April 20-21, 2020 DIRC

From GlueXWiki
Jump to: navigation, search

Meeting Information

GlueX DIRC Meeting
Monday-Tuesday April 20-21, 2020
Remote only

Connection Using Bluejeans

  1. To join via Polycom room system go to the IP Address: (bjn.vc) and enter the meeting ID: 846453895.
  2. To join via a Web Browser, go to the page [1] https://bluejeans.com/846453895 .
  3. To join via phone, use one of the following numbers and the Conference ID: 846453895
    • US or Canada: +1 408 740 7256 or
    • US or Canada: +1 888 240 2560
  4. More information on connecting to bluejeans is available.


Monday, April 20

Tuesday, April 21

  • 9:00 - FastDIRC status --- Yunjie
  • 9:45 - Followup discussion and planning future work --- All
    • Defining performance expectations:
      • PAC proposal: 3σ at 4 GeV which was equated to 2.5 mrad track resolution
      • 2015 review KPP: 2.5 mrad track resolution
    • Others?
  • 11:15 - GlueX bi-weekly meeting

Action items

Here we'll collect a list of specific action items from the discussion to accomplish in the short (prior to reconstruction launch) and long term:

  • Utilize bar box rotations from survey in geometric reconstruction prior to other corrections
  • Validate ReverseKalman filter from tracking group
  • 2D map in pixel space of time offsets and resolutions
  • Obtain expectation from MC:
    • per-track resolution
    • Nsigma separation
    • SPR
    • photon yield
  • Systematic study of event selection criteria for pions and kaons and performance of geometric reconstruction (follows initial MC study, if time allows)
  • Validation of correction methods with simulation with nominal geometry and known offset (dedicated samples)
    • Per-PMT thetaC correction
    • Ring fit correction to track angle
    • Chromatic correction
  • Efficiency and mis-identification for different cut requirements on DLL
    • Additional tagged samples of electrons and protons for mis-ID matrix
    • Study non-gaussian tails and implications for analyses
    • Provide guidance for utilizing DLL in physics analysis: define loose and tight conditions
    • Post-reconstruction best practices (avoid pitfalls of user defined corrections)
  • Longer-term items with large Spring 2020:
    • Cherenkov PDF
    • Time-based imaging
  • For future running:
    • Complete analysis of threshold, HV variations from commissioning data (require production with reconstruction launch)
    • Coherent noise seen in H12700 at HADES, LHCb, and CLAS12: many hits seen across detector plane. Need to look for evidence in our data.

Discussion notes


  • Slide 2: Kaon distributions have more background SPR plot and likelihood distribution is wider for kaon than pions.
    • Could the timing performance still be worse for kaons than pions?
    • What else would cause kaon background to be higher?
    • How much does this background contribute to to the larger photon yield for kaons?
  • Slide 3: Preferred correction order is always to begin with per-PMT Cherenkov angle correction, since it keeps ring segments connected. Smearing by either track resolution (ring fit) or chromatic correction
  • Per-PMT Cherenkov angle correction
    • Some PMTs with bad correction, could be checked with MC?
    • Have we tried generating some MC with a mis-aligned geometry and validate that the corrections obtained through this procedure get back to the optimal resolution?
  • Ring-fit correction
    • Sensitive to start values for the fits: use average of pion and kaon expected values.
    • How do the the correction factors vary over the DIRC plane?
      • Can we determine a map of expected track corrections as a function of r, phi?
  • Chromatic correction
    • Possible to use MC with varying timing resolutions to see if it follows the expectation. From this we could determine what time resolution would be required to make significant chromatic correction
    • Suggestion to divide time difference by path length to improve performance (from SLAC experience)
  • Comparison of corrections
    • Need to keep track of both SPR improvement and difference between thetaC mean with improvements
    • Kaon DeltaLL narrows as corrections improve performance
  • Kaon photon yield is significantly higher than pion
    • Need to look back at MC for comparison of photon yield between pions and kaons
    • Should cut on thetaC for counting photons be symmetric around pion and kaon peaks separately?
  • Event selection for calibration sample:
    • Look at side-band of phi mass distribution in thetaC plot, quantify pion background
    • For cleaner sample: could use comparison of 2pi and 2k kinfit chi2 values (a la Alex Barnes's study for the phi SDME)
    • For some reason don't see peak for opposite PID, e.g. kaons peaking at positive DeltaLL under pion sample. Concern that timing cut could be biasing tracks which make it into final sample for DeltaLL. Some options to consider:
      • Select poorer kaon sample (e.g. with inverted kinfit chi2 cut) and see if mis-ID'd pions appear
      • Study event selection plots: mass distribution (phi/rho yield), kinfit chi2, etc. after positively identifying a kaon


  • Per-track Cherenkov angle resolution vs photon yield
    • Not applying corrections for per-PMT offsets or ring fitting for this study
    • Correlated term of ~2.5 mrad larger than expected
    • Suggestion: repeat same procedure for kaons in data and
  • For charge separated samples: plot DeltaThetaC on slide 7 in different regions of the DIRC plane (bar and x-position)
    • Do we see any evidence of charge separation?
  • Cherenkov PDF: statistically limited thus far
    • Expected factor ~20 increase with full Spring 2020: 200 B triggers collected, compared to ~10 B triggers in commissioning period
    • Addition of other tagged kaon samples: reactions with Lambdas, K*, Kshort etc. could increase by factors of a "few" with possible additional backgrounds
    • Changing to DeltaThetaC for PDF may also decrease statistics requirement
    • Come back to this study after a significant fraction of Spring 2020 data is available for analysis


  • Tracking resolution for both position and angles at the DIRC are represented in radial and perpendicular coordinates, which have different resolutions
    • FDC geometry doesn't prefer radial vs perpendicular coordinates, so why is there a difference here?
    • CDC however has a radial orientation which could be the reason for this difference that is propagated to the extrapolated tracks
    • Possible that this will improve or become more symmetric with the reverse Kalman filter. Could also turn off CDC hits to see how things change?
  • Comparison to TDR plots
    • Resolution in perpendicular coordinate is worse than expectation, but from simulation still only expect ~1 mrad contribution to correlated term
    • Need to repeat per-track resolution studies by Ahmed and Roman for simulation as well, to confirm if they're consistent with this level of a correlated term
  • Possibility for magnetic field variations from nominal field map that induce charge-dependent bias in the angle/position residuals
    • Roman reported that for the x position = 0 there is very little difference between the SPR and thetaC centroid positions. He will look at other bins which may be more sensitive
    • z position alignment or momentum shift could give similar charge-dependent bias, but need to quantify the magnitude


  • Add LED time resolution as a function of run number to identify calibration issues
  • Add ratio of hit yield/# LED triggers for each PMT as a function of run number
  • Prepare cron job to be used in future runs to monitor performance in "real time"


  • Photon yield result (slide 19)
    • Observe similar results as geometric reconstruction for larger kaon yield than pion, indicating this is not specific to the reconstruction method
  • Automated way to import HDDS XML geometry to FastDIRC
    • Script exists to pick out parameters from XML files to update FastDIRC geometry
  • How close are we to assigning a pi/K separation from FastDIRC at a particular momentum with systematics?
    • Prerequisite is to incorporate into reconstruction framework: systematic studies could then be done in parallel for geometric and FastDIRC as a comprehensive study
  • How to best use this independent framework?
    • Incorporate into GlueX reconstruction framework, on Yunjie's to do list
    • Port geometrical propagation to hdgeant4 for simulation speedup?
    • Global alignment of optical elements?


  • Data/MC comparisons are needed in many cases where we've primarily looked at data only. Tables which show performance improvements with different corrections would be good to have both data and MC for verification.