150 likes | 268 Views
Pass0 – Status. M. Ivanov, J. Otwinowski , C. Zampolli with the collaboration of L. Betev Offline Week, 5-9 July 2010. Pass0. Pass0 is a dedicated implementation of the offline calibration strategy Reconstructing ITS, TPC, TOF and TRD only data on a subsample of events per each chunk
E N D
Pass0 – Status M. Ivanov, J. Otwinowski, C. Zampolli with the collaboration of L. Betev Offline Week, 5-9 July 2010
Pass0 • Pass0 is a dedicated implementation of the offline calibration strategy • Reconstructing ITS, TPC, TOF and TRD only data on a subsample of events per each chunk • Running the calibration for the TPC (TRD and ITS, see later) • Merging over the various chunks in the same run • Extracting calibration OCDB parameters • Inserting the calibration parameters in the OCDB (run-level files) 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 2
Status as March ‘10 Strategy – Run-level offline calibration Post-processing job Triggered by the end (~95%) of the previous sub-jobs (*) ESDs+ friends ESD friends_v1 Chunk 0 Calibration objects file i ESDs+ friends ESDs friends_v1 Chunk n ESDs+ friends ESDs friends_v1 Chunk 0 Calibration objects file j ESDs+ friends ESDs friends_v1 Chunk n (*) Details defined after some tests 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 3
Strategy – Pass 0 • 4% level central filtering has been moved within the reconstruction • Friends for all tracks for 4% of the events (randomly chosen) are kept • Additional filtering not needed for Pass 0 • Before inserting any object in the OCDB, validating the calibration results Post-processing job Triggered by the end (~95%) of the previous sub-jobs (*) ESDs+ friends ESD friends_v1 Chunk 0 Calibration objects file i ESDs+ friends ESDs friends_v1 Chunk n (*) Details defined after some tests 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 4
Ingredients • Reconstruction + Calibration (Pass 0) • Jdl: • recPass0.C • Standard reconstruction macro • Only ITS, TPC, TOF, TRD • Only on a limited number of events • Switching off friend filtering at the moment • ConfigCalibTrain.C • OCDB initialization • Magnetic field and Geometry from GRP • runCalibTrain.C • Setting the Pass 0 calibration train • Adding the “wagons” 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 5
Ingredients – II • Jdl (cont’d): • AddTaskTPCCalib.C • Setting the TPC Pass 0 calibration tasks • Reading the necessary information (start/end of run) from GRP • LoadLibraries.C • Loading the needed libraries (general to run analysis train + TPC specific) • Executables: • runPass0.sh • Only one executable needed, for both reconstruction and calibration • validation.sh • Validating the output of the reconstruction and the calibration 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 6
Ingredients – III • Merging and OCDB populating procedure • Jdl: • ConfigCalibTrain.C • To instantiate OCDB • LoadLibraries.C • To load the necessary libraries (general to run analysis train + TPC specific) • merge.C • To merge chunk-level calibration outputs • Uses AliFileMerger • makeOCDB.C • Calling AliTPCPreprocessorOffline • Generating the Gain ad Drift Velocity calibration parameters • Updating the OCDB 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 7
Ingredients – IV • Executables: • mergeMakeOCDB.sh • Only one executable to both merge and produce the OCDB entry • validationMerging.sh • Validating the output of the reconstruction and the calibration • Calibration parameter validation procedure • Still to be properly defined (see later) 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 8
Validation Procedure for TPC (scenario I) • Validation needed before populating the OCDB • Based on the comparison between corrections from: • TPC+ITS used as reference • Laser, if laser events are there among the selected ones • TPC+TOF, if TOF was included during data taking • (TPC+TRD, to be implemented) • Only when the 3 (4) corrections agree, the object should be stored in the OCDB • If they don’t, the run WON’T have the proper calibration (Pass 0 provides a run-level calibration) • After the validation, the object storing in the OCDB should be automatic • At a first stage, some manual intervention can be foreseen 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 9
Validation Procedure for TPC (scenario II) • Validation needed before populating the OCDB • Based on the comparison between corrections from: • TPC+ITS used as reference • Laser, if laser events are there among the selected ones • TPC+TOF, if TOF was included during data taking • (TPC+TRD, to be implemented) • If TPC+ITS corrections in the window → update OCDB (OCDB object flag 1) • If TPC+ITS & Laser in the window → update OCDB (OCDB object flag 2) • TPC+ITS & Laser & TPC+TOF in the window → update OCDB (OCDB object flag 3) • TPC+ITS & Laser & TPC+TOF & TPC+TRD in the window → update OCDB (OCDB object flag 4) • If they don’t in the window, the run WON’T have the proper offline calibration (Pass 0 provides a run-level calibration) 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 10
More Wagons • In the last months, new wagons asked to be added: • TRD • Tested on the Grid (C. Zampolli, R. Bailhache, J. Otwinowski) • Same structure as TPC (AddTaskTRDCalib.C, AliTRDPreprocessorOffline, TRD libraries needed) • AliTRDPreprocessorOffline committed lately in the trunk – to be tested on the Grid • Validation? • ITS • Alignment task, producing – exceptionally – a tree • While being tested on the grid, some changes in ANALYSIS classes had to be made now tests can be restarted • No need to populate the OCDB! 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 11
A few words on the Offline Calibration • Same strategy as for Pass 0 – with very few differences: • Pass 0: • Reconstruct (*) the ITS, TPC, TRD, TOF only data on a subsample of events per each chunk (more detectors could be needed) • Run the calibration for the TPC (TRD, ITS) • Merging over the various chunks • Calibration parameters put in the OCDB (not for ITS) • Pass 1: • Full reconstruction (*) • Additional filtering (see next slide) + Calibration for all the detectors • Merging of the chunks • Merging over different runs • Calibration parameters put in the OCDB (*) 4% central filtering on! 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 12
Filtering during Offline Calibration • As already presented in previous meetings, the offline calibration foresees to run on filtered events/tracks • On top of the 4% level central filtering occurring during the reconstruction, additional filtering procedures may be applied by each detector to filter at event/track level • Each calibration procedure should be implemented together with the corresponding filtering task • See: http://aliweb.cern.ch/Offline/Activities/Analysis/AnalysisFramework/FilteringAndCalibrationTrain.html 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 13
Open Issues • Finalization of the validation procedure • TPC, TRD, ITS testing • Should the macros ConfigCalibTrain and LoadLibraries be merged? • Folders in the AliESDfriends_v1.root files for each detector? • Triggering of merging (over chunks + over runs) + final calibration procedure to be defined AliEn experts 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 14
Open Issues and Comments • Any special requirement for the selection of the events for pass 0 has to be put in the rec.C • A macro to load all the necessary libraries may be created to configure the reconstruction+calibration job (to be called also when creating the calibration parameters to fill the OCDB) • Triggering of merging (over chunks + over runs) + final calibration procedure to be defined AliEn experts • Folders in the AliESDfriends_v1.root files for each detector? • Specific filtering algorithm to be defined • QA Train: when would it be run wrt reco+calibration? • If after Friends would be filtered! • Alignment: trees needed… 7/7/10 J. Otwinowski, C. Zampolli - Offl Week 15