preshower calibration

30
Preshower calibration S. Monteil, K. Sobczak 1/09/2010 1 01/09/2010 PRS calibration

Upload: macaulay-velazquez

Post on 02-Jan-2016

37 views

Category:

Documents


1 download

DESCRIPTION

Preshower calibration. S. Monteil, K. Sobczak 1/09/2010. Current status. Current status. A short reminder of the method: A distribution of ADC counts for tracks pointing towards a cell. Corrections for particle track length in side the cell applied. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Preshower calibration

PRS calibration 1

Preshower calibration

S. Monteil, K. Sobczak1/09/2010

01/09/2010

Page 2: Preshower calibration

PRS calibration 2

Current status

01/09/2010

Page 3: Preshower calibration

PRS calibration 3

• A short reminder of the method:– A distribution of ADC counts for tracks pointing towards a cell.

Corrections for particle track length inside the cell applied.– Fit by a convolution of Landau and Gaussian. MIP position is

the most probable value of Landau.– Uniformization inside boards with numeric gains (range [1,2])

and between boards with HV settings.

Current status

First 2010 april data

01/09/2010

Page 4: Preshower calibration

PRS calibration 4

Software point of view: yes• The last calibration was applied on

02/03/2010 using 2009 data (MIP position at 15 ADC counts).

• The detector is not uniformized (plots below):– 10% calibration inside regions,– the difference between middle and outer is

9% for the A side and 7% for the C side,– the difference between inner and outer is

16% for the A side and 12% for the C side.

• Corrective factors obtained in April applied into the software (thanks to Olivier).

• Calibration performance proved by energy flow method (thanks to Aurelien, the next slide is from his presentation in April) at the ~5% level.

Current status

Detector point of view: no

Is the detector calibrated?- It depends...

What can be improved?

01/09/2010

Page 5: Preshower calibration

PRS calibration 5

2010 real data : EFlow vs Landau

Inner C

Inner A

Middle C

Middle A

Outer C

Outer A

ρ = 64% ρ = 59% ρ = 58%

ρ = 77% ρ = 70% ρ = 54%

Correlation of the coefficients extracted from the EFlow using MagOff data (x-axis) and the Landau fit (y-axis) given by Stephane.

X and Y axis are in units of standard deviations.

A. MARTENS – Calorimeter WG meeting - 22/04/2010

The Mip calibration is ”contained” in the EF calibration which intrinsic precision is 5%.

01/09/2010

Page 6: Preshower calibration

PRS calibration 6

What is the precision of the current software calibration? Take May data as an example, split by BXID.

[MPV(meas)-MPV(average)]/MPV(meas.)

Current status

01/09/2010

Page 7: Preshower calibration

PRS calibration 7

Same for odd.

[MPV(meas)-MPV(average)]/MPV(meas.)

Current status

01/09/2010

Page 8: Preshower calibration

PRS calibration 8

• Lessons from this exercise:– Distributions of the spread of the channels are

basically gaussian. – Even BXID are software calibrated at the 5% level

(68 % of the channels have the same answer within ±5%).

– Odd BXID are software calibrated at the 10% (see later for the difference).

– Systematic shift between odd and even BXID at the 10 % level at maximum.

Current status

01/09/2010

Page 9: Preshower calibration

PRS calibration 9

We want the detector calibrated for the EM triggers: the same threshold ( > 100 ADC counts) is applied at the trigger level.

Motivation for a recalibration (reminder)

01/09/2010

Page 10: Preshower calibration

PRS calibration 10

Refined analysis

01/09/2010

Page 11: Preshower calibration

PRS calibration 11

• To improve the precision of the calibration following changes to the fitting procedure were done (since the last software corrections were applied):

– De-gain distribution before fitting to remove empty bins from histograms. Apply track length correction after.

– Added exponential distribution for a pedestal background shape. Final fitting model: Landau × Gaussian + exponential.

– Split in even / odd BXID. It should account for different VFE electronics gain and different pedestals.

• Evaluated on a sample from April and checked on a sample essentially from May (r4s5 w/ April data excluded).

• Latest data are currently under analysis.

Refined analysis

Legend for the next slides:

-Current sw calibration – previous calibration (old model) which is currently inside the software and which is used for the reconstruction.

- April – the same sample as the sample used for current sw calibration , fit with the new model, split even/odd.

-r4s5 – sample mostly from May, fit with new model used as a reference for checks, split even/odd.

01/09/2010

Page 12: Preshower calibration

PRS calibration 12

The improved fit result by regions

C side

Current sw April odd April even

Mean RMS Mean RMS Mean RMS

Inner 15.4 1.5 14.7 1.6 15.0 1.6

Middle 14.8 1.5 14.6 1.6 14.6 1.6

Outer 13.7 1.5 13.2 1.4 13.5 1.4

A side

Current sw April odd April even

Mean RMS Mean RMS Mean RMS

Inner 16.9 1.9 15.6 2.0 16.2 1.7

Middle 16.0 1.7 15.5 1.6 15.7 1.6

Outer 14.4 1.7 13.9 1.6 14.2 1.7

The differences between even and odd, between both of them and the previous calibration (currently in software) are presented on the next slides.

Statistics (number of tracks used)

Sample Odd Even

April 226 263 137 307 333 743

May 701 666 121 204 810 423

01/09/2010

Page 13: Preshower calibration

PRS calibration 13

Comparison even / odd

01/09/2010

Page 14: Preshower calibration

PRS calibration 14

Comparison even / odd

•The MPV values for even and odd BXID are not the same with an average shiftand RMS given on the plots.

•Typically MPV values from even BXID sample are greater than from odd BXID sample.

01/09/2010

Page 15: Preshower calibration

PRS calibration 15

Even / odd comparison by regions

• The difference between even and odd MPV value is higher in some regions especially for:

- PRS 2,

- PRS 7 crates 11 - 13.

• Pedestal measurement related ? Under investigation.

• Excluded channels don’t contain a MIP shaped signal (an example later).

01/09/2010

Page 16: Preshower calibration

PRS calibration 16

Comparison with the current calibration

01/09/2010

Page 17: Preshower calibration

PRS calibration 17

Even / odd comparison with current software calibrationO

dd

Eve

n

01/09/2010

• Odd sample has a larger spread (related to the fraction of odd and even in the initial calibration sample.

• Shifts of the mean values towards negative values are related to the model.

Page 18: Preshower calibration

PRS calibration 18

Even / odd comparison with current software calibration

• Correlation between split determination and averaged one (as in current sw).

• 2/3 of the events were in even BX.

01/09/2010

Page 19: Preshower calibration

PRS calibration 19

Detector stability. Comparison with May sample (r4s5).

01/09/2010

Page 20: Preshower calibration

PRS calibration 20

• Odd BXID:Shift not greater than 0.1%.RMS not greater than 0.5%.

• Even BXID:Positive shift of the mean value of MPV but still no more than 1%.RMS no more than 1.5%.

• It is still below 5%.

• The exact source of the difference is yet to be found (one but possibly not the only one is the statistics used).

Comparison with May sample (r4s5): Detector stability

C side A side

Odd Even Odd Even

Mean RMS Mean RMS Mean RMS Mean RMS

Inner 0.01 0.05 0.12 0.16 0.00 0.05 0.06 0.15

Middle 0.01 0.05 0.05 0.16 0.01 0.04 0.07 0.14

Outer 0.01 0.05 0.08 0.14 0.01 0.05 0.05 0.15

01/09/2010

Page 21: Preshower calibration

PRS calibration 21

• The comparison shows the correlations between the samples from April and from May (r4s5) are grater than 99% except of even BXID A side Middle sample which is 97%.

• It allows to conclude that the detector was very satisfactorily stable in terms of gains and pedestals during that period (and timing).

Comparison with May sample (r4s5): Detector stability

01/09/2010

Page 22: Preshower calibration

PRS calibration 22

Open questions and known issues.

01/09/2010

Page 23: Preshower calibration

PRS calibration 23

• Still a few questions to answer (but within 5% calibration!)

Open questions and known issues

• Even/odd difference: time stability plots.

Pedestal plots have to be investigated.

• Terrible channels- Dead channels ( 2 ) PRS 7 FEB 7– almost no statistics

Reference neighbor

01/09/2010

Page 24: Preshower calibration

PRS calibration 24

Open questions and known issues

- 16 channels + 3: no MIP shape ( a 1/4 board faulty - board to be replaced ) – Excluded channels on 2D plots.

They do correspond to 2 ADCs. Intermittent errors. Misadjusted sampling or hw failures ?

- Uncorrectable pedestals (~20 channels).

Reference neighbor

Pedestal 3627

01/09/2010

Page 25: Preshower calibration

PRS calibration 25

Open questions

- Some boards with higher difference between even and odd.

- Correlation of spotted „bad” channels with the EF was done once. Basically all problems spotted by the EF were identified. It might be worth to make a second pass with the latest data.

- Untypical shapes (~20 channels).

01/09/2010

Page 26: Preshower calibration

PRS calibration 26

Summary and plans

01/09/2010

Page 27: Preshower calibration

PRS calibration 27

• PRS is currently calibrated at the ~5% level.• Still, this is software calibration. The detector itself is

~15%. The EM triggers are not benefitting from the 5% calibration.

• A refined determination of the MIP position has been developped. Older results are reproduced.

• Calibration is split according to the BXID alternance to account for the different gains of the VFE electronics subchannels (and pedestal differences).

• Differences up to 10 % between odd and even BXID are observed.

• Still open questions here and there.

Summary

01/09/2010

Page 28: Preshower calibration

PRS calibration 28

• Corrections by subchannels are ready for the numeric gains to be set in the FE electronics.

• HV settings to uniformize tubes are accordingly ready (provided the last checks are completed).

• We’d like to calibrate the detector at the next

possible occasion.

• Latest data from the summer still to be analyzed.

Plans

01/09/2010

Page 29: Preshower calibration

PRS calibration 29

Backup

01/09/2010

Page 30: Preshower calibration

PRS calibration 30

Even / odd comparison with current software calibration

• Current calibration is closer to even MPV values than to odd MPV values because the sample used to obtain it had a ratio of even / odd BXID events ≈ 3:2.

01/09/2010