laser pulse width - california institute of technologyveverka/files/report_07-04-04.pdf · — adi...

Laser Pulse WidthLaser Pulse Width

Progress Report

J. VeverkaCaltech CMS Group Meeting

April 4, 2007

2

APD/PNAPD/PN--WidthWidth Linear FitLinear Fitss

• Linear fit of the APD/PN-width dependence for each channel of each SM

• Normalize APD/PN by the fit value at width = 30 ns

• Distributions and crystal maps for the slope, intercept, chi2, etc. of the linear fits for the normalizedAPD/PN values

error bars blown up by a factor of 10

normalization value

Example

3

Slope ValuesSlope Values

• Fit results for all channels of SM 4, 13, 17, 19, 20 and 22 (6 x 1700 = 10200 channels).

• The top plot shows the distribution of the linear-fit slopes.

• The bottom plot shows what fraction of crystals has a slope that fits in a given range around the mean value.

• Note that some 75% of the 10200 channels have a slope in the range (-2.372 +- 0.5) ‰/ns

4

Slope ValuesSlope Values

• Fit results for all channels of SM 4, 13, 17, 19, 20 and 22 (6 x 1700 = 10200 channels).

• The top plot shows the distribution of the linear-fit slopes.

• The bottom plot shows what fraction of crystals has a slope that fits in a given range around the mean value.

• Note that some 75% of the 10200 channels have a slope in the range (-2.372 +- 0.5) ‰/ns

75%

5

LM / TT FeaturesLM / TT FeaturesSM04

SM17

6

LM / TT FeaturesLM / TT FeaturesSM04

SM17

Crystal 353 Crystal 935

Crystal 558 Crystal 758

7

LM/TT FeaturesLM/TT Features

• Sometimes bad data, e.g. top plot

• Sometimes inherent property of the data, e.g. bottom plot

Crystal 558

Crystal 758

8

DQMDQM

APD/PN

9

APD/PN StabilityAPD/PN Stability

Plot sigmaof the fit

APD/PN

10

SM17SM17

11

SM17SM17

12

Run 20755Run 20755

13

Run 20755Run 20755

Disturbing tail

14

Run 20755Run 20755

Crystal 81

15

Run 20755Run 20755

Crystal 82

16

Run 2075Run 207566

Crystals 81, 82

17

APD/PN, SM17APD/PN, SM17Crystal 81 Crystal 82

Run

20755

Run

20756

18

PlanPlan

• Plot APD/PN Gaussian fit chi2’s • Plot crystal misfits run by run • Plot crystal pulls run by run• Devise DQM cuts

Backup SlidesBackup Slides

ReviewReview

Laser Monitoring Issues

21

Laser Monitoring Laser Monitoring

• Purpose: measure the ECAL crystal transparency change due to irradiation during the LHC running

• Goal: ~1 ‰ APD/PN stability• Need: understand systematic correlation between

APD/PN laser pulse– Width– Timing– Amplitude

• Here: concentrate on the width issues

22

APD/PNAPD/PN--Width DependenceWidth Dependence

• Simulated as a convolution of the laser pulse shape and electronics response:“The dependence observed in data can be reproduced based on the properties of the pulses alone.”

— Adi Bornheim, TB meeting, 20 Sep 2005“Slope (normalized APD/PN vs. width) : 2 ‰/ns”

— Adi Bornheim, TB meeting, 3 Nov 2005• Measured for a few channels of the 2004 SM10 data to be

linear with a slope of around 2.5 ‰/ns. For details, see talk by Adi Bornheim, TB meeting, 3 Nov 2005.

• Expected long-term width stability ~1-2 ns• Implication: The effect is larger than required precision, a

correction is needed. • Here: Measure the effect for the 2006 TB data on a larger

scale

Review: WidthReview: Width

Review of APD/PN and Laser Pulse Width Correlation

24

Pulse Shape ConvolutionPulse Shape Convolution

Time [ns]20 40 60 80 100 120 140

Pu

lse H

eig

ht

0

0.2

0.4

0.6

0.8

1

APD Pulses - Simulated

Laser Pulse Width 20 ns

Laser Pulse Width 25 ns

Laser Pulse Width 30 ns

Laser Pulse Width 35 ns

Laser Pulse Width 40 ns

Time [ns]400 500 600 700 800 900 1000 1100 1200

PN

Pu

lse H

eig

ht

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

Time [ns]780 785 790 795 800 805 810 815 820

PN

Pu

lse

He

igh

t

0.4079

0.40795

0.408

0.40805

0.4081

Time [ns]20 40 60 80 100 120 140 160 180

AP

D P

uls

e H

eig

ht

0

0.2

0.4

0.6

0.8

1

Laser Shape

Convoluted Shapes

Reminder :Pulse shape is a convolution of the electronic shape and the ‘line shape’ of the light. In case of a laser pulse, essentially a gaussianwith FWHM of 20 – 40 ns.Details : See talk on 20 Sep. 2005.

Remaining issue :The pulse width dependency extract from simulated shapes depends strongly on the a priori unknown electronic shape. This makes it difficult to predict the actual pulse width dependency.Solution :Tune the convoluted shape such that it matches the shape in data.

25

Pulse Width Correction on SM10Pulse Width Correction on SM10 in in 20042004

Uncorrected

Corrected

ECAL APD/PN, Single Channel Monitoring History

Laser Pulse Height

Laser Pulse Width

450 Hours

Fast Monitor Data

Data analysed : Part of Period 1 (not all the data was re-reprocessed to fix PN data) and Period 3. Period 2 is problematic -and thus not used.

Pulse width correction :APD/PN_cor = APD/PN+c·PW_Laser

26

Monitoring Stability Monitoring Stability vsvs Pulse Width Pulse Width CorrectionCorrection

Mon

itorin

g St

abili

ty [‰

]

Slope [1/ns]

With a linear correction we can vary the slope to study the sensitivity :

From SM10 data it appears that we don’t have to know the slope with great precision.

Review: AmplitudeReview: Amplitude

APD/PN and Laser Pulse Amplitude Correlation

28

APD [ADC Counts] APD [ADC Counts] vsvs APD/PN for SM22APD/PN for SM22

• For the SM22 PW scan, the intensity changes between 2000 and 4000 ADC counts. For that the APD/PN changes ~3.0 %

• From the linearity scan with the laser above we see that the nonlinearity as a function of the pulse intensity is of the order of ~0.1 %.

PrePreview: Timingview: Timing

APD/PN and Laser Pulse Timing Correlation

30

Amplitude, Timing and Width vs Pump Current

• At 25 A : -1.7 ns/A width / -46 ns/A timing

• At 22 A : ~ -4 ns/A width / -62.5 ns/A timing

• The timing changes ’10 times as fast’ as the width – in units of ns.

• The ageing should follow the same curves with respect to each other, otherwise the feedback does not work.

31

Timing and Width Timing and Width vsvs TimeTime

Shown is the drift of the pulse width (left) and the pulse timing (right) for the period corresponding to SM22 on the TB. While the timing drifts by about 2 ns between 9900 h and 10000 h the width drifts about 1.5 ns. That is clearly in contradiction to the behaviour on the previous slide.

New ResultsNew Results

APD/PN and Laser Pulse Width Correlation Measurements

33

Used DataUsed Data

• APD/PN data– Pulse width scans for 7 SMs: 2, 4, 13, 17, 19, 20, 22

(1700 channels each)– Total of ~90 useful laser runs (600 events each)– Standard online laser code used for reconstruction– Gaussian fit for each channel of each run:

• APD/PN value = mean of the fit• APD/PN value error = (sigma of the fit) / √600

• Laser pulse width data– Fast Monitor in the laser barracks used – All 2006 laser runs reconstructed and matched – Gaussian fit for each run:

• Width value and its error = same as for the APD/PN

34

Slope Distribution Example Slope Distribution Example -- SM17SM17

Max. single-value correction error ~ (0.18 ‰/ns)×(2 ns) = 0.36 ‰ < 1 ‰

For more plots like this one, see http://ultralight.caltech.edu/hepwiki/PulseWidthSystematics

Sigma / |Mean| = 6.9(1)%

35

Slope Crystal Map Example Slope Crystal Map Example -- SM17SM17

• Noticeable LM structures– Their scale is small compared to the slope values– They are a general feature of the APD/PN-width dependence for more or

less all studied SMs– Interesting but not yet thoroughly investigated

• For more plots like this one, see http://ultralight.caltech.edu/hepwiki/PulseWidthSystematics

36

SM04SM04

37

SM04SM04

38

SM13SM13

39

SM13SM13

40

SM19SM19

41

SM19SM19

42

SM20SM20

43

SM20SM20

44

SM22SM22

45

SM22SM22

46

SM02 SM02 –– Low IntensityLow Intensity

47

SM02 SM02 –– Low IntensityLow Intensity

48

SM17 SM17 –– Low GainLow Gain

49

SM17 SM17 –– Low GainLow Gain

50

ResultsResults

SM # Runs Run Numbers* Stand Slope (err) [‰/ns]04 15 25067-81 H4 cosmic -2.01(16)13 14 19811-24 H4 cosmic -2.91(12)17 5 20753-57 H4 cosmic -2.58(18)19 15 21683-99 H4 cosmic -2.28(11)20 9 23254-63 H4 cosmic -2.39(13)22 13 13582-96 H4 test beam -2.04(41)

• Legend:– Slope = mean of a Gaussian fit to a distribution of 1700 values as in slide #9– Err = sigma of the Gaussian fit

• Note that the slope values are compatible across the different SMs• Two SMs measured incidentally with improper intensity settings – results not listed

here since the slope values are not usable.• Assuming single-value correction for all SMs, it’s maximum error would be roughly

(0.5 ‰/ns) × (2 ns) = 1 ‰ → might/might not be good enough, evaluation needed

*Some run numbers in the range might be excluded

51

SummarySummary

• Correlation of APD/PN and laser pulse width measured for all channels of 7 super modules – linear dependence observed

• Results are consistent with expectations based on laser-pulse-shape and electronics convolution simulation

• Some LM systematic structures observed

52

OutlookOutlook

• Fit quality monitoring– plot APD/PN stability distributions and crystal

maps – plot fits for single crystals (groups by TT's) – make misfit crystal maps run by run – make pull crystal maps run by run

• Apply width-based correction to TB data• Significant APD/PN stability improvement

expected• Stay tuned for new results

53

Chi2 and Log10(Chi2) for SM17Chi2 and Log10(Chi2) for SM17

54

Chi2 and Log10(Chi2) for SM17Chi2 and Log10(Chi2) for SM17