EMCal Commissioning November 8 2010 Outline EMCal Operation
EMCal Commissioning November 8, 2010 Outline: • EMCal Operation • EMCal Performance • EMCal Beam Test • EMCal Calibrations • Outstanding issues (some of them!) November 8, 2010 EMCal Commissioning, T. Awes
EMCal Commissioning • Executive summary: – EMCal has operated extremely well during 2010 run period • EMCal DCS control and monitoring fully integrated in ALICE • Very few bad channels or problems (mostly RCU/BUSY related) – Minimal “EMCal On Call” attention required • EMCal L 0 trigger commissioned and successfully operated – Very successful Beam Test • Extensive set of measurements to fully understand EMCal (as built) response – EMCal Calibration • • November 8, 2010 Geometry issues still not fully resolved Apparent Non-Linearity as seen from mass not fully understood Cluster unfolding urgently needs further development Relative tower calibrations are quite good EMCal Commissioning, T. Awes 2
EMCal Monitoring & Control • EMCal fully integrated into ALICE DCS control – Essentially all work done by Jiro (Creighton) with early help from Kithsiri • STU DCS control done by Galdric (Grenoble) – EMCal respects various DCS control commands • Ready, Standby, Safe, Super. Safe – DAQ shifter can configure EMCal to take special pedestal runs between spills • Top Level EMCal PVSS Overview Panel (~April, not final) • EMCal also takes LED calibration triggers at ~0. 5 Hz during Physics datataking EMCal subsystem shifts ended May 5 (as for most of ALICE) – EMCal shift guide: http: //aliceinfo. cern. ch/Collaboration/Run_Coordinati on/Run 09/shift/ • Not updated since May! – Only EMCal “On Call” experts since then • • November 8, 2010 EMCal Commissioning, T. Awes ~12 qualified experts 1/2 shift credit per 24 hrs on call 3
EMCal Monitoring & Control Main EMCal DCS Panel to monitor the status of all Services of an EMCal group (a quadrant) DCS Monitoring Panel for SM group A 0, A 1, (A 2) November 8, 2010 EMCal Commissioning, T. Awes 4
EMCal DCS FEE configuration SMA 0 -RCU 0 SMA 0 -RCU 1 SMC 0 -RCU 0 SMC 0 -RCU 1 SMA 1 -RCU 0 SMA 1 -RCU 1 SMC 1 -RCU 0 SMC 1 -RCU 1 • Existing DCS control of FEE (&TRU) via scripts executed via ssh from PVSS – 2 DCS boards per SM so 8 DCS windows! + Busy. Box • Direct communication has been convenient for commissioning • During normal ALICE operation this is done “under the hood” of PVSS • EMCal functionally fully compliant, but technically not compliant as PVSS-DCS communication should be done via DIM server – STU DCS control by Galdric November 8, 2010 EMCal Commissioning, T. Awes 5
EMCal Monitoring - AMORE • EMCal Online monitoring is done in the context of AMORE Access the different plots by clicking on the top navigation tabs – Mostly work of Francesco (Houston), together with David and Yaxian – Full EMCal views of values and rms on High and Low gain, and Time bin for • Pedestals • LED reference – Same for EMCal Trigger FALTRO data November 8, 2010 EMCal Commissioning, T. Awes 6
EMCal Expert - AMORE Access the different plots by clicking on the top navigation tabs • Early AMORE plots showing examples of problems – Missing RCU, e. g. removed from readout due to BUSY problems – Noisy channels on T-card (due to bad APD+preamp; later fixed by disconnecting preamp) November 8, 2010 EMCal Commissioning, T. Awes 7
EMCal: Data Quality Monitoring Access the different plots by clicking on the top navigation tabs • EMCal Data Quality Monitoring (DQM) plots for DQM shifter are based on ratio of current LED peaks / LED peaks of Reference Run – Top plots for all towers in SM • Two weak LEDs in SM A 1 – Bottom plots for LED Monitoring Photo. Diodes November 8, 2010 EMCal Commissioning, T. Awes 8
EMCal Performance: First Look 70 o A-side C-side 90 o • Snapshot view from TED splashes (beam dumped upstream of ALICE) prior to start of p+p collisions – General uniformity (smooth variation) – Few noisy towers (1152 towers per SM) November 8, 2010 EMCal Commissioning, T. Awes 9
EMCal Performance • EU#2 EU#1 US#2 US#1 Dead/Noisy analysis (Francesco - February) – Noisy (32): Strips with T-sensors in SMs #1, Fixed in SMs#2 – Dead (9): Mostly US#1 with “cable issues” November 8, 2010 EMCal Commissioning, T. Awes 10
EMCal Performance • EU#2 EU#1 US#2 US#1 Dead/Noisy analysis (Francesco - March) – Noisy (10): Mostly in T-card due to single APD+preamp - disconnected later – Dead (6): Mostly US#1 with “cable issues” November 8, 2010 EMCal Commissioning, T. Awes 11
EMCal Performance • EU#2 EU#1 US#2 US#1 Dead/Noisy analysis (Francesco - April) – Noisy (9): Mostly in T-card due to single APD+preamp - disconnected later – Warm (3): US#1 with “cable issues” – Dead (1): US#1 with “cable issues” November 8, 2010 EMCal Commissioning, T. Awes 12
EMCal Performance: Online • First analysis of 7 Te. V p+p data – Obtained with relative calibrations==1 • I. e using online calibrations. – Good resolution on peak - “online” calibrations are quite good. November 8, 2010 EMCal Commissioning, T. Awes 13
EMCal Calibration: 0 Resolution Data: (Alexander) Status as of July • With initial online calibrations, mass resolution is not as good as expected according to simulation, even after recalibration with MIPs in ALICE. – Indicates a ~10% calibration error. We should understand why Cosmics Calib not so good (Beam Test data). • Indicates need for tower-by-tower 0 calibration analysis November 8, 2010 EMCal Commissioning, T. Awes 14
EMCal Calibration: 0 Resolution • Use ALICE Min Bias datas set to calibrate towers with mass peak. – Indicates a ~4% calibration error remains. – Dependence on run period (Temperature in ALICE) Status as of October Pass 0 Pass 3 Analysis done in Grenoble (Aug. 12): Gustavo, Julien, … 6 November 8, 2010 EMCal Commissioning, T. Awes 15
EMCal L 0 Trigger Performance ALICE High Luminosity runs with EMCal L 0 trigger enabled (but EMCal+PHOS only) Plot produced by Marcelo Figueredo: • EMCal L 0 trigger working very well - (too much detail to report) – Big effort by Jiri (Jyvaskyla) • From analysis of 5% of run 134908 (LHC 2010 f). Duration 10 hrs. – EMCal L 0 trigger at @300 Hz, ~90 k. Hz Interaction rate. – ~10 x 106 EMCal L 0 trigger events • Corresponds to ~3 x 109 sampled Min Bias events • I. e. ~4 x accumulated ALICE Min Bias sample • Perfect data sample for EMCal calibration with ’s November 8, 2010 EMCal Commissioning, T. Awes 16
EMCal Results Threshold Counts/200 Me. V Plot produced by Alexei Pavlinov Cluster Energy • From analysis of all of run 134908 (LHC 2010 f). – ~10 x 106 EMCal L 0 trigger events • Corresponds to ~3 x 109 sampled Min Bias events, – EMCal cluster energy spectrum to ~100 Ge. V • ~30 x 106 EMCal L 0 trigger events accumulated in PHYSICS_3 with EMCal+PHOS (rest in LHC 2010 g). – Also run with ~10 x 106 events with TPC, etc with EMCal L 0 and CINT 1 triggers November 8, 2010 EMCal Commissioning, T. Awes 17
EMCal Calibration • Immediate application of this data for EMCal calibration • EMCal was pre-calibrated with Cosmic MIPs to ~10% relative calibration • Final calibrations to be made using 0 mass peak for each tower – Run 134908 alone provides 4 x more data for this than all of ALICE Min Bias data set! – 10 x 106 events (E>1. 5 Ge. V)/4608 towers -> ~2000 /tower Plots produced by Marcelo Figueredo November 8, 2010 EMCal Commissioning, T. Awes 18
Single Tower Inv. Mass Spectra Use mass peak position to adjust tower gains E > 0. 5 Ge. V < 0. 5 Edge tower November 8, 2010 Plots from Gustavo Conesa Balbastra Fiducial tower EMCal Commissioning, T. Awes Edge tower 19/11 19
EMCal Calibration # entries per tower (4608 EMCal towers) Number of cluster pairs (I. e. invariant mass spectrum entries) in mass window per tower, for tower cluster energy > 0. 5 Ge. V and asymmetry < 0. 5 20/11
Single Tower Inv. Mass Spectra E > 1 Ge. V < 0. 5 Edge tower Plots from Gustavo Conesa Balbastra Fiducial tower Edge tower November 8, 2010 EMCal Commissioning, T. Awes 21/11 21
Full EMCal Inv. Mass (Run 134809) Analysis: E cluster > 0. 5 Ge. V; asymmetry < 0. 5; no edge clusters; Plots from Gustavo No cluster splitting ( loss above ~6 Ge. V/c) 0. 5 -1 Ge. V 1 -1. 5 Ge. V 3 -3. 5 Ge. V 3. 5 -4 Ge. V 5. 5 -6 Ge. V 6 -6. 5 Ge. V 8 -8. 5 Ge. V November 8, 2010 8. 5 -9 Ge. V 1. 5 -2 Ge. V 4 -4. 5 Ge. V 6. 5 -7 Ge. V 9 -9. 5 Ge. V EMCal Commissioning, T. Awes 2 -2. 5 Ge. V 4. 5 -5 Ge. V 7 -7. 5 Ge. V 9. 5 -10 Ge. V 2. 5 -3 Ge. V 5 -5. 5 Ge. V 7. 5 -8 Ge. V >10 Ge. V 22
EMCal Calibration: 0 Resolution Status as of October Current Status • Using high statistics Run 134908 to calibrate towers with mass peak (from Nicolas@ Grenoble). – Modest improvement in resolution – Suggests a ~3% calibration error remains, if simulation results are “truth” • Calibration error? Intrinsic energy resolution? Position calculation error? November 8, 2010 EMCal Commissioning, T. Awes 23
EMCal Calibration: E Resolution • Latest update (last week) of EMCal geometry including 1 cm Al plate and reflective paper decreases sampling fraction from 1/10. 5 to 1/12. 8 (from Alexei). – Worsens energy resolution by about 10% • Does Simulation now agree with Beam Test energy resolution? – Probably gives agreement between simulation and measured mass resolution • If true, then tower calibrations no longer an issue – Important to know the answer! November 8, 2010 EMCal Commissioning, T. Awes 24
EMCal/DCal Beam Test November 8, 2010 EMCal Commissioning, T. Awes 25
EMCal/DCal Beam Test • PS beam period in T 10 beamline August 2 -16 – 0. 5 Ge. V/c - 6 Ge. V/c • SPS beam period in T 2 -H 4 beam for 7 days August 23 -30 – 6 Ge. V/c - 250 Ge. V/c • Conditions: – Gains set to values determined from cosmics calibrations (Grenoble) as for SMs at P 2 – Temperature monitored, LED gain monitoring • Goal: �T o obtain full characterization of electon and hadron response of EMCal/Dcal as built and operated – – – – Energy resolution and non-linearity Position resolution Timing resolution Position/ uniformity of response TRU trigger data resolution Shower shape (clustering & unfolding) Hadron rejection by energy deposit (E/p) November 8, 2010 EMCal Commissioning, T. Awes 26
Beam Test Analysis • Data Production (Yaxian) – PS and SPS data processing complete • Data processed with different Peak Extraction methods • Temperature dependence corrections (Evi) – APD gain is temperature sensitive ~ -2%/C – Temperature variations of more than 6 degrees C occurred • Extracted analyzed by Bjorn and Andrew • Periods (runs) when the temperature information was not recorded Tower correction vs T November 8, 2010 EMCal Commissioning, T. Awes 27
Beam Test Analysis • Temperature dependence corrections - continued – Gain correction factor for time (Temperature) dependent gain variation: • Signal for tower i is corrected as: Si = Fi(t) * Si(t) – where Fi(t) = LEDi(t)/LEDi(t_reference) » where LEDi(t) = LEDi(t)/ LED_mon(t) » where t may simply be Run # – Gain corrections from LED data were determined for all Beam Test runs by Evi – Confirmed by Alexsei to improve electron resolution for Beam Test • Procedure works - Apply to ALICE data Correction Factor – Being prepared by Evi Bias not set? Correction vs Run (PS period) Tower correction vs T November 8, 2010 EMCal Commissioning, T. Awes 28
EMCal/DCal Beam Test Some examples of TB analyses relevant to Physics analysis: • Fast Peak extraction algorithms – Use as Benchmark data set to test Fast Peak extraction algorithms (HLT) • Parallel processing of data with different algorithms and comparison of results on Energy Resolution, Timing – Per Thomas (Yale), David (ORNL) • Clustering – Need to modify clustering algorithm to Nx. M around maximum tower (N, M=3 -5) • Optimize algorithm (N, M vs E) by checking E resolution and non-linearity – Mateusz (LBNL), Deepa (Utrecht) • Non-linearity & Resolution – Need to correct cluster energy for “non-linearity” • Extract non-linearity and resolution for varying Zero-Suppression thresholds, different clustering algorithms; Compare to simulation; Apply to data – Alexei(WSU), Evi (ORNL), Shingo (LBNL) • Shower shape – Photon ID, unfolding (especially in Pb+Pb) – Adam (Nantes), Mateusz (LBNL) • • Etc. Summary of situation: Analysis of Test Beam data has stalled… November 8, 2010 EMCal Commissioning, T. Awes 29
EMCal Calibration: Non-Linearity Non-linearity Correction deduced from Simulations (Evi, Shingo) Non-linearity Observed From analysis With tight asymmetry cut (Olga) • Cluster energy must be corrected for non-linearity – Expected due to effect of energy lost in towers below Suppression Threshold • Fractional energy increases for as shower energy decreases – Problem: Non-linearity apparent from E-dependence of mass appears stronger than observed in simulation November 8, 2010 EMCal Commissioning, T. Awes 30
EMCal Calibration: Non-Linearity From Beam. Test (with ZS>75 Me. V not ZS>30, Overestimate of non-linearity) • Beam Test data can be analyzed with same Zero Suppression thresholds as used for Data Analysis – First analysis by Alexei shown here was done with ZS thresholds too high – Non-Linearity from Beam Test data better agree with that observed with AND with the non-linearity extracted from simulation • If simulation doesn’t agree it’s a problem! November 8, 2010 EMCal Commissioning, T. Awes 31
EMCal Geometry Errors • Simple analytical calculation. – If EMCal further away than believed, opening angle is overestimated, cos is underestimated, and mass ~(1 -cos is overestimated. • ~1. 5% overestimate of 0 mass for 7 cm distance error - Translates to error on Energy calibration with resulting error on spectrum. – If geometry used has SMs incorrectly displaced relative to each other, opening angle is underestimated and 0 mass is again overestimated with non-linear effect • • Effect increases with increasing p. T - larger error on the opening angle We need to study the 0 peak mass and width vs p. T using pairs in single SM to investigate the magnitude of this problem. – Dangerous to be making relative calibrations with 0 peak across SMs until this is understood November 8, 2010 EMCal Commissioning, T. Awes 32
EMCal Calibration: Non-Linearity From Gustavo • Dependence on SM combinations of gamma pairs - early results – MIP calibrations – Ideal geometry (with errors) – Results highly dependent on SM combination - clearly a problem November 8, 2010 EMCal Commissioning, T. Awes 33
EMCal Calibration: Non-Linearity From Gustavo • Dependence on SM combinations of gamma pairs - latest results – High statistics calibrations – Correct local SM geometry (definitely correct, since 2 weeks only!) – Misalignments based on charge independence of residuals • Using a ~8 cm radial offset from Ideal Geometry November 8, 2010 EMCal Commissioning, T. Awes 34
EMCal Geometry Saga • Initial concerns about EMCal geometry arose after first look at trackcluster residuals months ago (Francesco, Shingo) – Large offsets, sign dependent, position dependent • A simplified transparent code was implemented to adjust alignments (Francesco) • Track projection methods investigated (Rongrong, Deepa) • Official geometry code understood and used to adjust alignments (Rongrong) • After long effort, comparisons resulted in corrected official geometry (0. 75 tilt error, smaller module size) and simplified geometry. – It is finally certain that the EMCal local SM geometry is correct! – Both independent analyses give consistent offsets based on tracking residuals - but concerns now about tracking… Sector number Sector 0 Sector 1 November 8, 2010 OFFSETs (cm) Coordinate Francesco X 1. 08 Y 8. 35 Z 1. 12 X -3. 17 Y 7. 56 Z 1. 55 Rongrong 1. 134 8. 2 1. 197 -3. 093 6. 82 1. 635 d. X(+) d. X(-) d. Y(+) d. Y(-) d. Z(+) d. Z(-) Sector 0 mean 0. 0006 -0. 02 sigma 2. 14 2. 51 0. 08 -0. 05 1. 88 2. 42 EMCal Commissioning, T. Awes Sector 1 mean 0. 03 -0. 02 -0. 01 0. 09 -0. 05 sigma 2. 15 2. 53 0. 8 0. 9 2. 24 2. 64 35
EMCal Geometry Alignment d. Y Residuals (cm) vs p. T for varying SM radial offsets (cm) for e+/- From Rongrong • d. Y residuals become sign and p. T independent for radial offsets of SMs by about 8 cm November 8, 2010 EMCal Commissioning, T. Awes 36
EMCal Geometry Alignment From Rongrong • However, 8 cm radial shift over-corrects d. Z vs Z residuals, which prefers a radial shift of only ~4 cm – 4 cm radial shift is in better agreement with Survey information – Implies we cannot trust track projection to EMCal… November 8, 2010 EMCal Commissioning, T. Awes 37
EMCal Calibration: Non-Linearity 8 cm 4 cm From Gustavo • Dependence on SM combinations of gamma pairs - latest results – High statistics calibrations – Smaller radial shift gives better consistency between different SM combinations • Further improvements by moving SM sectors…? • In contradiction to conclusions from Tracking residuals November 8, 2010 EMCal Commissioning, T. Awes 38
TPC tracking From Rongrong • Some strange “structure” seen in and distributions for tracks extrapolated to outer TPC. – Can we trust extrapolation to EMCal? ? ! November 8, 2010 EMCal Commissioning, T. Awes 39
EMCal Calibration: Non-Linearity From Olga • Non-linearity of mass from MC - latest results (today! Not latest geometry? ) – Tendency to plateau, as expected – But second rise starting at 3 Ge. V, not expected from gamma response • This is almost certainly due to bias on invariant mass distribution as showers merge, cutting away loss mass acceptance • Need to limit shower sizes, and/or unfold clusters November 8, 2010 EMCal Commissioning, T. Awes 40
EMCal Cluster Unfolding From Adam • Clustering algorithm in production considers all contiguous towers to be part of cluster – High p. T clusters begin to merge at ~6 Ge. V/c – Significant problem for high occupancy of central Pb+Pb • Need to limit cluster sizes (Deepa) and/or unfold clusters (Adam) November 8, 2010 EMCal Commissioning, T. Awes 41
Some Issues: Strange Events 3 PHOS modules • Strange high energy single cell clusters first reported by PHOS – One week old news • Should “go away” – Physical? • Uniform distributed in PHOS, but high E single tower =/= shower • APD? Electronics? • Checked EMCal (Alexei) – Don’t see PHOS effect but do see strange small clusters November 8, 2010 EMCal Commissioning, T. Awes 42
Some Issues: Trigger • Correlation between “trigger energy” (FALTRO) and reconstructed energy for 3 TRUs of 4 SMs – Good correlation in most TRUs, but some deviations that need to be investigated • Also, we have only started to check expected L 1 trigger in performance in offline November 8, 2010 EMCal Commissioning, T. Awes 43
EMCAL HLT Analysis Chain • On-line reconstruction for calibration and monitoring • Event rejection: relevant triggering observables in Pb-Pb: HLT ANALYSIS CHAIN – High-ET (cluster) trigger – Jet trigger federico. ronchetti@cern. ch November 8, 2010 EMCal Commissioning, T. Awes 44
EMCAL HLT Status Calibration & Monitoring: • on-line calibration: π invariant mass (CALO/OK) • Online track matching (CALO/OK) • L 0 efficiency and turn on (EMCAL/underway) • Histogramming component (EMCAL/OK) • Event display: proper representation of digits and clusters (OK) Reconstruction: • Signal extraction: crude, peak finder (OK) • EMCAL Geometry: row-col based algorithm which includes a first evaluation of misalignments (OK) • Clusterizer: (VC, Nx. N, tested with pp data) Triggering: • Software triggers: calibration trigger handling in HLT for running EMCAL in mode B/C in the global HLT chain • Threshold triggering component (CALO/OK) • V 0 reconstruction component (J. Thader, testing) for centrality and background estimation in EMCAL jet trigger November 8, 2010 federico. ronchetti@cern. ch EMCal Commissioning, T. Awes 45
EMCal Commissioning • Summary: – EMCal has operated extremely well during 2010 run period • EMCal DCS control and monitoring fully integrated in ALICE • Very few bad channels or problems (mostly RCU/BUSY related) – Minimal “EMCal On Call” attention required • EMCal L 0 trigger commissioned and successfully operated – Very successful Beam Test • Extensive set of measurements to fully understand EMCal (as built) response – EMCal Calibration • • Geometry issues still not fully resolved Apparent Non-Linearity as seen from mass not fully understood Cluster unfolding urgently needs further development Relative tower calibrations are quite good – A great deal of effort has been made (thanks to all, especially those I haven’t acknowleged) • Still a lot of work to do… • Some explanation for the slow progress: http: //www. nature. com/news/2010/101013/full/467775 a. html November 8, 2010 EMCal Commissioning, T. Awes 46
- Slides: 46