A view on electronics for the prototype of the GOSSIP detector in 0.13um CMOS Technology.

Vladimir Gromov

Electronics TechnologyNIKHEF, Amsterdam,

the NetherlandsDecember the 15th, 2004

Highlights. • Main functionalities of the detector and the principal block

diagram of the detector.• Main specifications of the electronics.(Single electron

efficiency, time resolution, power consumption, analog-to-digital compatibility issue).

• A choice of sensitive pad-preamplifier coupling in the very front-end (calculation of the parasitic capacitances).

• Design of the preamplifier in the 0.13um CMOS technology (signal response, noise, hardness to the spread caused by the fabrication process instability).

• Design of the analog part of the read-out electronics in 0.13um CMOS.

• Performance of the detector featuring the design (efficiency, signal time-walk , overall time resolution).

• Block diagram of the DLL-based TDC. • Current-steering logic is a way to eliminate switching noise

in the mixed analog-digital design.• Conclusion.

Preamp

Shaper

Discriminator

Threshold

Latch #1

4-bit DLL 1.6ns

Latch #2

Latch #16

Preamp

Shaper

Discriminator

Latch #1

Latch #2

Latch #16

to Read-0ut Clock 40MHz

Integral circuit in 0.13um CMOS technology

Cathode (drift) plane

Ingrid

Cluster2 Cluster1

Cluster3

Track of the particle

Drift distance

The principal block diagram of the detector.

Z Y

X

Main functionalities of the device:

1) The pixel structure with a fine pitch (30um …50um) can provide accurate information on X-Y coordinate of each cluster on the track. Thus the projection of the track is seen.2) With having measured the drift time of each cluster the angle between the track and X-Y plane can be found in order to depict a 3D picture.

Design objectives on the read-out electronics:

a) the fact that the pixel has been hit needs to be detected with high efficiency and low faulty. The hit should be correctly related to a proper bunch-crossing.b) drift time is to be measured as a latency of the hit arrival time in respect to the bunch-crossing signal accurate enough to determine Z - coordinate of the cluster.

Design objective

the desirable vs the possible

Main specifications of the electronics:

1) Single electron efficiency (input referred electronic noise).The fluctuations in the number of electrons in a single-electron avalanche is given by:

P(n) = 1/M * exp(-n/M) , where M is a gas gain factor.

With the input referred threshold at the level of 500e inefficiency will be 20% (gas gain M=2000)10% (gas gain M=4000)6% (gas gain M=8000) .

The threshold of 448e corresponds to ENC = 90e RMS

20%

10%

0 500 1000 1500 2000 2500 3000 3500 4000 4500 50000

0.2

0.4

0.6

0.8

11

0

P n 2000( )

P n 4000( )

P n 8000( )

n

50000 n n n 448

Gain=4000

Gain=2000

Gain=8000

Inefficiency

Threshold, electrons448e

2) Time resolution.Both the time resolution of the TDC and the time-walk in the discriminator are

independent contributors to the overall time resolution (σΣ ) of the electronics: σΣ = √ (σ2

TDS + σ2Time walk) ,

where σTDS = ∆t/√12 , ∆t – minimum bin size of the TDS

σTime walk is dispersion, related to the time-walk in the discriminator

For a 4-bit DLL-based TDC ∆t = 25ns/16 = 1.6ns it yields σTDS = 0.46ns .The electron drift velocity in the gas is about 20ns/mm therefore the TDC contribution to

the overall spatial resolution will be σspatial

TDS = 23um.

Under these conditions time walk in the discriminator become the main contributor.

Time-walk. Where does it come from?

Signal at the output of discriminator (Sout(t)) is a convolution integral of input current i(t) and pulse response function (H(t)) of the electronics.

0 20 40 60 8020

15

10

5

00

19.355

S t 1 10( ) 448

13

f t 1 10( ) 250

5

900 t

Low threshold

High threshold

Time, ns

Sout(t)Signals at the

input of discriminator,

arbitrary unit

Range of time-walk for the fast shaped signal

Range of time-walk for the slow shaped signal

A fast shaped signal

A slow shaped signal

Sout t( )0

tH ( ) i t ( ) d

i(t) - input current :Ion current occurs in the Micromegas-pad gap in the period

∆tion= (∆L)2/μ U ≈ 30ns, where ∆L ≈ 50um is the Micromegas-pad distance

U ≈ 400V is Micromegas-pad voltageμ= 1.72cm2V-1 sec-1 is mobility of ions in Argon

Single-electron current in the detector

20 0 20 400.15

0.1

0.05

0

s t( )

t

20 0 20 401.5

1

0.5

0

S t1( )

t1

Integral of current induced by a single electron.

time,ns time,ns

i(t)

electron component

ion component

10% is electron contribution to

theoverall charge 90% is ion

contribution to theoverall charge

H(t) - shaping function (δ-pulse response)

Let us take shaping function of the electronics as follows F(p)=1/[(p 1 + 1) (p 2+1)]. It demonstrates pulse response f (t, 1 2 )

f t 1 2( ) 11 2( )

exp t1

exp t2

0 10 20 30 40 50 600

0.02

0.04

0.06

0.08

f t 1 10( )

f t 8 10( )

ttime,ns

Pulse response of the electronics (τ1=1ns, τ2=10ns)

Pulse response of the electronics (τ1=8ns, τ2=10ns)

Distribution of the threshold-crossing time. Monte-Carlo simulations.

0 2 4 6 8 100

200

400

600

Integrals of the distributions.

0 2 4 6 8 100

5000

1 104

EntriesGain=2000, Thr = 448e, τ1=1ns, τ2=10ns

Gain=2000, Thr = 448e, τ1=8ns, τ2=10ns

time,ns

time,ns

Entries Gain=2000, Thr = 448e, τ1=1ns, τ2=10ns

Gain=2000, Thr = 448e, τ1=8ns, τ2=10nsinefficiency=20%

!!! Fast shaping enables us to get much better time resolution at a given gas gain (threshold).!!! Fast shaping enables us to get much better time resolution at a given gas gain (threshold).

3) Power consumption.

Number of channels per wafer = π D2/(4 pitch2) = 3.14 * 106

with power consumption 10W/wafer (possible to cool it down by gas flow) !!! Power consumption per channel = 3.2uW !!! Power consumption per channel = 3.2uW

4) Switching noise.

In a mixed-mode design switching noise coming from digital part of the circuit back to high sensitive analog front-end is a very important issue. The most most common way to eliminate switching noise is using current-steering logic. Although it reduces speed and increase static power consumption.

D=10cm

pitch =50um

Preamp Preamp

Ingrid

Preamp

i(t)

C1

C2C2 C3 C0 ~ 20pF

R0

U1=-300V…-400V

C4

R4

-800V

Cathode (drift) plane

R1

 The very front-end. DC or AC coupling to the Preamp.

 The very front-end. DC or AC coupling to the Preamp.

Safety DC-coupling AC-coupling

C0

C3 4*C2

Zin≈0Zin/4≈0

Qd=U1*C0

Discharge trajectory

C0

C34*C2

C1

Zin≈0

Zin/4≈0

R1

Qd=U1*C5

Discharge trajectory

C5=C0*C1/(C0+C1) ≈ C1 C1 C0

i(t)

C0

C3 4*C2Zin≈0

Zin/4≈0

i1(t) iin(t)

i(t)

C0

C34*C2

C1

Zin≈0Zin/4≈0

i1(t) iin(t)

R1

In order to collect much of the charge iin(t) ≈ i(t) the following condition must be met C1 C3+4*C2.For better safety C1 0. Therefore values of the parasitic capacitors C3,C4 are important to know.

Signal collection DC-coupling AC-coupling

C R d( ) 2 3.14 0

0

dz

0

1rr z

z2 r2 R2

32

d d

0 5 10 6 1 10 5 1.5 10 5 2 10 5 2.5 10 50

0.5

1

1.5

2

2.5

C R 25 10 6

C R 50 10 6

C R 75 10 6

R

C=1.8fF whenR=25um, d=50um

Pad-to-Micromegas grid capacitance calculations.

R - is a radius of the pad. The pad is a circle. d - is pad-to-Micromegas distance.0 - is vacuum dielectric constant.

DIdeal boundless plane

Ideal uniformly charged disk

R

0 5 10 6 1 10 5 1.5 10 5 2 10 5 2.5 10 5 3 10 50.1

0.2

0.3

0.4

0.5

0.6

0.645

0.111

C1 b( )

3.5 10 55 10 6 b

C=0.62fF whenb=30um, a=20um,pitch=50um

Conclusion: C3+4*C2 = 1.8fF +4*0.62fF = 4.32fFConclusion: C3+4*C2 = 1.8fF +4*0.62fF = 4.32fFIn order to collect much of the charge iIn order to collect much of the charge iinin(t) ≈ i(t) the following condition must be met C1 4fF(t) ≈ i(t) the following condition must be met C1 4fF

C1 b( ) 4 3.14 0 r

2 10 10

a b( )2

z

0.5

0.5xz

z2 x2 b2

32

d d

a

b

b

Ideal uniformly charged square padIdeal uniformly charged square pad

b - is length of the pad. The pad is a squire . 50um is a pitch[50um-b] - is a pad-to-pad distance.0 - is vacuum dielectric constant.r=4 - is relative permittivity of the dielectric.

Pad-to-Pad capacitance calculations.

Input Output

T249Id= -1uGm = 2.4uGds = 26nVgs= -963mVVds=-957mV (Vds_sat=-658mV)Cgg= 57.7fFCdd+Cjd=0.451fF+0.150fF=0.6f

T245Id= 1uGm = 23.2uGds = 0.7uVgs= 234mVVds=243mV (Vds_sat=45mV)Cgg= 2.6fFCdg=0.8fFCdd+Cjd=0.8fF+0.7fF=1.5f

Schematic of the Preamplifier. Main specifications.Technology: 0.13um CMOS.Power supply voltage:1.2VPower consumption: 1uA * 1.2V = 1.2uW.Charge sensitivity (real detector current pulse): 33mv/448e. Shaping function: rise time is 6ns, decay time is 100ns.Output noise (RMS): 4.3mVEquivalent input noise (RMS): (4.3mV/33mV)*448e = 58e

Cdg=0.8fF100MΩ

Idc=6nA

Input signal is a current δ-pulse

Output signal as a response to the δ-pulse

Input signal is a real current pulse

Output signal as a response to the real current pulse

Spectral density of the squire of the noise output

voltage|V2

n(jw)|

18.5uV2

4.3mV

The preamplifier. Simulation results.

Input OutputIdc

The preamplifier. Monte-Carlo analysis in Affirma Spectre.

Channel-to-channel variations of the bias current Inom=1uAσ=5%

Channel-to-channel variations of output voltage Unom=247mVσ=6.5%

Channel-to-channel variations of gain (charge sensitivity) GAINnom=33mV/448eσ=8.7%

Schematic of the Preamp + Shaper + Discriminator. Main specifications.Technology: 0.13um CMOS.Power supply voltage:1.2VPower consumption: 1.6uA * 1.2V = 1.92uW (3.3*106 channels per wafer or 6.3W per wafer) .Charge sensitivity (real detector current pulse at the shaper output): 254mv/448e. Shaping function: rise time is 23ns, decay time is 100ns.Output noise (RMS): 37mVEquivalent input noise (RMS) at the shaper’s output: (37mV/254mV)*448e = 65e

Input

Output240nA

100nA

300nA1000nA

Input current

Preamp output

Shaper output + threshold

Output of the first stage of the discriminator

Output of the discriminator

Channel-to-channel gain variations at the output of the shaper GAINnom=254mV/448eσ=10%

Channel-to-channel variations of the voltage at the output of the shaper σ = 18mV Uthr=190mV.

Preamplifier + Shaper + Discriminator. Simulation results and Monte-Carlo analysis in Affirma Spectre.

Walk-time as a function of the signal amplitude (THR=448e)

0.8*448e

1*448e1.5*448e

2*448e3*448e

5*448e

12*448e

20*448e

50*448e

Preamplifier + Shaper + Discriminator. Statistical analysis.

0 5000 1 104 1.5 104 2 104 2.5 104 3 1040

10

20

30

34

0

P n 2000( ) 32

P n 4000( ) 32

P n 8000( ) 32

n

Twalk jt 1

Tw n( )

300000 n n n 448 Twalk jt 0 448 n

Time-walk vs pulse height distribution

Signal, electronsTHR=448e

Gain=8000

Gain=4000

Gain=2000

Time-walk curve

0 5 10 15 20 25 30 35 400

1000

2000

3000

40004000

0

DISTRT2000 nt

DISTRT4000 nt

DISTRT8000 nt

1

nt

nt1

DISTRT2000 nt125

=

1

nt

nt1

DISTRT4000 nt125

=

1

nt

nt1

DISTRT8000 nt125

=

39.90 inttnttime,ns

Entriesefficiency=100%

Time resolution (time distribution of the threshold crossing events). Statistics is 10000.

Gain = 8000

Gain = 4000

Gain = 2000

Preamplifier + Shaper + Discriminator. Statistical analysis. Time resolution

!!! It is feasible to reach time resolution of order of σ=2ns (100um) with a realistic gas gain.!!! It is feasible to reach time resolution of order of σ=2ns (100um) with a realistic gas gain.

Phase detector

Clk40MHz

Delay chain. DLL

#1 #2 #3 #4 #16

Block diagram of the DLL.

Vdd=1.2V

bias1

bias2

bias3

In +

In -Out+

Out-

An inverter in low-voltage current-steering logic.

±200mV

Conclusion..The TimePix detector is going to be a powerful tool for future experiments.. Definition of the topology and specifications of the detector is in progress on the basis of

the potentialities of the modern deep sub-micron CMOS technology . The following specification have been found feasible so far:

Gas gain: 2000-8000. Single electron efficiency: 80%-94%. Input referred threshold: 500e. Time resolution: σ = 2ns corresponding to spatial resolution σ = 100um. Power dissipation: 3.2uW/channel (10W/wafer). AC coupling to the preamplifier looks preferable from safety point of view. Not

much of the signal will be lost if the coupling capacitor is as tiny as 30fF…40fF. .First trial to design an analog circuit in the 0.13um CMOS technology capable to meet the

specification has shown a promising result.. DLL-based TDC structure is a possible candidate for time-to-digital conversion block. . More efforts needs to be made to design switching noise free logic cells.