status of global tracking and plans for run2 (for tpc related tasks see marian’s presentation) 1...
DESCRIPTION
3 Including the TRD in the tracking (RefitInward) Gain: factor ~ 2 improvement in the p T resolution Currently prevented by insufficient alignment of TRD (depend on the global alignment task) Need to improve TRD tracklet fit (outlier clusters): in the current MC to see improvement in p T resolution we need to apply a cut on TRD χ 2, which kills 30-40% of TRD matches Uniformity problem: even with complete TRD ~30% of tracks will not be matched → keep copy of tracks with and w/o TRD in the refit? (will require repeating the TPC and ITS RefitInward step twice) Attempt to use TRD in tracking with current data is in progress (with standalone TRD alignment) https://alice.its.cern.ch/jira/browse/PWGPP-1... PWGPP-8 https://alice.its.cern.ch/jira/browse/PWGPP-1 MC anchored to LHC13bTRANSCRIPT
1
Status of global tracking and plans for Run2
(for TPC related tasks see Marian’s presentation)
R.Shahoyan, 19/03/14
2Global alignment/calibration framework
Need to realign the detectors before Run2 o SPD filters drilling affects ITS alignmento new SPD modules will appearo Current TRD alignment is not satisfactory
Currently ITS and TPC are aligned internally, then they are aligned globally one with respect to other Outer detectors alignment is relying on TPC tracks, w/o any feedback on TPC alignment calibration
Better constraint would be imposed by the global alignment/calibrationÞ Use Millepede algorithm for simultaneous refitting of detector’s alignment/calibration parameters and tracks
(already used for ITS and Muon alignment).
The task is planned since > 1 year, but no much progress due to the other priorities
Principal problem to solve: track model for the alignment:Millepede needs derivatives of residuals vs track parameters AND global cov. matrix – not provided by the Kalman
Need to evaluate 2 different approaches: o Using Kalman + smoother (LHCb): mathematically difficult in ALICE (rotations between tracking frames)o Global fit with explicit scattering kinks (ATLAS): numerically expensive
Central unit
Steering
Millepede minimization
ESD+Friends(or filtered
TrackPoints)
ITS
TPC
TRD
TOF
…
parameters to optimize
tracks, clusters
residuals + their derivatives
geometryOCDB
Detector-specific unitsDerived from template class
Declare DOFs and calculates track-cluster residuals wrt these DOFs
3Including the TRD in the tracking (RefitInward)
Gain: factor ~2 improvement in the pT resolution Currently prevented by insufficient alignment of TRD (depend on the global alignment task) Need to improve TRD tracklet fit (outlier clusters): in the current MC to see improvement in pT
resolution we need to apply a cut on TRD χ2, which kills 30-40% of TRD matches Uniformity problem: even with complete TRD ~30% of tracks will not be matched
→ keep copy of tracks with and w/o TRD in the refit? (will require repeating the TPC and ITS RefitInward step twice)
Attempt to use TRD in tracking with current data is in progress(with standalone TRD alignment)https://alice.its.cern.ch/jira/browse/PWGPP-1 ... PWGPP-8
MC anchored to LHC13b
4Finalization/validation of HLT seeded offline TPC reconstruction
Need to add to HLT output the track->cluster associations
Consider different options: o If the HLT track finding is reliable (lost cluster, fake clusters), consider only track refit in the
offline instead of seeding + afterburner for tracks not found in HLTo If HLT track finding is efficient at low pt, consider:
cluster reduction foreseen for Run3: only clusters attached to HLT tracks are stored(and in tracks’ proximity if the cluster attachment efficiency/purity is not good enough)
store HLT tracks/clusters within certain Z range from IP ;Z, pT ranges need to tuned to preserve tracks for loopers, cosmics tracks, some sample of events unbiased by trigger
In Run2 the TPC cluster data will be dominated by the pile-up → suppressing unnecessary tracks/clusters will lead to huge gain in raw data size, CPU speed and memory footprint
Update from Sergey in the end of March – beginning of April
Sergey, PWGPP-trk, 02/05/2013
5Standalone ITS reconstruction in the HLT
Not really global tracking but should be done in Run2 (“consolidation meeting, 14/03/14, https://indico.cern.ch/event/307253/) for:o TPC drift speed calibration (statistics → CPU speed is not an issue)o Vertexing with tracks for Luminous Region measurement with ~1 minute periodicity:
statistics/efficiency at pT>0.5 GeV is important but this task can be performed also byTPC/ITS tracks in HLT if the drift speed calibration is reliable
Rely on SPD + SSD points (+ vertex constraint, since secondaries reconstruction is irrelevant) Default plan is that the CA tracker will be adapted to SPD+SSD by I.Kisel group
(as prototype for Run3 main ITS tracker).First version was pledged for May 2014 but no reassuring news yet.PhD studen (M.Puccio) spent 2 weeks in GSI to start development of the tracker.
Alternative (proposed by Sergey at yesterday HLT training): use tracklets from current HLT SPD vertex finder as seeds to be completed by SSD points.o Not very large efforts to developo Worth to have it as a backup solution in case the CA tracker is not fully operational
by the end of the year
6Reduction of reconstruction memory footprint, disk output
Requires routine (periodic) optimization. Trying more economic base track parameterization.For the TPC: https://indico.cern.ch/event/302336/contribution/1/material/slides/1.pdf (Marian, Mikolaj)
Setting up the system for routine validation of simulation/reconstruction
Check quality of new releases/tags using standard set of dataStarted by Marian and Mikolaj…
ITS-SA seeded reconstruction of low-pT short tracks in TPC
TPC standalone tracking imposes certain requirements on the track length and seed pointsLow pT tracks (for pions: 100-150 MeV) are seen only in the IROC (looping) May require merging of the loop segmentsMore relevant for Run3, where ITS will lack the PID, hence it is important to decrease as much as possible the TPC pT threshold.