1. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏1/16 Final Pixel CRAFT results Pixel General Meeting Romain Rougny, University of Antwerpen

2. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏2/16 All CRAFT Analysis Calibrations at VME level (K. Ecklund, A. Ryd) Threshold and Noise measurement (D. Kotlinski) Gain Calibration results (F. Bleckman, B. Heyburn, R. Rougny) List of Active Components (K. Rose, L. Wehrli, D. Kotlinski) Noisy Pixels from DQM (P. Merkel) Comparison between Data and MC (T. Rommerskirchen, G. Giurgiu, V. Chiochia) Lorentz Angle Measurement (L. Wilke, A. Kumar, M. Swartz) Hit and Impact Parameter Residual (A. Bonato, N. Trahn, A. York) Pixel Hit Efficiency (L. Mucibello, R. Rougny, N. van Remortel) Track Detection Efficiency (M. Lebourgeois, B. Mangano) Pixel Hit Resolution (K. Ulmer) All analysis are detailed on the following twiki : https://twiki.cern.ch/twiki/bin/viewauth/CMS/TkCraftNotePixel

3. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏3/16 CRAFT Statistics For pT > 20 GeV, all 3 algorithms identify the same number of tracks (37967) All analyses are done with the CTF algorithm unless specified otherwise

4. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏4/16 Address Level Calibration For each pixel hit, the address is encoded with a 6 level scheme over 5 clock cycles, the 6 th one giving the charge. Example of the 6 level peaks for one ROC Separation between the mean of 2 adjacent peaks in units of the peak RMS RMS of all peaks for all ROCs The levels are well-defined and well-separated

5. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏5/16 Threshold and Noise Noise Treshold VCal efficiency The threshold is the VCal value for which the efficiency reaches 50% The noise is the width of the region where the efficiency goes from 0% to 100% The conversion Vcal-electron varies between pixels. On average, we have 1 VCAL = 65.5 electrons Threshold and noise are measured during a S-curve run, where only 81 cells per ROC are measured due to time constraints

6. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏6/16 Thresholds and Noise BARREL END-CAP Mean = 3690 e - RMS = 414 e - Mean = 2870 e - RMS = 194 e - Mean = 141 e - RMS = 35 e - Mean = 85 e - RMS = 27 e - MIP = 22000e -, so noise is small once considering a charged particle

7. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏7/16 Gain Calibration The gain and pedestals are used to calculate the charge deposited in clusters. Like for the S-curve, the gain calibration is done with a dedicated run For each pixel, we plot the response in ADC of an input charge in VCAL. We then fit the part not on the plateau by a straight line Gain = 1 / slope Pedestal = offset [ADC]

8. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏8/16 Gain Calibration END-CAPBARREL Gains are stored per column Pedestals are stored per pixel Those values were included in the DB since CRAFT_ALL_V9, so were used for the 2 nd and 3 rd reprocessing

9. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏9/16 Noisy Pixels Noisy pixels are detected via the Pixel Data Quality Monitoring (DQM) package – Can be done in real time or offline (running on reprocessed data) – Counts the number of events in which a pixel registers a charge above threshold, and divides by the total number of events – the “digi event rate” Cutoff: digi event rate > 0.001 – Barrel: 235 noisy pixels – Forward: 17 noisy pixels Noisy pixels were masked during CRAFT data taking If cutoff is tightened to digi event rate > 0.0001, only 13 additional pixels would be declared noisy (these are not currently masked) Number of noisy pixels is very small:.00056% of total pixels

10. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏10/16 Data/MC Comparison Cluster charge is corrected for the way the particle travels in a module. pre-selection : – Clusters must be on tracks – Tracks must have > 4 GeV momentum – For each cluster, the muon time must be >0 OR its charge must be > 10k e - MC has IDEAL gain calibration New MC has : higher thresholds, new lorentz angles, new value for vcal Many cosmic ray effects (time-walk, broken clusters) still not simulated in MC (and not needed for collisions). Necessary for good comparison with MC

11. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏11/16 Lorentz Angle Charge carriers are affected by the Lorentz force (deflected at Lorentz Angle -LA). To determine the value of the LA, the spread of the drifting charge distribution is measured as a function of the track incidence angle (minimized at LA) Comparison to PIXELAV simulation shows good agreement: – BPIX 3.8 T, 100V, 20 C: cot( ) = -0.452 +/- 0.002 – FPIX 3.8 T, 300V, 20 C: cot( ) = -0.080 +/- 0.005 BARREL END-CAP LA = 24.7° @ 3.8 T LA = 4.0°

12. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏12/16 Hit Residuals Cuts : momentum > 10 Gev Edge Hits removed 1 pixel hits with charge < 10k e - This is the TOTAL residual : It includes the track extrapolation error, the detector intrinsic resolution, the multiple scattering contribution, and uncertainties due to residual detector misalignement

13. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏13/16 Barrel Pixel Track Efficiency Definition of efficiency Pixel integrates over 1 BX, whereas the tracker does over 3 BX Due to geometry (and nature of cosmic rays), only 3% of tracks through BPIX, 1.5% through FPIX Cuts: Loose Timing Cut (above 70% eff vs muon timing) If cut of pT > 20 GeV applied, RS similar to CTF/COS Comparison of CTF to MC gives similar results This is not a measurement of the sensor efficiency, but is rather a result of several factors: Tracking efficiency, alignment, detector performance, timing synchronization, etc.

14. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏14/16 Pixel Hit Efficiency Layer 1Layer 3Layer 2 Cuts : - keep only events with exactly 1 track, with all FEDs in - keep only tracks with pT > 10GeV - rejected hits on the borders of the modules - for each hits, must be at least 1 other valid hit on the track with Y 0 - muon arrival time must be in a 10ns window MAIN SOURCE OF SYSTEMATIC Once cleaned (no modules with problems or nHits < 10) the mean module efficiency is : Layer 1 : 97.12% | Layer 2 : 97.06% | Layer 3 : 96.39%

15. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏15/16 Pixel Hit Resolution Cuts : Only tracks passing through barrel and overlaping modules Track momentum > 5 GeV Only tracks with angle < 30° from normal Exclusion of edge clusters Exclusion of clusters with more than 2 pixels wrt expected Final Fit values for 1 module : X : 16 + 2 micron Y : 29 + 3 micron

16. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏16/16 Impact Hit Parameters Based on track-splitting method Cuts: – At least 3 Pixel Barrel hits per leg – Momentum >= 10 GeV to reduce MS effect APEs affect parameter resolution : inclusion of pixels improves impact parameter resolution – Ongoing studies to find optimal APEs

17. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏17/16 Conclusion A lot of analyses were performed on the CRAFT data All analyses are currently finishing, minor updates needed, like rerunning on the latest datasets They will all be included in the pixel paper that is currently being written by V. Chiochia and K. Ecklund

18. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏18/16 BACKUP

19. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏19/16 Pixel hit resolution Use tracks passing through two sensors in the same layer – overlaps Combine forward and backward predicted states to get best track prediction without information from the layer under study Compare the difference in hit and predicted positions on the two overlap modules: ΔXhit=Xhit1-Xhit2, ΔXpred=Xpred1-Xpred2

20. UA - Antwerp - 21/04/09 Romain Rougny (CMS)‏20/16 Lorentz Angle