Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... ·...

48
1 Readout Electronics the path of the signal from detection to acquisition

Transcript of Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... ·...

Page 1: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

1

Readout Electronics the path of the signal from detection to acquisition

2

Detector(s)

A large variety of detectors But similar modeling

6x6 pixels4x4 mm2

HgTe absorbers 65 mK 12 eV 6 keV

ATLAS LAr calorimeter CMS pixel module

PMT in ANTARES

3

Detector Signal

bull Detector signal generally a short current pulse

i =VR (R = 50Ω oscilloscope termination) ndash thin silicon detector (10 ndash300 microm) 100 psndash30 ns ndash thick (~cm) Si or Ge detector 1 ndash10 micros ndash proportional chamber 10 ns ndash10 micros ndash Microstrip Gas Chamber 10 ndash50 ns ndash Scintillator+ PMTAPD 100 psndash10 micros

int=prop dttiQE ss )(

t

V

oscilloscope picture

4

Signal measurements

Various measurements of this signal are possible Depending on information required -  Signal above threshold

digital response event count

-  Integral of current = charge egrave energy deposited

-  Time of leading edge egrave time of arrival (ToA) or time of flight (ToF)

- Time of signal above threshold egrave energy deposited by TOT

and many more hellip

t

V

oscilloscope picture

threshold

5

Energy measurement

Determine energy deposited in a detector bull Necessary to integrate detector signal current ndash integrate charge on input capacitance ndash use integrating (ldquocharge sensitiverdquo) preamplifier ndash amplify current pulse and use integrating ADC (analog to digital converter)

int=prop dttiQE ss )(

Examples

E [GeV] 10 20 30 40 50

E meas of e- in a calorimeter

6

Electronics requirements for E meas Dynamic range maximum signalminimum signal (or noise) typically 103ndash105

often specified in dB (=20log VmaxVmin) = 60 ndash100 dB also in bits 2n= VmaxVmin= 10 ndash18 bits

The large dynamic range is a key parameter for calorimeter electronics

7

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

8

Detector modeling (approximation)

Detector = capacitance Cd ndash  Pixels 01-10 pF ndash  PMs 3-30 pF ndash  Ionization chambers 10-1000 pF ndash  Sometimes effect of transmission line

Signal current source ndash  Pixels ~100 e-microm ndash  PMs 1 photoelectron -gt 105-107 e- ndash  Modeled as an impulse (Dirac)

i(t)=Q0δ(t)

Missing ndash  High Voltage bias ndash  Connections grounding ndash  Neighbors ndash  Calibrationhellip

I in C d

Detector model

Typical PM signal

10ns

Reality is different

9

Ionization chamber signal shape

-ΔU

derive signal for single ionizing particle [R = infin]

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 2: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

2

Detector(s)

A large variety of detectors But similar modeling

6x6 pixels4x4 mm2

HgTe absorbers 65 mK 12 eV 6 keV

ATLAS LAr calorimeter CMS pixel module

PMT in ANTARES

3

Detector Signal

bull Detector signal generally a short current pulse

i =VR (R = 50Ω oscilloscope termination) ndash thin silicon detector (10 ndash300 microm) 100 psndash30 ns ndash thick (~cm) Si or Ge detector 1 ndash10 micros ndash proportional chamber 10 ns ndash10 micros ndash Microstrip Gas Chamber 10 ndash50 ns ndash Scintillator+ PMTAPD 100 psndash10 micros

int=prop dttiQE ss )(

t

V

oscilloscope picture

4

Signal measurements

Various measurements of this signal are possible Depending on information required -  Signal above threshold

digital response event count

-  Integral of current = charge egrave energy deposited

-  Time of leading edge egrave time of arrival (ToA) or time of flight (ToF)

- Time of signal above threshold egrave energy deposited by TOT

and many more hellip

t

V

oscilloscope picture

threshold

5

Energy measurement

Determine energy deposited in a detector bull Necessary to integrate detector signal current ndash integrate charge on input capacitance ndash use integrating (ldquocharge sensitiverdquo) preamplifier ndash amplify current pulse and use integrating ADC (analog to digital converter)

int=prop dttiQE ss )(

Examples

E [GeV] 10 20 30 40 50

E meas of e- in a calorimeter

6

Electronics requirements for E meas Dynamic range maximum signalminimum signal (or noise) typically 103ndash105

often specified in dB (=20log VmaxVmin) = 60 ndash100 dB also in bits 2n= VmaxVmin= 10 ndash18 bits

The large dynamic range is a key parameter for calorimeter electronics

7

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

8

Detector modeling (approximation)

Detector = capacitance Cd ndash  Pixels 01-10 pF ndash  PMs 3-30 pF ndash  Ionization chambers 10-1000 pF ndash  Sometimes effect of transmission line

Signal current source ndash  Pixels ~100 e-microm ndash  PMs 1 photoelectron -gt 105-107 e- ndash  Modeled as an impulse (Dirac)

i(t)=Q0δ(t)

Missing ndash  High Voltage bias ndash  Connections grounding ndash  Neighbors ndash  Calibrationhellip

I in C d

Detector model

Typical PM signal

10ns

Reality is different

9

Ionization chamber signal shape

-ΔU

derive signal for single ionizing particle [R = infin]

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 3: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

3

Detector Signal

bull Detector signal generally a short current pulse

i =VR (R = 50Ω oscilloscope termination) ndash thin silicon detector (10 ndash300 microm) 100 psndash30 ns ndash thick (~cm) Si or Ge detector 1 ndash10 micros ndash proportional chamber 10 ns ndash10 micros ndash Microstrip Gas Chamber 10 ndash50 ns ndash Scintillator+ PMTAPD 100 psndash10 micros

int=prop dttiQE ss )(

t

V

oscilloscope picture

4

Signal measurements

Various measurements of this signal are possible Depending on information required -  Signal above threshold

digital response event count

-  Integral of current = charge egrave energy deposited

-  Time of leading edge egrave time of arrival (ToA) or time of flight (ToF)

- Time of signal above threshold egrave energy deposited by TOT

and many more hellip

t

V

oscilloscope picture

threshold

5

Energy measurement

Determine energy deposited in a detector bull Necessary to integrate detector signal current ndash integrate charge on input capacitance ndash use integrating (ldquocharge sensitiverdquo) preamplifier ndash amplify current pulse and use integrating ADC (analog to digital converter)

int=prop dttiQE ss )(

Examples

E [GeV] 10 20 30 40 50

E meas of e- in a calorimeter

6

Electronics requirements for E meas Dynamic range maximum signalminimum signal (or noise) typically 103ndash105

often specified in dB (=20log VmaxVmin) = 60 ndash100 dB also in bits 2n= VmaxVmin= 10 ndash18 bits

The large dynamic range is a key parameter for calorimeter electronics

7

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

8

Detector modeling (approximation)

Detector = capacitance Cd ndash  Pixels 01-10 pF ndash  PMs 3-30 pF ndash  Ionization chambers 10-1000 pF ndash  Sometimes effect of transmission line

Signal current source ndash  Pixels ~100 e-microm ndash  PMs 1 photoelectron -gt 105-107 e- ndash  Modeled as an impulse (Dirac)

i(t)=Q0δ(t)

Missing ndash  High Voltage bias ndash  Connections grounding ndash  Neighbors ndash  Calibrationhellip

I in C d

Detector model

Typical PM signal

10ns

Reality is different

9

Ionization chamber signal shape

-ΔU

derive signal for single ionizing particle [R = infin]

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 4: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

4

Signal measurements

Various measurements of this signal are possible Depending on information required -  Signal above threshold

digital response event count

-  Integral of current = charge egrave energy deposited

-  Time of leading edge egrave time of arrival (ToA) or time of flight (ToF)

- Time of signal above threshold egrave energy deposited by TOT

and many more hellip

t

V

oscilloscope picture

threshold

5

Energy measurement

Determine energy deposited in a detector bull Necessary to integrate detector signal current ndash integrate charge on input capacitance ndash use integrating (ldquocharge sensitiverdquo) preamplifier ndash amplify current pulse and use integrating ADC (analog to digital converter)

int=prop dttiQE ss )(

Examples

E [GeV] 10 20 30 40 50

E meas of e- in a calorimeter

6

Electronics requirements for E meas Dynamic range maximum signalminimum signal (or noise) typically 103ndash105

often specified in dB (=20log VmaxVmin) = 60 ndash100 dB also in bits 2n= VmaxVmin= 10 ndash18 bits

The large dynamic range is a key parameter for calorimeter electronics

7

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

8

Detector modeling (approximation)

Detector = capacitance Cd ndash  Pixels 01-10 pF ndash  PMs 3-30 pF ndash  Ionization chambers 10-1000 pF ndash  Sometimes effect of transmission line

Signal current source ndash  Pixels ~100 e-microm ndash  PMs 1 photoelectron -gt 105-107 e- ndash  Modeled as an impulse (Dirac)

i(t)=Q0δ(t)

Missing ndash  High Voltage bias ndash  Connections grounding ndash  Neighbors ndash  Calibrationhellip

I in C d

Detector model

Typical PM signal

10ns

Reality is different

9

Ionization chamber signal shape

-ΔU

derive signal for single ionizing particle [R = infin]

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 5: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

5

Energy measurement

Determine energy deposited in a detector bull Necessary to integrate detector signal current ndash integrate charge on input capacitance ndash use integrating (ldquocharge sensitiverdquo) preamplifier ndash amplify current pulse and use integrating ADC (analog to digital converter)

int=prop dttiQE ss )(

Examples

E [GeV] 10 20 30 40 50

E meas of e- in a calorimeter

6

Electronics requirements for E meas Dynamic range maximum signalminimum signal (or noise) typically 103ndash105

often specified in dB (=20log VmaxVmin) = 60 ndash100 dB also in bits 2n= VmaxVmin= 10 ndash18 bits

The large dynamic range is a key parameter for calorimeter electronics

7

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

8

Detector modeling (approximation)

Detector = capacitance Cd ndash  Pixels 01-10 pF ndash  PMs 3-30 pF ndash  Ionization chambers 10-1000 pF ndash  Sometimes effect of transmission line

Signal current source ndash  Pixels ~100 e-microm ndash  PMs 1 photoelectron -gt 105-107 e- ndash  Modeled as an impulse (Dirac)

i(t)=Q0δ(t)

Missing ndash  High Voltage bias ndash  Connections grounding ndash  Neighbors ndash  Calibrationhellip

I in C d

Detector model

Typical PM signal

10ns

Reality is different

9

Ionization chamber signal shape

-ΔU

derive signal for single ionizing particle [R = infin]

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 6: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

6

Electronics requirements for E meas Dynamic range maximum signalminimum signal (or noise) typically 103ndash105

often specified in dB (=20log VmaxVmin) = 60 ndash100 dB also in bits 2n= VmaxVmin= 10 ndash18 bits

The large dynamic range is a key parameter for calorimeter electronics

7

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

8

Detector modeling (approximation)

Detector = capacitance Cd ndash  Pixels 01-10 pF ndash  PMs 3-30 pF ndash  Ionization chambers 10-1000 pF ndash  Sometimes effect of transmission line

Signal current source ndash  Pixels ~100 e-microm ndash  PMs 1 photoelectron -gt 105-107 e- ndash  Modeled as an impulse (Dirac)

i(t)=Q0δ(t)

Missing ndash  High Voltage bias ndash  Connections grounding ndash  Neighbors ndash  Calibrationhellip

I in C d

Detector model

Typical PM signal

10ns

Reality is different

9

Ionization chamber signal shape

-ΔU

derive signal for single ionizing particle [R = infin]

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 7: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

7

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

8

Detector modeling (approximation)

Detector = capacitance Cd ndash  Pixels 01-10 pF ndash  PMs 3-30 pF ndash  Ionization chambers 10-1000 pF ndash  Sometimes effect of transmission line

Signal current source ndash  Pixels ~100 e-microm ndash  PMs 1 photoelectron -gt 105-107 e- ndash  Modeled as an impulse (Dirac)

i(t)=Q0δ(t)

Missing ndash  High Voltage bias ndash  Connections grounding ndash  Neighbors ndash  Calibrationhellip

I in C d

Detector model

Typical PM signal

10ns

Reality is different

9

Ionization chamber signal shape

-ΔU

derive signal for single ionizing particle [R = infin]

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 8: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

8

Detector modeling (approximation)

Detector = capacitance Cd ndash  Pixels 01-10 pF ndash  PMs 3-30 pF ndash  Ionization chambers 10-1000 pF ndash  Sometimes effect of transmission line

Signal current source ndash  Pixels ~100 e-microm ndash  PMs 1 photoelectron -gt 105-107 e- ndash  Modeled as an impulse (Dirac)

i(t)=Q0δ(t)

Missing ndash  High Voltage bias ndash  Connections grounding ndash  Neighbors ndash  Calibrationhellip

I in C d

Detector model

Typical PM signal

10ns

Reality is different

9

Ionization chamber signal shape

-ΔU

derive signal for single ionizing particle [R = infin]

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 9: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

9

Ionization chamber signal shape

-ΔU

derive signal for single ionizing particle [R = infin]

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 10: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

10

Ionization chamber signal shape

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 11: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

11

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 12: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

12

Reading the signal

Signal ndash  Signal = current source ndash  Detector = capacitance Cd

ndash  Quantity to measure bull  Charge egrave integrator needed bull  Time egrave discriminator + TDC

Integrating on Cd

ndash  Simple V = QCd

ndash  laquo Gain raquo 1Cd 1 pF egrave 1 mVfC ndash  Need a follower to buffer the voltagehellip

egrave parasitic capacitance ndash  Gain loss possible non-linearity ndash  crosstalk ndash  Need to empty Cdhellip

I in C d

- +

QCd

Impulse response

Voltage readout

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 13: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

13

Reading the signal (II)

Depends on the detector capacitance

If the input time constant of the amplifier τ = CiRi is large compared to the duration of the current pulse of the detector tc the current pulse will be integrated on the capacitance Ci The resulting voltage at Ci and Ri is

vi = vs = Qs (Cdet + Ci)

The fraction of the signal charge measured is The dynamic input capacitance Ci should be gtgt Cdet to get a good ratio close to 1

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+=

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 14: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

14

Charge sensitive amplifier

Voltage gain dVodVi=-A egrave vo=-Avi Input impedance = infin (no signal current flows into amplifier input) Voltage diff across Cf vf=(A+1)vi egrave Charge deposited on Cf Qf=Cfvf Qi = Qf (since Zi= infin ) egrave Effective input capacitance

Ci= Qivi = Cf(A+1)

ldquodynamic input capacitancerdquo

Amplifier gain

)1(111

gtgtasymp+

==sdot

sdot== A

CCAA

CA

vCvA

dQdVA

ffiii

i

i

oQ

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 15: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

15

Charge sensitive amplifier (II)

So finally the fraction of charge signal measured by the amplifier is Example A = 103

Cf = 1pF egrave Ci = 1nF

Cdet = 10pF egrave QiQs = 099 (CigtgtCdet)

Cdet = 500pF egrave QiQs = 067 (Ci ~ Cdet)

iii

ii

s

i

CCCCvvC

QQ

11

)( detdet +=

+= )1( gtgtasymp A

CACi

f

Si det 50um thick 500mm2 area

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 16: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

16

Charge sensitive (pre)-amplifier

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 17: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

17

Charge vs Current preamps

Current Charge

bull  Charge preamps ndash  Best noise performance ndash  Best with short signals ndash  Best with small capacitance

bull  Current preamps ndash  Best for long signals ndash  Best for high counting rate ndash  Significant parallel noise

bull  Charge preamps are not slow they are long

bull  Current preamps are not faster they are shorter (but easily unstable)

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 18: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

18

Readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADCs) bull  Thousands to millions of channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 19: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

19

Pulse shaping Two conflicting objectives 1)  Limit the bandwidth to match

the measurement time egrave too large bandwidth increases the noise

2)  Contain the pulse width so that successive signal pulses can be measured without overlap (pile-up)

egrave Short pulse duration increases the allowed signal rate but also noise

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 20: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

20

CR-RC shaper Example of a simple shaper CR-RC bull  the high-pass filter sets the duration of the pulse to have a decay time τd bull  the low-pass filter increases the rise time to limit the noise bandwidth key design parameter peaking time egrave it dominates the noise bandwidth

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 21: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

21

CR-RC shaper (II)

Effect of a CR-RC shaper with fix integrator time constant = 10ns and variable differentiator time constant

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 22: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

22

readout architecture for E meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement of amplitude (ADC) bull  Thousands to millions of channels

Black box assumed to be known

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 23: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

23

Analog memories

bull  Switched Capacitor Arrays (SCAs) ndash  Store signal on capacitors (~pF) ndash  Fast write (~ GHz) ndash  Slower read (~10MHz) ndash  Dynamic range 10-13 bits ndash  depth 100-2000 capacitors ndash  Insensitive to absolute value of capa

(voltage write voltage read) ndash  Low power ndash  Possible loss in signal integrity (droop

leakage current)

bull  The base of 90 of digital oscilloscopes

In

Write

Read Out

Principle of a laquo voltage-write voltage-read raquo analog memory

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 24: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

24

Readout architecture for t meas

Most front-ends follow a similar architecture

bull  Very small signals (fC) -gt need amplification bull  Measurement time (discriminator TDCs) bull  Thousands to millions of channels

discriminator13 TDC13

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 25: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

25

Discriminator

Discriminator techniques leading edge triggering and constant fraction triggering

threshold

Working principle compare voltage level of signal with fixed voltage level (threshold) if the signal level exceeds the threshold a standard logic signal is generated

A

B

A

B

Output A

Output A

time walk no time walk

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 26: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

26

Time measurements

Time measurements are characterized by their slope-to-noise ratio

Two main effects contribute to the deterioration of a time measurement ie time of threshold crossing fluctuates due to

jitter time walk

Often driven by the time constant of the shaper which determines rise time amp amplifier bandwidth

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 27: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

From theory to reality

27

Every signal comes together with noise hellip

Noise is a quite complex topic (see backup) we do not have time to discuss it But always remember

egrave what matters in a detector is SN

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 28: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

28

Examples of readout elerctronics

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 29: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

29

Ionization calorimeters

Examples DOslash(LAr) NA48 (LKr) ATLAS (LAr) H1

Stable Linear Easy to calibrate () Moderate resolution

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 30: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

30

ATLAS LAr Front End boards

Amplify shape store and digitize Lar signals10485761048577104857810485791048580104858110485821048583104858410485851048586104858710485881048589104859010485911048592104859310485941048595104859610485971048598104859910486001048601104860210486031048604104860510486061048607104860810486091048610104861110486121048613104861410486151048616104861710486181048619104862010486211048622104862310486241048625104862610486271048628104862910486301048631104863210486331048634104863510486361048637104863810486391048640104864110486421048643104864410486451048646104864710486481048649104865010486511048652104865310486541048655104865610486571048658104865910486601048661104866210486631048664104866510486661048667104866810486691048670104867110486721048673104867410486751048676104867710486781048679104868010486811048682104868310486841048685104868610486871048688104868910486901048691104869210486931048694104869510486961048697104869810486991048700104870110487021048703 128 warm preamps 128 tri-gain shapers 128 quad pipelines 32 ADCs (12bits 5 MHz) 1 optical output (Glink)

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 31: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

31

ATLAS LAr shaper Goal optimize signal to noise ratio between electronics noise and pileup noise Ionization signal ~500ns=20 LHC bunch Xings Reduced to 5 bunch Xings with fast shaper egrave worse SN due to loss of charge Choice of peak time varies with luminosity egrave 45ns at L=1034cm-2s-1

serial noise LAr

parallel noise LAr

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 32: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

32

Crystal calorimeters

Babar(CsI) Kloe(CsI) CMS (PbWO4) L3 CLEO Belle ALICE

Fast Best resolution Difficult to calibrate expensive

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 33: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

33

CMS ECAL Electronics building block Trigger Tower (25 channels)

-1 mother board -1 LV regulator board -5 VFE boards (5 channels each) -1 FE board

2 fibres per TT sending -trigger primitives (every beam crossing) -data (on level 1 trigger request)

Trigger Tower

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 34: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

34

Scintillating calorimeters

CMS hadronic LHCb OPERA ILC hadronic ILC em ATLAS hadronic

Fast Cheap Moderate resolution Difficult to calibrate

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 35: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

35

ILC hadronic calorimeter (CALICE)

ASIC amplification + shaping + multiplexing (18 ch)

VFE control board for 12 ASICs layer connect to SiPMs

pixel device operated in Geiger mode

Single tile readout with WLS fiber + SiPM

Read out 216 tilesmodule 38 sampling layers ~8000 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 36: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

36

ILC-SIPM

Signal hold

16 selectable gain factors from 067 to10 VpC

Charge preamplifier

16 CR (RC)2 shapers 40-180 ns shape time

Single channel HV tune

decoupling

ILC HCAL readout chip

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 37: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

37

Trends amp Future

More channels more functionality in the chip (analog+digital) egrave more integration

Detector imbedded electronics egrave reduce cable volume = dead volume egrave ultra-low power consumption

FLC_PHY3 18ch 1010mm 5mWch ILC 100microWch ATLAS LAr FEB 128ch 400500mm 1 Wch

FE chip (1mm) Wafer (400microm) PCB (600microm) Tungsten (1 mm)

Readout chip integrated in active layer (Si-W ECAL for ILC)

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 38: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

38

Imbedded electronics (ILC HCAL)

AHCAL Slab 6 HBUs in a row

HBU HCAL Base Unit 12 x 12 tiles

SPIROC 4 on a HBU

HEB HCAL Endcap Board Hosts mezzanine modules DIF CALIB and POWER HLD

HCAL Layer Distributor

Power 40 microW channel Heat T grad 03K2m Time constant 6 d

Front end ASICs embedded Interfaces accessible

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 39: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

39

More pixels more functionality

36 8bit

5V DAC

Dual DAC Bandgap

3616 Analog

memory

36 Preamp Shaper discri

362 Wilkinson

ADC

SRAM Readout

SPIROC layout (CALICE chip for Analog HCAL readout)

Output multiplexed digitized signals

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 40: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

40

Support material

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 41: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

41

Current preamplifiers bull  Transimpedance configuration

ndash  Vout(ω)iin(ω) = - Rf (1+ZfGZd)

ndash  Gain = Rf

ndash  High counting rate ndash  Typically optical link receivers

bull  Easily oscillatory ndash  Unstable with capacitive detector ndash  Inductive input impedance

Leq = Rf ωC

ndash  Resonance at fres = 12π radicLeqCd

ndash  Quality factor Q = R radicLeqCd bull  Q gt 12 -gt ringing

ndash  Damping with capacitance Cf

bull  Cf=2 radic(CdRf G0ω0) bull  Easier with fast amplifiers

Current sensitive preamp

Step response of current sensitive preamp

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 42: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

42

Resolution and noise

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 43: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

43

Why bother about resolution and noise

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 44: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

44

Electronics noise

bull  Definition of Noise ndash  Random fluctuation

superimposed to interesting signal

ndash  Statistical treatment

bull  Three types of noise ndash  Fundamental noise

(Thermal noise shot noise)

ndash  Excess noise (1f hellip) ndash  Parasitic agrave EMCEMI

(pickup noise ground loopshellip)

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 45: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

45

Calculating electronics noise bull  Fundamental noise

ndash  Thermal noise (resistors) Sv(f) = 4kTR ndash  Shot noise (junctions) Si(f) = 2qI

bull  Noise referred to the input ndash  All noise generators can be referred to the

input as 2 noise generators ndash  A voltage one en in series series noise ndash  A current one in in parallel parallel noise ndash  Two generators no more no lesshellip

Noisy

Noiseless

en

Noise generators referred to the input

n  To take into account the Source impedance

n  Golden rule n  Always calculate the signal before the noise

what counts is the signal to noise ratio

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 46: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

46

Parallel noise

Series noise

Noise in charge pre-amplifiers

bull  2 noise generators at the input ndash  Parallel noise ( in2) (leakage

currents) ndash  Series noise (en

2) (preamp)

bull  Output noise spectral density ndash  Sv(ω) = ( in2 + en

2|Zd|2 ) ω2Cf2

= in2 ω2Cf2 + en

2 Cd2Cf

2 ndash  Parallel noise in 1ω2

ndash  Series noise is flat with a laquo noise gain raquo of CdCf

bull  rms noise Vn

ndash  Vn2 = int Sv(ω) dω2π -gt infin ()

ndash  Benefit of shapinghellip

Noise generators in charge preamp

Noise density at Preamp output

f([Hz]

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 47: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

47

Equivalent Noise Charge (ENC)

Two basic noise mechanisms input noise current in [pAHz] input noise voltage en [nVHz]

Equivalent Noise Charge where Ts = characteristic shaping time (eg peaking time)

Fi Fe Form Factors that are determined by the shape of the pulse (calculated in the frequency or time domain) Ci = total capacitance at the input node (detector capacitance + input capacitance of preamplifier + stray capacitance + hellip )

egrave Current noise contribution increases with T egrave Voltage noise contribution decreases with increasing T only for ldquowhiterdquo voltage amp current noise sources + capacitive load ldquo1f rdquo voltage noise contribution constant in T

from Front End from Shaper

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system

Page 48: Readout Electronics the path of the signal from detection ...garutti/LECTURES/Particle... · Trigger Tower (25 channels) -1 mother board -1 LV regulator board -5 VFE boards (5 channels

48

Coherent noise problem Noise adds linearly instead of

quadritically Particularly sensitive in calorimetry as

sums are performed to reconstruct jets or Et

miss

Σ aisup2 = n σsup2incoh + nsup2 σsup2coh (i=channels)

Coherent noise estimation Perform Direct and Alternate sums to

extract coherent noise SDsup2 = Σ aisup2 SAsup2 = Σ (-1)i aisup2 SAsup2 = n σsup2incoh Incoherent amp coherent noise

σsup2incoh = SAsup2n σsup2coh = (SDsup2-SAsup2)nsup2

Usually σcoh σincoh lt~ 20

Coherent noise in a multi-channel system