Mattione Update 04252012

From GlueXWiki
Jump to: navigation, search

Summary

  • Promising start to single-track event pion spiral reconstruction, MUCH work still to do.

Spiral Fitting - Reconstructed Track Candidate Viewer Comparison

  • Notes:
    • All events have one π+ thrown at p = 150 MeV/c, θ = 90 degrees, φ = random.
    • Current Code = Code as of 04/08/12
    • Current Code # Reconstructed Track Candidates in events below: 4, 6, 5, 12
    • Spiral Code # Reconstructed Track Candidates in events below: 2, 1, 0, 1
    • Event 1: Spiral Code reconstructs two tracks due to turn near beamline, will be difficult to reduce to one track.
    • Event 2: Spiral Code reconstructs both spiral arms (from the beamline to the edge) and then links them together to form one track
    • Event 3: Not sure yet why spiral code fails to reconstruct anything.
    • Event 294: Current code ignores stereo hits while trying to link axial hits together by phi: explosion in track candidates
    • Event 294: Spiral code has memory explosion (leak?) while attempting to link axial and stereo hits together, need to improve somehow...
Event # Hits (CDC & Truth) Current (04/08/12) Code Track Candidates Spiral Code Track Candidates
Event 1
Mattione Update 04252012 SlowPiPlus Event1 Hits.png
Mattione Update 04252012 SlowPiPlus Event1 Current.png
Mattione Update 04252012 SlowPiPlus Event1 Spiral.png
Event 2
Mattione Update 04252012 SlowPiPlus Event2 Hits.png
Mattione Update 04252012 SlowPiPlus Event2 Current.png
Mattione Update 04252012 SlowPiPlus Event2 Spiral.png
Event 3
Mattione Update 04252012 SlowPiPlus Event3 Hits.png
Mattione Update 04252012 SlowPiPlus Event3 Current.png
Mattione Update 04252012 SlowPiPlus Event3 Spiral.png
Event 294
Mattione Update 04252012 SlowPiPlus Event294 Hits.png
Mattione Update 04252012 SlowPiPlus Event294 Current.png
Mattione Update 04252012 SlowPiPlus Event294 Spiral.png


Spiral Fitting - Histograms

  • # Reconstructed Tracks / Event:
    • NOTE: Hit selectors disabled for wire based & time based fits: just use hits from track candidates (and ignore hits after spiral turn back inwards (for now anyway))
    • Not sure why so few wire-based & time-based tracks reconstructed, possibly due to bad θ reconstruction.
Track Candidates
Wire Based Tracks
Time Based Tracks


  • Track Candidates: Track Parameter Differences (BEST Reconstructed - Thrown)
    • Best Reconstructed = Track Candidate that is the closest match to the thrown track.
    • NOTE: Δφ = 20 not real, due to artifact: reconstructed track candidate parameters are reported at a point along the track (~)just outside the start counter, whereas thrown parameters reported at thrown vertex (beamline)
    • Need to work on θ reconstruction.
pReconstructed - pThrown
θReconstructed - θThrown
φReconstructed - φThrown

Spiral Fitting - Procedure Overview

  • Link adjacent hits together within each superlayer
  • Link adjacent superlayer hits group together (all possible combinations)
  • Fit the axial superlayer hits (and the beamline) to a circle (for each unique axial group combination)
  • Reject bad axial group combinations (bad combo if one of its axial groups was already used in a fit that got a lower chisq/ndf)
  • Compare the different stereo superlayer hit group combinations to the circle fit, choose the best combination (best = combo with lowest standard deviation of reconstructed z-vertex hits)
  • Calculate track parameters (p, theta, phi, x, y, z)
  • Link spiral arms together (look for rings with many consecutive hits shared between spiral arms)