1. Pulse Shape Fitting Beam Test September, October 2007 @ CERN
EMCal Meeting, Nantes
Raphaelle Ichou
July,

2. Outline 8x8 Prototype EMCal towers Fitted pulse shape distribution
- The fit quality
- The fit parameters: Tau
High-Low gain correlation
AliRoot : v4.13.Rev-02
July,
Raphaelle Ichou

3. Fitted pulse shape distribution

4. The fit procedure A  P  tmax t0
AliEMCALRawUtils::RawResponseFunction()
5 parameters: (P, A, tmax, , N)
AliEMCALRawUtils::FitRaw()
Initialisation:
Pedestal :
tmax: Max (ADC)
N: fixed at 2 (Gain in CPU time: Factor 2 )
Fit range:
Min: time=0
Max: time=time(ADC< P+3)+5=22
Saturated channel: if ADC >1023, points not taken into the fit A t0 
signal
Fit  P
tmax
July,
Raphaelle Ichou

5. Typical fitted pulse shape 4x4
Saturation: amp > 1023
Run 190
Electron beam 80 GeV/c
At the center of 4 tower corners
July,
Raphaelle Ichou

6. Typical fitted pulse shape 2x2
Run 196
Electron beam 80 GeV/c
Run 202
Electron beam 80 GeV/c
Run 190
Electron beam 80 GeV/c
July,
Raphaelle Ichou

7. The fit Quality

8. The fit « Quality » 2 not significant (no statistical distribution)
Evaluate the difference (%) between ADC counts and the fit function:
A new variable defined : Q
July,
Raphaelle Ichou

9. Quantify the fit Quality
Run 190
1 tower : A ~ 600
signal
Fit Q
Here : Q < 15%
Time(x 100ns)
July,
Raphaelle Ichou

10. Fit Quality vs amplitude(Low gain)
Mean Q
(0 < A < 200)
Q depends on A
Q < 6%
=> Good fit quality
July,
Raphaelle Ichou

11. Fit Quality vs amplitude(High gain)
Mean Q
(200< A <1050)
Q < 6% in Low gain
Q < 15% in High gain
=> Good fit quality (low amplitudes)
July,
Raphaelle Ichou

12. Fit Quality vs Fit Range
Improve the fit Quality (high amplitudes):
Choose a smaller fit range
Min=0
Max={90%-5%}A
90%
20% 5%
July,
Raphaelle Ichou

13. Fit Quality vs Fit Range
Mean Q
A Max for the fit range at 20% A :
Mean Q smaller than 4%.
Q < 4%
Q > 4%
% A %
90 %
80%
70%
60%
50%
40%
30%
20%
July,
Raphaelle Ichou
15%

14. Fit parameters vs Fit Range
tmax A
% A
% A
% A
80 %
30%
80 %
30%
80 %
30%
Fit parameters not affected by the fit range from 80% to 30% A
Larger fit range : parameters variation< 1%
=> Parameters not sensitive to the fit range
July,
Raphaelle Ichou

15. Fit Quality vs Amplitude
Mean Q
Using a fit range with a Max at 30%
High gain (200<A<1050)
Better mean Q
with a Max at 30% A
July,
Raphaelle Ichou

16. Conclusion on the fit Quality
Good fit quality (Q < 16%)
Can be improved using restrained Fit Range at 30% of the Amplitude (Gain in CPU time: 5-15%)
Small influence on the fit parameters
July,
Raphaelle Ichou

17. Study of 
 : decay constant, property of the signal shaper (electronics)
 fixed -> more stable fits
(For the moment  fixed for small amplitudes signals)

18.   studied for each tower:  vs Amplitude (High gain)
2.2<  <2.5 Quite constant
Low dispersion (bigger for lower amplitudes)
July,
Raphaelle Ichou

19.  for all towers
Mean
17 Runs used, first periods of SPS :
[190, 202, 188, 192, 196, 217, 221, 245, 285, 372, 387, 390, 518, 875, 177, 219, 227]
2000 events in each
Between towers: similar  values (2.33<  <2.4)
Similar results than David
Can be fixed in the fit for each tower?
(Gain in CPU time :
10-20%)
July,
Raphaelle Ichou

20. High-Low gain correlation
From which amplitude can we use low gain rather than high gain?
(Saturation at ADC =1023)

21. High vs Low Gain
17 Runs used, first periods of SPS :
[190, 202, 188, 192, 196, 217, 221, 245, 285, 372, 387, 390, 518, 875, 177, 219, 227]
2000 events in each
A High
(For the digits: we use the High gain except when A > 950)
Here :Good correlation for A < 1000
A Low
July,
Raphaelle Ichou

22. High-Low Gain ratio Ratio 16 <Ratio < 16.8
Similar results than David
tower
July,
Raphaelle Ichou

23. Conclusion Fit : Good fit Quality
Can be improved using restrained Fit Range
But small influence on the fit parameters and on CPU time.
Study of  : Quite constant values for all amplitudes in each tower and similar values for all towers (small dispersion)
Not so much place for improvement
Study of the High-Low gain correlation:
Good correlation (factor ~ 16.4) up to 1000.
July,
Raphaelle Ichou

24. Back-up slides

25.  vs Event number 
July,
Raphaelle Ichou

26. High vs Low Gain
A High
July,
Raphaelle Ichou
A Low