FGT time shape analysis

 FGT Time Dependence of the Raw Signal - Initial Observations

 

Currently, I am looking at the cosmic test data with signal to look at the time shape of the signal. For no good reason, I'm only looking at the first 1000 events.

For now, I do not do a proper pedestal subtraction. Instead, I simply take the lowest ADC count of the 7 time bins in any given events, and subtract it out from all time bins (per Steve's suggestion). Therefore at least one of the time bins will have zero count. This lowest count approach seems to be not outrageiously wrong, and for the purpose of time depdence it's likely not a bad place to start.

 

Fig. 1 APV vs. Minimum ADC distribution

 

Fig. 2. APV vs. ADCMAX - ADCMIN

 

I've tried fitting the raw ADC vs. time bin distribution using the following simple function, per Gerard's suggestion.

f(t) = [0]*x*x*exp(-x/[1])

As for the error, I am using 35 count average pedestal width scaled by the ADCMIN I use in place of the pedestal relative to the actual average pedestal value of 745 count (Anselm's note). I then add simple sqrt(ADCMAX - ADCMIN) to this "pedestal error" in quadrature. 

The following plots are for the events whose maximum ADC count is more than 700 counts higher than the minimum ADC. For these events, the chi-squared distribution looks like this. 

 

Fig. 3. Chi-squared distribution for events with ADCMAX - ADCMIN > 700 counts 

 

 

First, events with "good" fits = chi^2/dof < 1. 

 

Fig. 4. The first 40 events with chi^2/dof < 1., ADCMAX-ADCMIN>700, with no repeat on channel numbers. Also, the maximum location of the fit is within the first 4 time bins. (PDF)

 

Fig. 5. The same as above, but the maximum location of the fit is beyond the 4th time bin. (PDF)

 

Fig. 6. APV vs. t_max (time at which the fit function is maximum) distribution.

 

 

Second, events with "bad" fits = chi^2/dof > 30.  

 

Fig. 7. The first 40 events with chi^2/dof > 30., ADCMAX-ADCMIN>700, with no repeat on channel numbers. Also, "high-count" less than 4 (explained later). (PDF)

 
 
These are again just the first 40 of each type of events, but so far even the "bad" fit events seem mostly ok. The only obvious pathology I've found so far is that on rare occasions, we have multiple time bins all having very high values. If I select the events where more than 3 time bins (other than the one giving ADCMAX) had within 10% of the maximum ADC count, they look like the following. 
 
 
Fig. 8. Odd events where more than 3 time bins have ADC count very close (>90%) to that of the maximum ADC. (ADCMAX-ADCMIN>700) (PDF)
 
 
 
Fig. 9. ADCMAX - ADCMIN distribution for these odd events. 
 
 
 
Fig. 10. For ADCMAX - ADCMIN > 350, fraction of these events vs. APV