Pedestal Shift and High Side Tail Correlation

1
Pedestal Shift and High Side Tail Correlation

• Melanie Day
• University of Rochester
• 6/25/09

2
Background

• Pedestal
• ADC Counts seen with no signal
• Usually 300-400 ADC counts
• High Side Tail-Counts after Pedestal, may be
  signal, may be cross talk
• Expect signal to cause pedestal shift
• Read one board on mapper, 64 channels, 50
  positions, CS 137 source

3
Initial Methods

• Assume signal causes at least 50 ADC ped shift
• Assume pedestal width 50 ADC Counts
• Use Jaewons fit to find pedestal mean
• Use positions with no signal for initial pedestal
  value, average five values
• TailSum Sum(ADCBin Events) for all events
  after PedMean 50

4
Preliminary TailSum vs. PedShift
5
Suggested Tweaks

• Jaewon suggests channel mapping
• 50 ADC shift could be crosstalk
• No channel mapping done for current setup
• Bob suggests better signal fitting program
• Program slower
• Poor fit for pedestal with no signal
• Possibly use parameters to more precisely measure
  beginning of tail. A gaussian is
• Ae-(x-µ)2/(2s2) so the integral gives
  Assqrt(2?), the estimated number of ADC counts
  under the gaussian

6
Deviation from Gaussian Pedestal Measured Events
7
Implementing New Fit

• Fit is slow, and innacurate for pedestal
• Made if loop, simple JaeWon loop for PedMeangt390,
  complex fit for PedMeanlt390
• Probably better way to distinguish?
• Shortened fit time from 2 hours to 30 mins
• Found Jaewons simple fit works very poorly for
  large signals because of .1PedMean fit range
• Complex fit gives larger PedShift

8
JaeWons Fit With High Signal
9
Preliminary TailSum vs. PedShift with Complex Fit
10
Fit Using Estimated End of Gaussian
11
Gauss Integration Fail

• Almost always underestimates sum
• Indicates integration is overestimating events
  under pedestal
• Hypothesis Estimated gaussian extends into
  signal region
• Still, doesnt quite explain extent of disparity,
  may be bad sigma or amplitude fit
• Would simple JaeWon fit give better accuracy?

12
Estimated Gaussian On Fit
Presumed Gaussian Range
Signal in Range
13
Other attempts

• Tried
• Estimating minimum between PedMean and PedMean
  100 as beginning of signal
• Estimating change of slope from highly negative
  to an average of zero as beginning of signal
• As minimum overestimates and slope average
  underestimates, tried average of the two
• None of these as good as preliminary guess of
  PedMean 50

14
Future

• Channel Mapping?
• Give up on complex fit?
• Try and increase accuracy of simple JaeWon fit
  for high signal?
• Other method for finding beginning of signal?

15
The End