time Spread • Effect of the scintillation photon arrival at the photo detector we refer to

as Optical Transit Time Spread.

• Experimental program to explore ultimate timing resolution, in particular

the impact of the optical transit time spread.

γ x

γ x

t1

t2

EM shower propagation

snapshot Scintillation light propagation

cS < c

100 GeV γ

23 cm

Time evolution of a shower from photon in CMS ECAL PbWO crystal (25 cm long).

1.5 [ns] 0.0 0.5 1.0 3

17.05.2016 Adi Bornheim, Calor 2016, Calorimeter Precision Timing

Long Crystal timing measurementsCMS ECAL current timing performance

• Timing resolution of CM S ECAL better then 1 ns was not foreseen in

the original design, despite this:

: excellent t iming resolution already achieved in 2012 (LHC collision @8 TeV).

Z æ ee events.

nγ/eff

A

2103

10

)[n

s]

2-t

1(t

γ

-110

1

C2 γ nγ/effA

N(t) = γ

2.0 ns±N = 33.2

E in EB [GeV]20 40 60 80100

CMS Preliminary - Run1 EB Z study

0.001 ns±C = 0.154

• Timing resolut ion est imated from fit

to: tchannel 1 ≠ tchannel 2.

• Take the two most energet ic channel

for each electron cluster.

Simone Pigazzini Precision t iming with PbWO crystals CALOR 2016 4 / 12Fast Timing for Collider Detectors - CERN Academic Training Program 32

Improvements expected with clock distributionCMS ECAL current timing performance

• Timing resolution improves for channels of the same cluster.

• Further gain when considering channels that belongs to the same readout unit.

Channels in the same shower but

different readout units.

Channels in the same shower and same

readout units.


Less shaping and higher analog bandwidthCMS ECAL electronics for HL-LHC

Improvements:

• Noise from APD leakage current.

: increased by long exposure to radiat ion.

• Allow higher trigger rates.

• Mit igate pileup from previous and following bunch crossings.

• Mit igate signal contaminat ion from concurrent interactions in the same bunch

crossing (through t iming).

• Different solut ions are under evaluat ion.

• Current ECAL electronics with faster shaping

t ime could sat isfy the requirements.

: Shorter signal

: Larger Amplitude/ noise

: Better timing resolution.

Test beam: digitized APD signal



March 2nd 2017 MD - MIP timing layer review 7

Proposed architecture Proposed architecture

Digitaloutput

TIA Filter ADC

Rg

LGAD – Avalanche Region is localized

Low Gain Avalanche Detectors (LGADs)

3

The LGAD concept has been proposed and manufactured first by CNM

(National Center for Micro-electronics, Barcelona)

Field needed: E ~ 300 kV/cm

High field obtained by

(1) adding an extra doping layer

(2) by external VBias

Gain layer High field

Nic

olo

Ca

rtig

lia,

INFN

, To

rin

o -

ETL

re

vie

w 2

/03/1

7

The LGAD project has been developed initially by the RD50 collaboration


Light generation in scintillators

Rare Earth4f

5d


• Wide emission spectrum from UV to IR

• Ultrafast emission in the ps range

• Independant of temperature

• Independant of defects

• Absolute Quantum Yield

Whn/Wphonon = 10-8/(10-11-10-12)

≈ 10-3 to 10-4 ph/eh pair

• Higher yield if structures or dips in CB?

Interesting to look at CeF3

Hot intraband luminescence

More details in SCINT2013 paper TNS-00194-2013

M. Korzhik, P. Lecoq, A. Vasil’ev


Documents

Fast Timing for Collider Detectors · 2018. 11. 20. · ' H R J H L M 6 P3 0 HKKL K PN Hz Fast Timing for Collider Detectors - CERN Academic Training Program 10. Larger, thinner crystals