Trackersoftwarereview-mtg7 » History » Revision 4

« Previous | Revision 4/11 (diff) | Next »
Dobbs, Adam, 08 September 2017 12:23

Tracker Software Review: Meeting 7

September 8, 2017: 1400 BST


  1. Introduction - K. Long
  2. Actions from previous meetings
    • PR:
  1. AD: Optimize Chi2 cut for “MINUIT” variant
  2. AD: Show individual residual distributions to understand details - is the station 1 distribution a binning artifact or a constraint from the fit?
  3. AD, CH: Check and implement MCS errors
  4. AD: Plot chisq per dof for 'perfect' events
  5. Define multi-track requirements
  6. AD: Compare data/MC efficiencies for 8681 Done
  7. AD: Show distributions with sign per station per tracker Done
  • Track fit:
    1. Repeat study that was done to investigate whether field integration (e.g. using Runge Kutta) is required to compensate for magnetic field non-uniformity
    2. Make study of magnetic-field alignment with a view to establishing whether the present algorithm is sufficiently insensitive to reasonable assumptions of maximum field-misalignment
    3. Study p-value: split up contributions to p-value shape using MC and study shape in data, e.g. is there an error in the resolution per plane, handling of MCS, handing of energy loss, b field..
  • Tracker, MC:
    1. MU/EO: Understand reasons for shape of hits-per-station histogram - low-level
    2. PK: Check dead channel & noise handling in MC.
    3. Tracker group: Summarize, show what has been studied so far for systematics (alignment, scattering, energy loss, field uniformity, etc).
  1. Report on actions: Pattern recognition: A. Dobbs
  2. Report on actions: Track fit: C. Hunt
  3. Next Steps for review
  4. AoB

Dial-in information

Attendance: KL, CR, PK, AD, CH, MU, DR

Notes from Meeting 5: Aug 16, 2017:

  • AD: Compare data/MC efficiencies for 8681
    • done. 8681, 2.8.5 MC reprocessed with 2.9.1-PatRec,Kalman -- results consistent with 2.9.1 reco
    • Illustrative plots added
  • AD: Optimize Chi2 cut for “MINUIT” variant
    • pushing cuts lower gives higher efficiencies, though purity may be an issue. Noted that critical parameter is efficiency
    • Q: how loose should the LSQ cuts be? AD points to loosest cut in table from previous meeting.
  • AD: Show residual distributions with sign per station per tracker
    • sent to email list, added to meeting wiki
  • AD: Plot chisq per dof for 'perfect’ events
    • stands
  • Define multi-track requirements
    • stands. Need discussion and inputs on how to define
  • Adding errors:
    • for circle fit: suggestion is to define a mean error and apply to every station rather than station-by-station errors.
    • Q: how to define mean error? based on the geometric mean of the errors?
    • Q: does taking the chi^2/mean_error^2 bias the pr fit?
    • errors on the resulting circle fit parameters will feed into the longitudinal fit
  • Discussion about aperture cut
    • CR notes that aperture cut should be standardized by tracker group. AD notes that the cut in PR is loose -- 150.
    • PK suggests a flag to indicate if track passed or failed an aperture cut
    • All agree that fiducial cut should be standardized,
    • Q: does the current aperture cut throw out any potential tracks? If it does, cut should be loosened, but consensus seems to be that tracking should not do any "analysis-style" cuts but rather optimize track-finding.
  • Effect of magnetic field non-uniformity, alignment on track fit -- CH had issues reading mag field map from Joe Langland's program. CH will check against comsol map for 7469.

Updated by Dobbs, Adam almost 7 years ago · 4 revisions