1

MGPA Linearity

Mark Raymond (Dec.2004)

Non-linearity measurements in the lab

hardware description method results

2

Linearity test bench

Digital Sequencer

AWG(AG33250)

programmableattenuator

(HP8494+HP8495)1 db steps

MGPA 14-bit ADC(SIS3301)

prog. delay(0 – 250 ns1 ns steps)

trig.

trig.10ns

AWG

prog.attenuator

14-bitVMEADC

MGPAtest

board

prog.delay

levelshift

LabView automated

use prog. attenuator, ratherthan changing amp. of AWG O/Pto avoid contribution from AWGO/P stage linearity

14-bit ADC -> negligible INLand DNL contribution tomeasurement

for MGPA test board use either RAL PCB with socket easy chip exchangeor VFE card verify performance in final configuration

40 MHz clock and ADC trig.

Cinj

3

MGPA test board photosRAL board with socket

modified VFE card (latest version)

APD inlight tightenclosure diff. to s.e.

conversion

differentialMGPA O/Psignals accessedbetween MGPAand ADC

chargeinjection

4

Linearity measurement procedure

Digital Sequencer

AWG(AG33250)

programmableattenuator

(HP8494+HP8495)1 db steps

14-bit ADC(SIS3301)

prog. delay(0 – 250 ns1 ns steps)

trig.

trig.

levelshift

Digital Sequencer

AWG(AG33250)

programmableattenuator

(HP8494+HP8495)1 db steps

MGPA 14-bit ADC(SIS3301)

prog. delay(0 – 250 ns1 ns steps)

trig.

trig.40

MHz10ns

levelshift

1) remove MGPA test board and characterise system => make precision measurement of attenuator steps

50ns CR-RC shape

2) replace MGPA and measure pulse shapes for different signal sizes

pulse shapes measured with 1ns resolution by sweeping time of charge injection using prog. delay

40 MHz clock and ADC trig.

40 MHz clock and ADC trig.

Cinj

5

Typical non-linearity measurement (1)

-1.0

-0.5

0.0

0.5

1.0

[Vol

ts]

200150100500

time [1 nsec. steps]

1.5

1.0

0.5

0.0

200150100500

time [1 nsec. steps]

subtractpedestals

high gain channel (VFE card)180 pF Cin addedfullscale signal = 5.8 pCsignal sizes not linearly spaced becauseof logarithmic attenuator steps

AD

C ra

nge

5.8 pC

6

1.5

1.0

0.5

0.0

200150100500

time [1 nsec. steps]

2.0

1.5

1.0

0.5

0.0pe

ak s

ampl

e va

lue

[vol

ts]

1.00.80.60.40.20.0

attenuator steps

peak samples fit to peak samples

take peaksamples

and fit

Typical non-linearity measurement (2)pedestal subtracted pulse shapes peak sample vs. attenuator value

1.0 correspondsto 5.8 pC here

7

2.0

1.5

1.0

0.5

0.0

peak

sam

ple

valu

e [v

olts

]

1.00.80.60.40.20.0

attenuator steps

-0.2

-0.1

0.0

0.1

0.2

non-linearity [% fullscale]

peak samples fit to peak samples non-linearity

finally calculate linearity

Non-linearity [% fullscale] =

peak pulse ht. – fit (to pk pulse ht) x 100 fullscale signal (1.8 V)

Typical non-linearity measurement (3)

8

Pulse shape matching

Pulse Shape Matching = (PSMF – Average PSMF) Average PSMF

Average PSMF = average over all pulse shapes

-1.0

-0.5

0.0

0.5

1.0

[Vol

ts]

200150100500

time [1 nsec. steps]

can also calculate pulse shape matching forsame data

PSMF = peak sample value / sample 25ns before

-2

-1

0

1

2

PS

M [%

]

1.00.80.60.40.20.0attenuator steps

9

Some non-linearity results (old)

-0.2

0.0

0.2

6040200

-0.2

0.0

0.2

12840

-0.2

0.0

0.2

6543210

-0.2

0.0

0.2

6040200

-0.2

0.0

0.2

1086420

-0.2

0.0

0.2

6543210

MGPA Version 1 MGPA Version 2

10 chips measured for eachMGPA version

v. similar results V1 cf. V2

nonlinearity within (or close to)± 0.1% specification

these measurements madeon RAL card using simulatedAPD capacitance

Non

linea

rity

[% fu

llsca

le]

Non

linea

rity

[% fu

llsca

le]

charge injected [pC] charge injected [pC]

high gain range

mid

low

mid

low

high

10

-0.2

-0.1

0.0

0.1

0.2

non-

linea

rity

[% fu

llsca

le]

1.00.80.60.40.20.0

attenuator steps

180 pF APD 180 pF, 5 ns risetime

New results

Why is non-linearity worse in H4 beam? (noticeable fall-off for small signals on high gain channel)

Could there be an APD effect?

No – no significant difference between180 pF and APD biased at 350 V

Could signal risetime be shorter than expected?

No - changing charge injection 10ns -> 5ns doesn’t make much difference (at least for high gain channel)

Hi gain channel non-linearity measured on VFE card

11

Final thoughts

non-linearity measured using procedure described here does not explain test beam results results here within or very close to +/- 0.1% spec.

method used here will not include any contribution from ADC, but don’t expect to be significant

could test beam linearity studies be more sensitive to pulse shape?

PSM within design spec. but does tend to show similar trend to test beam linearity plots

-2

-1

0

1

2

PS

M [%

]

1.00.80.60.40.20.0attenuator steps