Common mode and ADC data simulation part 2: fjlters SAMPA Tests - - PowerPoint PPT Presentation

Konstantin Münning Universität Bonn Helmholtz-Institut für Strahlen- und Kernphysik 10.12.2014

Common mode and ADC data simulation

part 2: fjlters SAMPA Tests Meeting

2

Simulation of common mode efgect

• the simulation of the common

mode efgect is based on superposition of all charges arriving at the pad plane attenuated by the capacitive coupling factor

• the baseline shift is a result of

the pile-up of the charge signals

• at suffjcient rate this shift and

its fmuctuations become signifjcant

3

Filtering

• to reduce the impact of the baseline shift a digital

fjlter, the baseline correction circuit, has been implemented in the data path of the ALTRO chip:

• Filter performance needs to be evaluated for an
• ptimal choice for the upcoming SAMPA chip

from ALTRO manual DRAFT 0.2, CERN 2002

4

• to evaluate the quality of the fjlter a measure of its

efgect is needed

• approach: using well known input signal spectra and

measuring the resulting deterioration due to the baseline shift and fmuctuations without and with fjlter

• here: single line spectrum of peaks with 60 ADC

amplitude and 160ns peaking time with random time distribution

• evaluation of peak areas
• comparison with ideal values and theoretical values

Evaluation

5

Parameter extraction

• usually peak parameters are extracted by pulse

shape analysis → without proper selection and tuning results may be misleading

• as the data is simulated using the intrinsic

information when peaks signals are generated is the most reliable peak identifjcation

• overlaps and peak multiplicity

can easy be detected

peaking time of 160ns, signal amplitude 60ADC, pedestal 10ADC simulation without common mode efgect with 10MHz sampling rate

ROI

• verlap

6

Peak areas without common mode

• ideal condition, just the peaks, no common mode
• peak area is calculated in a fjxed region relative to

peak synthesis time

• reference for best achievable result

with this setup

analytical value

7

Peak areas with common mode

• impact and magnitude of baseline shift on peak

area is clearly seen

analytical value

8

AL TRO Moving Average Filter

• average of 8 samples
• upper and lower threshold set to 4 ADC
• baseline is restored for all tested charge rates
• constant overcorrection depends on

threshold value

9

Alternative: slope based fjlter

• slope based fjlter is following directly the signal but

is limited by maximal slopes for rising and falling

• has some advantages compared to ALTRO MAF

upward slope downward slope

simplifjed functional description of slope based fjlter

10

Slope based fjlter

• upward slope 1/20, downward slope 1/1
• baseline is restored for all tested charge rates
• lower noise
• slight overcorrection depends on charge rate

and slope parameter

11

Direct fjlter comparison

12

Hypothesis

• fjlter quality can be measured by comparing the

restored baseline to the analytical baseline value

analytical value difgerence to analytical value

13

• with given parameters both tested fjlters efgectively

remove the baseline shift

• advantages of moving average fjlter: well tested,

known behavior

• advantages of slope fjlter: lower noise, easy to

implement in hardware → low energy consumption, self-healing → no start-up sequence needed

• results of baseline comparison support hypothesis →

easier method to analytically evaluate fjlters

• discussion of fjnal fjlter choice necessary

Conclusion

14

• add formulas and numbers
• optimize parameters, try known threshold values,

try with realistic peaking times

• use random signal amplitudes or specifjc spectra
• add noise and distortion
• simulate with fast changing charge rates
• add PSA and try recorded detector data
• Further more exotic tests are also possible like

straying random bit fmips in the fjlter registers to simulate radiation induced errors

Further steps

15

Comments? Questions?

16

Spares

17

Data reduction with slope based fjlter

• slope based fjlter allows data reduction by easily

extracting peak parameters by on-chip DSP

• instead of transmitting ADC values, only few data

words need to be transmitted

• data amount is not scaling with ADC sampling rate
• easy correction of missing areas/amplitudes
• new noise/zero suppression schemes possible

peak start peak area peak amplitude

18

fjne structure of peak parameters

• when restoring peak parameters from digital data

with simple means, i.e. without fjtting functions, a fjne structure is revealed when the amplitude resolution is not limited