You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 4 Next »

Agenda:

  1. Update on Kalman Filter                — Robert
  2. Update on MC                                — Tongtong
    1. Displaced vertexes and EGS5 target simulation (?)
    2. Moving to using mc-tools.
  3. AOB/Question

Notes:

Robert: 

Does not see much effect of using a non-uniform field vs uniform field. Code is working either way, but non-uniform field does not seem needed.
The puzzling issues with the single track events are now resolved. The momentum distributions look good. The chi-squared distributions look OK but the mean is higher (expected 12). Pulls look good for the fit but are 20% higher than unity (ideal). This could be due to non-gaussian scattering and the hit-errors.  Residuals are =< 8microns. Seem okay.

Helix parameter pull distributions are slightly larger than unity, but looks ok. Mean of the pt is not quite zero, but this could be many different things. Comparing the fit results between Kalman and GBL: the Kalman sees more tracks: Kalman distribution has larger tails, but also 44% more entries. 

PF: the GBL does not find tracks, it just re-fits. The Kalman finds tracks.

Looked at 2015 A' MC, with 1000 events. Comparing GBL and Kalman, biggest difference is in the chi-squared, but it is not clear if this is a 1-to-1 comparison on the definition of chi2.

Showed some individual events, some that seem to have a hard (larger angle) scatter. There are cases where some (probably recoil) tracks are not found. 

 

Tongtong:


 

 

  • No labels