Difference between revisions of "BCAL Reconstruction Meeting 2013-04-04"

From GlueXWiki
Jump to: navigation, search
(Minutes)
m (Text replacement - "/halldweb1.jlab.org/" to "/halldweb.jlab.org/")
 
(One intermediate revision by one other user not shown)
Line 32: Line 32:
 
# Updates
 
# Updates
 
## Will:
 
## Will:
### [https://halldweb1.jlab.org/wiki/images/7/7e/Bcal_offline_mtg_20130403.pdf Changes to BCAL Reconstruction]
+
### [https://halldweb.jlab.org/wiki/images/7/7e/Bcal_offline_mtg_20130403.pdf Changes to BCAL Reconstruction]
 
### [[BCAL_Reconstruction_Issues]]
 
### [[BCAL_Reconstruction_Issues]]
 
## Beni: [[BCAL reconstruction test with pi0]]
 
## Beni: [[BCAL reconstruction test with pi0]]
Line 47: Line 47:
 
## Will: Checked in code to do timewalk corrections in BCAL reconstruction. This is necessary to do reconstruction with David's new, more realistic BCAL simulation in HDGeant/mcsmear. In yesterday's Software Group meeting it was decided to switch over to the new simulation scheme by default. Discussion of KLOE vs. IU clustering algorithm continues, need more data.
 
## Will: Checked in code to do timewalk corrections in BCAL reconstruction. This is necessary to do reconstruction with David's new, more realistic BCAL simulation in HDGeant/mcsmear. In yesterday's Software Group meeting it was decided to switch over to the new simulation scheme by default. Discussion of KLOE vs. IU clustering algorithm continues, need more data.
 
## Will's slides (Changes to BCAL Reconstruction) generated much discussion related to how clusters should be formed, how events from the outer cells (no TDCs) be treated, likewise for events that may not have a TDC hit in the inner layers, and the KLOE vs IU behaviour (most important difference is cuts in z vs cuts in theta).  It became clear that the current HDGEANT has the (unrealistic) energy weighted time and we should be careful about drawing any conclusions from those simulations. Other discussion revolved around the order of doing things (e.g. match ADC-TDC info on one side and do time walk, or combine both sides in doing the time walk, a.o.)
 
## Will's slides (Changes to BCAL Reconstruction) generated much discussion related to how clusters should be formed, how events from the outer cells (no TDCs) be treated, likewise for events that may not have a TDC hit in the inner layers, and the KLOE vs IU behaviour (most important difference is cuts in z vs cuts in theta).  It became clear that the current HDGEANT has the (unrealistic) energy weighted time and we should be careful about drawing any conclusions from those simulations. Other discussion revolved around the order of doing things (e.g. match ADC-TDC info on one side and do time walk, or combine both sides in doing the time walk, a.o.)
## David's proposal: (a two-pass approach)
+
## David's proposal:
###
+
##* Clusters are formed using timing info from the fADC only
###
+
##* Timewalk corrections to TDC hits are applied at the individual hit level (i.e. single ended)
###
+
##* Final timing calibration for TDCs is done via [[Andrei's recipe for BCAL Timing Calibration|Andrei's recipe]]
 +
##* After clusters are formed, hi-res TDC info is used to define properties of cluster as a whole (z-position and time of impact)
 
## Will has not used Andrei's algorithm yet and questions arose later (offline) about how applicable it would be to real data.  We should indeed strive for producing code that would work on both real and MC data.
 
## Will has not used Andrei's algorithm yet and questions arose later (offline) about how applicable it would be to real data.  We should indeed strive for producing code that would work on both real and MC data.
 
## We need to consider whether to process based on Delta-z (mm) or Delta-time (ps).
 
## We need to consider whether to process based on Delta-z (mm) or Delta-time (ps).

Latest revision as of 10:31, 31 March 2015

Teleconference Time: 11:00 a.m. EST

  • ESNET (Number is 8542553) and EVO session (GlueX Calorimetry meeting room)
  • Phone connection only upon request.
    • +1-866-740-1260 : US+Canada
    • +1-303-248-0285 : International
    • then enter participant code: 3421244# (remember the "#").
    • or www.readytalk.com (and code without the #)

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 Videoconference Suite is 306-585-4204

Action Items

  1. Mar 7 Meeting
    1. Do we need time-walk on FADC timing? (David)
    2. Do we have 12 bits or more in the FADC? (David)
    3. Show plots with timewalk correction applied. (Will)
    4. Will's Slide 8 spikes in spectrum? Yellow bumps in Slide 4?
  2. Mar 14 Meeting
    1. Slide 2 expand scale; Slide 3 add error bars; Slide 4 Project on y-axis to compare fits (peak vs mean on Slide 5)
    2. Time-walk (t-w) correction methodology. (Andrei)
    3. Check PoissonD generator for discreteness and will examine quantization part of MCSMEAR. (Andrei)
  3. Mar 28 Meeting
    1. David will send the code where the random generator is called to Andrei.

Agenda

  1. Updates
    1. Will:
      1. Changes to BCAL Reconstruction
      2. BCAL_Reconstruction_Issues
    2. Beni: BCAL reconstruction test with pi0
  2. Any other business

Minutes

Attendees: David, Beni (JLab), Will Levine (CMU), Andrei, Zisis (Regina)

  1. Action Items
    1. David verified that we have 12 bits in the FADC. 12 bits corresponds to 2 V pulses, and the effective bits are something like 12.5 or 13.5. Another example: 8192 does not mean 14 bits, since samples are correlated.
    2. David sent the code (~1000 lines) to Andrei. Andrei looked and did not see any problems in it. We should revisit the application of random number generators in tandem and see if they are producing spikes in the spectra.
  2. Updates
    1. Will: Checked in code to do timewalk corrections in BCAL reconstruction. This is necessary to do reconstruction with David's new, more realistic BCAL simulation in HDGeant/mcsmear. In yesterday's Software Group meeting it was decided to switch over to the new simulation scheme by default. Discussion of KLOE vs. IU clustering algorithm continues, need more data.
    2. Will's slides (Changes to BCAL Reconstruction) generated much discussion related to how clusters should be formed, how events from the outer cells (no TDCs) be treated, likewise for events that may not have a TDC hit in the inner layers, and the KLOE vs IU behaviour (most important difference is cuts in z vs cuts in theta). It became clear that the current HDGEANT has the (unrealistic) energy weighted time and we should be careful about drawing any conclusions from those simulations. Other discussion revolved around the order of doing things (e.g. match ADC-TDC info on one side and do time walk, or combine both sides in doing the time walk, a.o.)
    3. David's proposal:
      • Clusters are formed using timing info from the fADC only
      • Timewalk corrections to TDC hits are applied at the individual hit level (i.e. single ended)
      • Final timing calibration for TDCs is done via Andrei's recipe
      • After clusters are formed, hi-res TDC info is used to define properties of cluster as a whole (z-position and time of impact)
    4. Will has not used Andrei's algorithm yet and questions arose later (offline) about how applicable it would be to real data. We should indeed strive for producing code that would work on both real and MC data.
    5. We need to consider whether to process based on Delta-z (mm) or Delta-time (ps).
    6. Other topics: energy-weighted position (hits), shapes of energy depositions in cells, and other effects reported in literature. Risetime, jitter and dynamic range (E and theta) complicate matters.
    7. KLOE vs IU: A year ago Will looked at a Pythia pi^0 sample and kinematics such as 1GeV at 50deg are not present. These kinematics used by Beni might bias results. Will will send realistic kinematics to Beni so that he can repeat his study with those. Also, KLOE is less aggressive than the IU code in separating clusters in the same phi sector and uses a fixed cut expressed in cm for z. The latter is not a linear effect from mid to forward angles. IU uses a tight theta cut, and this is also responsible for accentuating the difference between the KLOE and IU simulations in Beni's kinematics. Beni will post a flowchart of his code.
    8. Will has gathered many of these issues in a list: BCAL_Reconstruction_Issues. We will mull over this and David's proposal above.