Difference between revisions of "BCAL Reconstruction Meeting 2014-09-18"

From GlueXWiki
Jump to: navigation, search
(Minutes)
(Minutes)
 
(One intermediate revision by the same user not shown)
Line 50: Line 50:
 
# Announcements: Zisis is planning to come to JLab for shifts, Oct 1-11, subject to the accelerator schedule.
 
# Announcements: Zisis is planning to come to JLab for shifts, Oct 1-11, subject to the accelerator schedule.
 
# Action Items:  Items 1 and 2 are being worked on; see Tegan's reports below.
 
# Action Items:  Items 1 and 2 are being worked on; see Tegan's reports below.
# BDT Update: Tegan gave a report (see link above) on what he and Justin have been working on.  The objective is to implement the BDT in the BCAL Reconstruction to see if it can improve over Will's cuts analysis (Looking for similar efficiency with higher background rejection and signal purity); also to import the BDT knowledge to Regina.  
+
# Update on BDT: Tegan gave a report (see link above) on what he and Justin have been working on.  The objective is to implement the BDT in the BCAL Reconstruction to see if it can improve over Will's cuts analysis (Looking for similar efficiency with higher background rejection and signal purity); also to import the BDT knowledge to Regina.  
 
## Overview: The events are generated by bggen (full hadronic photoproduction spectrum) requiring at least 2 showers/clusters in the BCAL.  The FCAL is not involved at this stage but will be in the future. The BDT was trained on a single cluster (gamma).   
 
## Overview: The events are generated by bggen (full hadronic photoproduction spectrum) requiring at least 2 showers/clusters in the BCAL.  The FCAL is not involved at this stage but will be in the future. The BDT was trained on a single cluster (gamma).   
 
## Observations:  
 
## Observations:  
Line 59: Line 59:
 
### (DL) DBCALIncidentParticle factory was written to capture Andrei/Irina's Sampling Fraction tables.
 
### (DL) DBCALIncidentParticle factory was written to capture Andrei/Irina's Sampling Fraction tables.
 
## Questions:
 
## Questions:
### (ES) Why the theta ranges? A: Theta is binned in ranges (instead of letting the BDT handle it) because it allows the extraction of finer dependencies over limited angular range, instead of a global dependency.
+
### (ES) Why the theta ranges? JS: Theta is binned in ranges (instead of letting the BDT handle it) because it allows the extraction of finer dependencies over limited angular range, instead of a global dependency.
### (WL) Is there a hard cut to eliminate noise? A: No.
+
### (WL) Is there a hard cut to eliminate noise? TB: No.
### (WL) Why Point>1 cut?  A: It is a global cut, which is done in BDT to simplify life ahead of time, by removing obviously unnecessary sorting of such a variable in BDT.  Nevertheless, even after the cut the Point variable is still used in the BDT classifier.
+
### (WL) Why Point>1 cut?  JS: It is a global cut, which is done in BDT to simplify life ahead of time, by removing obviously unnecessary sorting of such a variable in BDT.  Nevertheless, even after the cut the Point variable is still used in the BDT classifier.
### (ZP) What is z-coordinate cut? A: it is almost the full length of the BCAL, in the global coordinate frame, where the BCAL goes from 17-407cm.
+
### (ZP) What is z-coordinate cut? TB: it is almost the full length of the BCAL, in the global coordinate frame, where the BCAL goes from 17-407cm.
 
### (ZP) The ranking of the variables changes in the last few slides.  TB: The top 4 variables are always the same, and there is no significant difference in their ranking.
 
### (ZP) The ranking of the variables changes in the last few slides.  TB: The top 4 variables are always the same, and there is no significant difference in their ranking.
 
### (DL) tproj looks narrow.  TB: this is an artifact of the plotting.  Promised to circulate expanded scale plot after the meeting (and he did).
 
### (DL) tproj looks narrow.  TB: this is an artifact of the plotting.  Promised to circulate expanded scale plot after the meeting (and he did).
Line 68: Line 68:
 
### (ES) What are the horizontal lines in slide 5? TB: Those represent the hard cuts from Will, ie if we match the efficiency we get a better purity.
 
### (ES) What are the horizontal lines in slide 5? TB: Those represent the hard cuts from Will, ie if we match the efficiency we get a better purity.
 
## Conclusion: Will's hard cuts did a pretty good job.  More BDT work is needed as mentioned above.
 
## Conclusion: Will's hard cuts did a pretty good job.  More BDT work is needed as mentioned above.
# Timing code: Will said that it looks good.  Tegan took the time-walk out, but it needs to be replaced with new code, following the flash ADC timing. Tegan will look at time vs ADC to study the t-w.
+
# Update on Timing code: Will said that it looks good.  Tegan took the time-walk out, but it needs to be replaced with new code, following the flash ADC timing. Tegan will look at time vs ADC to study the t-w.
 
# Any other business: Zisis suggested that Tegan/Justin present their work at the Collaboration Meeting, and everyone agreed.  Zisis will email Mark Ito to schedule this.
 
# Any other business: Zisis suggested that Tegan/Justin present their work at the Collaboration Meeting, and everyone agreed.  Zisis will email Mark Ito to schedule this.

Latest revision as of 12:17, 22 September 2014

Video Conferencing Information

Meeting Time: 12:00 p.m. EDT

  1. To join via a Web Browser, go to the page [1] https://bluejeans.com/907185247.
  2. To join via Polycom room system go to the IP Address: 199.48.152.152 (bjn.vc) and enter the meeting ID: 907185247.
  3. To join via phone, use one of the following numbers and the Conference ID: 907185247.
    • US or Canada: +1 408 740 7256 or
    • US or Canada: +1 888 240 2560
  4. Upon connection all microphones are automatically muted. To unmute your mike on a Polycom or equivalent unit, enter *4. Unmuting on a computer is trivial as there is a microphone button than can be clicked.
  5. More information on connecting to bluejeans is available.

Participant Direct Lines

  • JLab Phone: in CC F326 is 757-269-6460 (usual room)
  • JLab Phone in CC L207 is 757-269-7084
  • Phone in the Regina Video-conference Suite is 306-585-4204

References

  1. BCAL Reconstruction Issues
  2. BCAL Reconstruction Algorithms
  3. Will's reconstructed energy info

Action Items

  1. Reconstruction code
    1. Timing: extracted from F1TDC and compared to fADC; check need for time-walk (fits, iterations, etc)
    2. Tegan: BDT presentation and timing plots at next meeting
    3. Sampling fraction tables
    4. Shower curvature tables: compare KLOE to IU codes: check gap in cluster handling between codes.
    5. z-coordinate determination from up/down amplitude ratio
    6. set variable “failures” to a number other than zero (e.g. -9999 or similar)
    7. comb code for constants to be pulled out into a database
    8. measure data pedestals using LV=0 setting (Mark)

Agenda

  1. Announcements
  2. Action Items
  3. Updates
    1. BDT Update: Presentation:File:BDT 2014-09-18.pdf (Tegan)
    2. fADC time-walk code: see emails from Aug 21 and Sep 17; (Will and Tegan)
  4. Any other business

Minutes

Attendees: David, Elton, Will M., Irina, Andrei (JLab), Will L. (CMU), Justin (MIT), Tegan, Zisis

  1. Announcements: Zisis is planning to come to JLab for shifts, Oct 1-11, subject to the accelerator schedule.
  2. Action Items: Items 1 and 2 are being worked on; see Tegan's reports below.
  3. Update on BDT: Tegan gave a report (see link above) on what he and Justin have been working on. The objective is to implement the BDT in the BCAL Reconstruction to see if it can improve over Will's cuts analysis (Looking for similar efficiency with higher background rejection and signal purity); also to import the BDT knowledge to Regina.
    1. Overview: The events are generated by bggen (full hadronic photoproduction spectrum) requiring at least 2 showers/clusters in the BCAL. The FCAL is not involved at this stage but will be in the future. The BDT was trained on a single cluster (gamma).
    2. Observations:
      1. (ES) pi0 peak is at 130 MeV resolution is 30 MeV (slide 5) and has a tail. This should be looked at more carefully. Red curves are signal, blue background.
      2. (WL) Observation on Slide 8. Will had communication with Richard Jones on a similar issue, namely that the truth info from HDGEANT is not behaving as it should. Secondaries are being reordered. Some showers start right outside the BCAL. Zisis will follow up.
      3. (DL) primary flag =1 comes from generator.
      4. (JS) It is good to have detector-level matching in the BDT.
      5. (DL) DBCALIncidentParticle factory was written to capture Andrei/Irina's Sampling Fraction tables.
    3. Questions:
      1. (ES) Why the theta ranges? JS: Theta is binned in ranges (instead of letting the BDT handle it) because it allows the extraction of finer dependencies over limited angular range, instead of a global dependency.
      2. (WL) Is there a hard cut to eliminate noise? TB: No.
      3. (WL) Why Point>1 cut? JS: It is a global cut, which is done in BDT to simplify life ahead of time, by removing obviously unnecessary sorting of such a variable in BDT. Nevertheless, even after the cut the Point variable is still used in the BDT classifier.
      4. (ZP) What is z-coordinate cut? TB: it is almost the full length of the BCAL, in the global coordinate frame, where the BCAL goes from 17-407cm.
      5. (ZP) The ranking of the variables changes in the last few slides. TB: The top 4 variables are always the same, and there is no significant difference in their ranking.
      6. (DL) tproj looks narrow. TB: this is an artifact of the plotting. Promised to circulate expanded scale plot after the meeting (and he did).
      7. (AS) Can you train the BDT in pairs of photons? JS: Training it on di-photon sample may occur in the future. It is more complicated.
      8. (ES) What are the horizontal lines in slide 5? TB: Those represent the hard cuts from Will, ie if we match the efficiency we get a better purity.
    4. Conclusion: Will's hard cuts did a pretty good job. More BDT work is needed as mentioned above.
  4. Update on Timing code: Will said that it looks good. Tegan took the time-walk out, but it needs to be replaced with new code, following the flash ADC timing. Tegan will look at time vs ADC to study the t-w.
  5. Any other business: Zisis suggested that Tegan/Justin present their work at the Collaboration Meeting, and everyone agreed. Zisis will email Mark Ito to schedule this.