BCAL Reconstruction Issues

From GlueXWiki
Revision as of 18:34, 19 June 2013 by Wilevine (Talk | contribs)

Jump to: navigation, search
  • Simulation (mcsmear)
    • fADC timing
      • Uses threshold crossing time rather than FPGA-ish algorithm
      • Resolution is too poor
    • mcsmear doesn't seem to smear data enough
      • TDC timing resolution is too good
      • single photo-electron peaks in energy spectrum
    • spikes in time histogram at low energy
    • max ADC amplitude (12 bits?)
    • Sampling fluctuations not applied on per-particle basis
    • Very low energy ADC hits coming from mcsmear
    • Short attenuation length
  • Timewalk corrections
    • Establish procedure for doing these corrections based on simulated data (how closely should this match the correction procedure for real data?)
    • Determine form of timewalk corrections
    • What if we have TDC info only on one end?
  • DBCALPoint (reconstructing E,z,t from double-ended hit)
    • When determining energy, give more weight to higher energy hit? (also in averaging time?)
    • Best way to determine sigma_z?
    • How to deal with multiple hits in same channel?
  • Clustering
    • Use only timing information from fADCs?
      • If we use TDC info also, we need to appropriately make use of the different timing errors of different hits
  • Determining cluster properties (E, x, y, z, t)
    • How to average cell-by-cell position to get cluster position
      • Weight averages by...? E? E^2? 1/sigma^2?
      • Average rectangular coordinates? Cylindrical? Spherical?
    • Energy corrections (z-dependent, non-linear)
      • Is current scheme good enough?
      • If so, write a plugin to automate this process.
      • Need extra special treatment of ends of calorimeter?
    • How to determine errors (and correlations) on cluster properties (important for kinematic fitting)
  • Matching with tracks
    • How to determine if cluster matches track
    • How to determine TOF if more than one matching cluster
      • This is a minor issue since very few (none?) will have >1 match