PHOS offline status report Dmitri Peressounko ALICE offline week,
Reconstruction status Raw data reconstruction: – Fast fitting with GammaN is implemented. No much improvement neither in Amp, no by using 2 cut => do not use. – Correction for a time stamp by the L1 phase read out of ALTRO trailer. Time gate was introduced to RecoParameters and to Clusterizer to suppress noise. Not used by default, still under investigation. More flexibility on non-linearity correction added: one can choose functional form and parameters from OCDB. Default values not changed yet PHOS offline status2
Online DQM PHOS data quality monitor is implemented as QA AMORE agent. For the DQM shifter a single histogram is shown: the number of cells per event. In pp run the physics events are rare, hence the number of cells is determined by noisy channels. In Pb-Pb run the occupancy of the PHOS detector may be high, so the number of cells is determined by the sum of noise and physics signals PHOS offline status3
PHOS OCDB objects Energy calibration: conversion from ADC counts to GeV High gain/low gain ratio Bad channel map Alignment Reconstruction parameters All objects are produced offline PHOS offline status4
PHOS energy calibration Pre-calibration: adjusting the high-voltage APD bias to provide the same APD gain for all channels. Achieved calibration accuracy: 20-50%; Physics data with pp collisions allowed to improve calibration using equalization of the mean deposited energy per channel from physics events. Achieved calibration accuracy: 6.5%; Final calibration will be finalized using equalization of the 0 peak per channel. Accumulated statistics is still not enough for it. The goal is to achieve calibration accuracy of 1%. Difference between pass1 and pass2 reconstructions for PHOS: – LHC10b: pass1, pass2 – both initial calibration – LHC10c: pass1 – raw calibration, pass2 – latest calibration – LHC10d: pass1 was corrupted due to OCDB error can’t be used, pass 2 OK – LHC10f, g: latest calibration in both passes PHOS offline status
Calibration with PHOS offline status6 For each PHOS cell calculate mean energy in the range E>0.06 GeV Calculate corrections to the CC: C i = const/, where const was adjusted to put π 0 peak to 135 MeV/c 2. Use C i to calculate of the next iteration: recalibrate cell energy with C i recalculate energy and position of the cluster Calibration was found on 70M event sample
Module 4 before calibration PHOS offline status7
Module 4 after calibration PHOS offline status8
PHOS HG/LG calibration PHOS FEE measures energy by 2 ADCs with different gains: High gain: 2 MeV – 2 GeV, 10 bits Low gain: 32 MeV – 32 GeV, 10 bits Ratio HG/LG may vary from channel to channel. Precise value HG/LG is needed to ensure a continues spectrum in the whole range 2 MeV – 32 GeV HG/LG was measured in dedicated LED runs with variable amplitudes. HG/LG is a property of the electric circuit and cannot change in time. One high-statistics LED run is enough to calculate it once and forever PHOS offline status
HG/LG measurements PHOS offline status10 HG/LG ratio varies within RMS 2%. Mean value is different in module 4 compared with modules 2 and 3.
Bad channel map PHOS offline status11 Bad channel map was found off-line from different sources of information: Pedestal runs LED runs with variable amplitude LED runs with zero amplitude Physics runs In total 1371 channels are bad (out of 10752) All three Pedestal+Quality Quality+Noise Quality Pedestal Noise
Track Matching, LHC10c pass2 dz vs z No slope in dz(z) distribution. Still there is some jump between + and – z: 0.33 cm for mod.3 and 0.2 cm for mod 4. (residual TPC decalibration + TPC sector dependence?) PHOS offline status
Track Matching, LHC10c pass2, phi No unique slope for all distributions. Overall offsets: Module 2: -0.7 cm; Module 3: -0.8 cm; Module 4: -0.9 cm PHOS offline status
Timing: Pileup PHOS performance in pp14 Runs analyzed (LHC10e): ,130799,130802, ,130834,130842, ,130848, Wrong bands disappear if PhysicsSelection +Zvtx cut applyed Pileup:
Timing: L1phase shift PHOS performance in pp15 Subtracting: EMCAL/TPC Adding: PHOS
Timing: L1phase shift - Side peaks Single peak at 0 Single peak at -1 Single peak at +1 2 peaks at -1,0 2 peaks at 0,+1 2 peaks at -1,+1 Correct operation: “+”; Side peaks: FEE mis-configuration
Shower shape PHOS offline status17 LHC10e: photons P t >1 GeV/c, contributing to 0 peak 0 = 2.0±0.01 0 = 0.71± = 1.22±0.01 1 = 0.42±0.01 c = -0.59±0.03 MC: simple photons Position of lambda bump is ~10% higher than in MC, width is same. => Result of decalibration. 1 0
Purity and efficiency PHOS offline status18 EfficiencyS/Bg No Cuts100%0.06 Neutral96%0.10 Dispersion75%0.23 Disp+Neutral72%0.32 Runs 130xxx
Good run selection Runs suitable for physics analysis were selected by studying the output histograms of the analysis train (wagon CaloQA): Cell multiplicity Cell energy spectrum Cluster multiplicity Cluster mean energy Slope of the cluster energy spectrum Number of reconstructed 0 per event PHOS offline status19
QA: Cluster multiplicity and Mean energy PHOS offline status20 E cluster >0.3 GeVE cluster >0.5 GeV
QA Energy slope and Number 0 of per event PHOS offline status21
Invariant mass: 0 and in pp PHOS offline status LHC10e pass1: 215M events
Tuning MC to reproduce real data PHOS offline status23 Decalibration is imitated by the Gaussian smearing of calibration parameters with =6.5%. Non-linearity correction was introduced to MC. All corrections to MC are applied in analysis. MC in mass production run always with ideal (residual) OCDB.
peak position and width in pp Measured mass of 0 and are consistent with PDG values a proof of good energy linearity PHOS offline status
Preparation for HI Keep same thresholds and other parameters as for pp =>Use same calibration Possible caveats: too high occupancy and too large clusters Two more branches were switched off in last 2 months =>revisit BadMap
Invariant mass: 0 in Pb-Pb PHOS offline status26
Simulation status Simulation of anchor runs is performed with the PHOS OCDB objects from raw:// - bad channel map and reconstruction parameters. Calibration is taken from residual OCDB, because calibration improves with time, and simulation will not re-run. Applying real calibration parameters is passed to analysis PHOS offline status27
Summary PHOS calibration is an offline task which requires the full available statistics. Conditions: calibration parameters, bad channel map, HG/LG ration are stable. Once they are found, then can be used in pass1 reconstruction for future LHC periods. Further improvement of PHOS calibration requires 10 9 pp events. Until this statistics is accumulated, pass1 and pass2 reconstructions have equally good quality. Dispersion: Still 10% difference between data and MC – investigating Track matching: Accuracy is sufficient for PID cuts. In direction matching can be improved by modifying “distance” to PHOS for hadrons PHOS offline status28
Backup PHOS offline status29
Track Matching, LHC10e pass1 dz vs z Big jumps between + and – z. Within constant sign regions slope consistent with zero. Average offsets: Module 2: -1.3 cm; Module 3: cm; Module 4: -0.7 cm. Consistent with LHC10c pass PHOS offline status
Track Matching, LHC10e pass1, phi PHOS offline status
MC LHC10d4: Hadrons, dz(z) Slope ~0.2% =>dR~0.9 cm. Should be attributed to the deeper hadron shower than photon PHOS offline status
MC LHC10d4: hadrons, phi PHOS offline status