Repository navigation
Conversation
A TPC-only track has no time measurement: its time0 is the tracker's guess (GPUTPCGMTrackParam::ShiftZ: z = 0 at the beam line if the helix can come from it, otherwise the latest cluster at z = 0.521 * x, then clamped to the drift volume). For a cosmic, which does not come from the beam line, the two legs get independent guesses, typically tens of mus apart, so their clusters were distortion-corrected at different z and the legs disagree (y, snp, q/pt) although they are one track. For legs on opposite sides the time comes from z continuity, but legs on the same side were compared as tracked. New MatchCosmicsParams refitSameSideAtCommonTime (default false, not in the preset physics-v1, see below): after the cheap cuts (tgl, q/pt, same half), same-side TPC-only legs are refitted at the centre of their brackets' overlap (the time the refit of the winners uses) and compared there (MatchCosmics::refitSeedAtTime: TPC refit from the outer parameters, then the propagation to the DCA as for the seeds; energy loss along the muon's flight as in refitWinners). The TPC refitter is created once per TF in process() and shared with refitWinners. The matcher debug tree writes the legs compared and a commonTime flag. Cosmic MC with PbPb-like distortions (190 TFs, preset physics-v1, off -> on): correctly matched muons 6848 -> 7324 of 8639 (79.3 -> 84.8 %); same side 77.0 -> 84.2 %, its loss to the crude chi2 cut 6.9 -> 0.8 %; A×C unchanged; 0 wrong-muon pairs in both; matcher +12 ms per TF (~100 refits). With the true MC time instead of the window centre the same-side pass rate of the chi2 cut would be 99.2 % (centre 98.0 %, legs at their own time0s 84.7 %). PbPb 2025 (4 TFs): ~530 refits per TF, +15 ms per TF. Not in the preset: in PbPb (568041, 310 CTFs, 16541 TFs) it adds mostly pairs of collision tracks, whose time0 was about right and which the refit at the window centre mis-corrects: +352 same-side cosmics but only +5 TOF-tagged (2.6 % of the new ones vs 15.3 % of the common ones), and ~35 A×C cosmics lose a leg to a refitted same-side pair. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
New MatchCosmicsParams tofFlightSelection (default off, on in the preset physics-v1; needs TOF clusters, the workflow then adds them to the matcher's inputs): every pair of TPC-only legs accepted by checkPair is tested for a top / bottom TOF cluster pair along the outward continuations of its legs whose time difference matches the muon's flight between them along the helix (tofFlightTolerance), with the mean time in the pair's time window and both legs' y and z within tofRoad at that time (the test of the cluster collector). A confirmed pair takes the TOF time as its common time (error tofTimeError), so the winner is refitted at it, and is put in front of all unconfirmed partners of both legs (chi2 order within each group): the mutual-best selection prefers it, and a random pair can no longer take a leg of a TOF-confirmed cosmic. If the refit of a confirmed winner at the TOF time fails (in practice chi2Match > maxChi2Match), it is refitted at the time it has without its TOF flight pair (z-continuity time of legs on opposite sides, else the centre of the overlap of the legs' time brackets): the winner selection has already released the other candidates of both legs, so dropping the winner loses both legs. PbPb 568041: 24 TOF-tagged same-side cosmics, all real (legs on one line far from the beam line, TOF pair in the physical order without a reversed-order alternative), mostly on the C side nearer the readout at the TOF time; A-C cosmics show no TOF-TPC time offset there. The refit loop body is a lambda refitting a winner at a given time. The TOF candidates of a leg are searched once per TF, in the time-sorted TOF clusters of its time bracket and in the sector it points to and its neighbours. Debug tree "match": tofScore. Distorted cosmic MC (19 x 10 TFs, PbPb-like distortions, physics-v1 without the common-time refit, without -> with tofFlightSelection): matched muons 6848 -> 6871 of 8639, 0 wrong pairs both, 2134 confirmed pairs, 9 fallbacks. PbPb 568041 reconstructed from CTF (310 CTFs, 16541 TFs, vs. the same CTFs without the TOF selection, older code, before the fallback): cosmics 4782 -> 4760, TOF-tagged 681 -> 697; A-C: 11 untagged pairings replaced by 12 TOF-tagged ones. Matcher CPU 0.536 -> 0.545 s per TF. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
matthias-kleiner
force-pushed
the
cosmics-matcher-writer-clean
branch
from
October 8, 2026 14:04
18a44b6 to
9a493f4
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Your PR AliceO2Group#15915 cannot be merged as is. You should either run
clang-formatyourself and update the pull request, or merge this PR in yours.You can find the AliceO2 coding conventions at https://github.com/AliceO2Group/CodingGuidelines.