1. 1 EMCAL Reconstruction in Pass6 2009 - pp 900 GeV 29/03/2010 Gustavo Conesa Balbastre

2. 2 Intro Changes in the raw fitting with respect to pass5 (David S. has the details): –Raw Sample selection –Time measurement –Hardcoded default parameters –… Fitter is kStandard (TMinuit) –Other fast fitters (3) under test but we are still trying to decide which one will replace the kStandard. –Decision to be made doing tests with data … but 2009 data has a problem … We discovered last week that our raw data is corrupted … –possible explanation of most of the pi0 peak shift. –Fixed (hopefully) for next data taking. –More details in last slides.

3. 3 Energy distributions Low cell energy rejected in pass 5, clear effect in all the cluster Energy. In pass6 new 3 noisy channels in all the events, with low energy, 0.1-0.2 GeV

4. 4 Energy distributions pass5/pass6 Low cell energy rejected in pass 5, clear effect in all the cluster Energy. In pass6 new 3 noisy channels in all the events, with low energy, 0.1-0.2 GeV

5. 5 Number of cells per cluster: Pass5/6 Pass5 Pass6 Increase the number of cells per cluster from pass5 to pass6

6. 6 Number of cells per cluster: Simulation (LHC09d10) vs Pass5 Pass5 number of cells per cluster much smaller than in simulation

7. 7 Number of cells per cluster: Simulation (LHC09d10) vs Pass6 Better agreement between the simulation and data, although simulation has not the same calibration parameters so we are maybe not comparing really apples to apples.

8. 8 Pi0 Invariant Mass Distribution pass6pass5 Peak shifted, in the good direction, about 10 MeV upper in pass6. More pairs in new pass due to the higher number of small signal channels. Background also higher due to more noisy (low signal) channels. different event statistic in both passes, less analyzed in pass6

9. 9 Pi0 Mass pass 5 pass6 There is still a shift of 30 MeV (30% decalibration) to the expected position (M  = 135 MeV/c 2 ) No improvement seen on the width.

10. 10 Cell Time measurement Time of clusters concentrated in 500-600 ns (peaks at 500 and 600 ns for pass6). More simmetric distributions in pass6. Need to access to the T0 of the event to calculate the Time of Flight. The RCU clocks are running at 10 MHz and we are not sure that they are all on the same phase with respect to the beam clock, or if the phase is stable over time. This might effect, need to do more check comparing different RCU regions

11. 11 Track Matching -Problems with TPC? No changes in TPC for this pass so it shouldn’t. -Change in our code? In principle nothing changed in the tracking code or the geometry between the passes. Somehow the larger number of low energy noisy clusters destroys matching?

12. 12 Cluster-Track in pass6 Track propagation done with AliEMCALTracker, same used for matching. Looks reasonable

13. 13 Cluster - Track in pass 6 Residuals Track propagation done with AliEMCALTracker, same used for matching Looks reasonable except “y” which is shifted 3 cm. This does not happen with cosmic data.

14. 14 Cluster-Track Pass2 Cosmic Residuals

15. 15 EMCAL Calibration EMCal “Calibration Mystery” solved… –Due to “Variable Pedestal” subtraction method used for 900 GeV p+p data Variable pedestal calculated automatically and subtracted from signal Tested and confirmed to give same result (max-min) as for no pedestal subtraction, for EMCal with LED signals –But, for LEDs triggered by TRU on receipt of L0, »I.e. LED pulse arrives AFTER L0 in post-samples –Problem first reported by PHOS about 3 weeks ago… Apparently (obviously), the pedestal calculation stops on receipt of L0 signal and USES the presample region for pedestal calculation. –Signal cannot fall into presample region!!! –But L0 arrives 1.2us after interaction… –It’s in the RCU manual… –EMCal signal entirely in presample region… Effect on EMCal signals? L0 PHOS Time bin

16. 16 EMCAL Calibration Observe that signal is ~0.7 smaller with variable pedestal subtraction Does not seem to depend on amplitude, but does depend on RCU (time bin?) Further studies needed with varying LED delays to investigate dependence. Hopefully, it will be a single overall scale correction.

17. 17 Back up

18. 18 Energy spectra

19. 19 Energy Ratio kX /pass6 What happens at 0.5 GeV with kCrude? Pass6 like kCrude, similar trend, a bit higher except at 0.5. Pass5 similar to kCrude for high energy cells, we throw away low energy cells, < 0.5 GeV. At cluster level smaller number because of less low energy cells? kPeakFinder 90% of kCrude, same trend. kLMS almost = kFastFit lower increasing energy. kNeuralNet very different.

20. 20 Cell Time Time : All new fitters have some discrete values, kPeakFinder almost only discrete, kCrude only discrete.

21. 21 Pi0 Invariant Mass Distribution kPeakFinderpass6kNeuralNet kFastFit kLMS kCrude

22. 22 Pi0 Mass pass6kPeakFinderkNeuralNet kFastFitkLMS kCrude

23. 23 Pi0 Width Pass5Pass6 No improvement observed on the width

24. 24 Pi0 Width Pass6kPeakFinderkNeuralNet kFastFitkLMS kCrude