1. Solving pedestal problem
Mauro Raggi
Sapienza Universita’ di Roma

2. Observing no beam time run654
Observed the pedestal in events in between and when linac is switching from electrons to positrons
Try to compare the prediction of Average of 100 samples to the average of all 1000 samples.

3. Different pedestal evaluation
Average of 1000 sample much more stable within groups of events and in between the two samples.
Shape of the red distribution (Avg100) much broader and with different shape in the two samples
Average of 1000 better estimate of the pedestal!
2000<evt<3000
9000<evt<10000
Avg 1000 samples
Avg samples
Avg 1000 samples Avg 100 samples

4. Effect on the charge
Converted the difference into Charge to see if it was the right scale for Qped summing on 1000 samples:
Qdiff=(-PedCh[7]+MeanCh[7])/(4096*1.)/50*1.e-9/1e-12*1000.
Wow seems of the right scale!

5. Correct the effect Reconstruction Program Offline program
Define a sample of pseudo pedestals by asking: abs(Avg100- Avg1000)<3 counts
Fill histograms with Avg1000
Offline program
Fit Avg1000 histogram with gaussian
Get ped value in counts for each channel
Reconstruction program
Define a pedestal vector PedCh[ch] and fill it with fitted value
SamRec[s] = (Double_t) (chn->GetSample(s)-PedCh[ch])/4096*1.;
QCh[ch]+= - SamRec[s]/50*1E-9/1E-12;
Residual effects
The precision on the pedestal in counts is ~1/sqrt(12) cnt going to charge:
charge we get for 1000s: 0.3/4096/50*1e-9/1E-12*1000 = 1.5 pC rms

6. No more double pedestal peak

7. Fitting the pedestal Ch QMax Qfit PedQ 0 = 1.375 1.54139

8. Zero of the pedestal distribution

9. Improving the correction
Turn the Pedestal into a double variable
You can bin the histogram well below the 1 count level thanks to the average and fit the pedestal distribution with a gaussian
Use floating points values for the pedestal in the computation.

10. Pedestal fits for run 669 Ch hMax pedfit Ped 0 = 3769.65 3769.59
We most probably don’t need to fit at all the Histogram max works very well.

11. Improved pedestal fit run 669
Gaussian pedestal with average 1.3 pC sigma
Mean pedestal is at zero with 0.2 pC precision!

12. Time stability tests run 669
Time dependent oscillations of the order of a fraction of a count can be observed. They reflect on different values of the pedestal in different runs.
Further sub pC threshold in charge maybe achievable by following time variation inside the same run.
Ch7 run 669

13. Pedestal zero run 661
Using pedestal calibration from 669 into 661 a worst quality of the pedestal 0 is obtained.
Pedestal sigma still of the order of 1.5pC.
Need run by run calibration of the to get the best zero alignment.
Mandatory if a zero suppression has to be applied at LvL1

14. PED trend ch7 run 661
Real fit value

15. Linearity

16. Total Charge 100 MeV after pedestal calibration
Very good spectrum for the low energy photon with reasonable energy scaling. End point at around 20 MeV
Low energy photon spectrum

17. Applying zero suppression
Effect of a 5 pC ~4s zero suppression on each of the crystals produces a remarkable improvement of the resolution.
With proper calibration of the pedestal zero suppression is a useful tool at low energy

18. Out of run pedestal distribution
2000<Nevt< || 8000<Nevt <12000 || 32000<Nevt<33000 || 44500<Nevt <46500"
Data coming from different times but still 1.1 pC rms
Pedestal is stable in the same run!

19. Predicted co60 spectrum
Co60 spectrum obtained by calculating response for 1.1 and 1.3 MeV photon with 13.8 pC/MeV introducing a fluctuation of 1/sqrt(100*E) on the photoelectron.
Data pedestal
Expected Co60 signal with 100 pe/MeV

20. Conclusion
The average of 1000 samples in empty events provides a stable pedestal
Stable in time
Very small 1.3pC RMS (~100 KeV) and no double peak.
The linearity seems good and the spectrum of low energy photons shows the expected drop with energy.
end point of the spectrum around 20MeV.
With new pedestal I expect that Co60 source can be used for calibration porpouse.