Difference between revisions of "BCAL Readout Segmentation Task group Meeting: Jun 17, 2011"

From GlueXWiki
Jump to: navigation, search
(New Agenda items)
(New Agenda items)
 
Line 30: Line 30:
 
== New Agenda items==
 
== New Agenda items==
 
* [http://www.jlab.org/Hall-D/software/wiki/images/9/94/Calibration_uofr_segm_meet110617.pdf BCal Calibration] (I.Semenova, A.Semenov)
 
* [http://www.jlab.org/Hall-D/software/wiki/images/9/94/Calibration_uofr_segm_meet110617.pdf BCal Calibration] (I.Semenova, A.Semenov)
* [http://www.jlab.org/~semenov/time_simulation_110617.pdf Simulation of time and amplitude from the shower] (A.Semenov)
+
* [http://www.jlab.org/Hall-D/software/wiki/images/b/b4/Time_simulation_uofr_meeting110617.pdf Simulation of time and amplitude from the shower] (A.Semenov)
 
* [[Media:20110617_bcal.pdf|BCAL Energy Resolution Contributors]] (David L.)
 
* [[Media:20110617_bcal.pdf|BCAL Energy Resolution Contributors]] (David L.)
  

Latest revision as of 17:44, 17 August 2011

Meeting Info

Time:

  • 11:30 JLab
  • 9:30 Regina

Location:

  • CC F326-327

Vid-con:

  • ESNet 8542553

Agenda

Review action items from previous meeting

  • Send e-mail to Andrei and Irina pointing them to the BCAL smearing code where the new Poisson sampling was introduced (David)
    Completed
  • Measure inherent energy resolution of hdgeant/mcsmear chain without sampling fluctuations turned on. (David)
    See today's slides
  • Write description of how we should incorporate sampling fluctuations in the smearing code (Elton)
    ?
  • Write description of how we should incorporate sampling fluctuations in the smearing code (Andrei)
    Briefly outlined in June 10th email:
    • "About the implementation of realistic energy resolution in readout cells (because of the sampling fraction): I do believe that we can use the Fluka simulation of the BCal modules with detail structure that Irina did for 90 and 20 degree emission of photons in the range up to 2 GeV. We can compare the relative width of Fluka spectra from the cells with the relative width of the correspondent HDGEANT spectra (energy depositions in the homogeneous material only, all extra-widenings are off); the difference will give us the correction we need to introduce into the simulation as a function of cell, energy, angle etc."
  • Check and document input distributions to profile histograms to make sure tails and second peaks don't significantly affect the mean. (David)
    Not necessarily completed. Plots circulated on June 11th.
  • Follow up on status of GlueX code documentation. (David)
    Work has started on updating/enhancing documentation as reported at the June 15, 2011 offline software meeting.

New Agenda items

Minutes

Participants

Andrei S., Irina S., Matt S., David L., Elton S., Beni Z. (partial), Eugene C. (partial)

Review action items from previous meeting

We briefly reviewed the action items from last week's meeting. Some of them were to be addressed by specific talks at this meeting so see notes specific to those presentations below. Some noted points:

  1. Andrei had sent a brief description of a procedure for calculating the dependence of the sampling fraction
    as a function of energy, angle, and cell using FLUKA. Elton had not had a chance to look at and consider
    this proposal so will do so later and send comments (if any).
    Andrei suggested holding off on doing the MC calculation until the more detailed timing structure is implemented
    in hdgeant/hddm.
  2. It is believed that the double peak structure observed in the reconstructed energy distributions by Irina was
    an artifact of the IU algorithm that is currently under development. Therefore, the item marked
    Not necessarily completed can be considered completed.

BCAL Calibration

Slides 2 & 3

Andrei went over some slides describing a calibration procedure he and Irina implemented. This involved fitting the E_generated to Eraw ratio to a two-sided Gaussian (i.e. different sigma's on either side of the peak) + const level on the right side. The histograms covered a fairly narrow range in energy (50MeV) and z(5cm). For all cases, only events with exactly 1 reconstructed shower were used. The simulation generated 50k single γ events/cm with a flat energy distribution in the 0-4GeV range.

A cut of 1.5σ's was then placed on the ratio distributions where σ was different on either side of the peak. The mean of the histogram in the range (peak-1.5σ1) to (peak+1.5σ2) was taken as the value of Egen/Eraw at the center of the bin in Eraw and Zentry.

It was noted that the course segmentation case resulted in far fewer events than the fine making it to the histogram. Andrei speculated this was due to the cut on the number of showers. He agreed to look at the number of showers per event for the fine and course data samples.


Slide 4

The mean Egen/Eraw ratio values were plotted as a function of Eraw for each bin in z (n.b. Andrei noted that the x-axes were mislabeled as E_rec when they should have been E_raw). These were fit to a function described on slide 2. The function was similar to what had been used previously, but used the ratio of Egen to Eraw as a function of Eraw rather than the other way around.

Andrei noted that the dip from 3.5-4 GeV was an end-point effect and was observed to move when the end point was moved from 4 GeV to 6 GeV.


Slide 5

The parameters of each of the fits like those shown on slide 4 were plot as a function of z. These were fit to functions with 13 free parameters, 9 of which defined Gaussians located at positions corresponding to FDC frames. For these fits, the errors used on each bin came from the reported errors on the first stage fits of Egen/Eraw, but scaled by the √χ2/NDF.

Scripts were written to produce a dataset and fit it using the resulting parameters. These will be sent to Dave so they may be incorporated in the repository to use as a common calibration mechanism for the current studies.

Andrei noted the calibration is very sensitive to the simulated data set, segmentation, smearing, injected noise, used thresholds as well as cluster reconstruction procedure. Any changes to either of these should trigger a recalibration.

Simulation of time and amplitude from the shower

Andrei went over a procedure that has been used in the past to implement the time development of a shower signal at a photodetector. He supplied example code with how this may be implemented in Fortran in the existing hdgeant program. This essentially involved propagating each energy "chunk" to the photosensors individually, building a histogram of the attenuated, deposited energy as a function of time.

Dave noted that dark pulses will need to be added before a threshold can be applied. This means either moving all of the dark pulse generation code out of mcsmear and into hdgeant, or emitting the histogram info from hdgeant so that it can be used in mcsmear.

Matt noted that there is a mechanism in the KLOE code to split clusters based on the RMS of the times of the cells making up the cluster. This has not been optimized for GlueX and changes to the timing mechanism in the sim-recon package may require changes to this parameter (BREAK_THRESH_TRMS).

Dave volunteered to implement this and will report on it at the next meeting.

BCAL Energy Resolution Contributors

Dave showed several slides reporting on a study of the contribution to the resolution of each of the effects that have been introduced in the BCAL smearing to add realism. The study involved 90o γ's incident on the center of the BCAL, mimicking the 2006 beam test.

Questions arose concerned that the 1/E term appeared to remain 0, even when effects (e.g. electronic noise) were included that should have contributed to that term. Dave agreed to check this once more.

The final recommendation of the talk was to remove the explicit smearing due to the floor term since that is already being included via energy leakage in hdgeant. There was no disagreement.


Action Items

  • Review Andrei's proposed procedure for calculating sampling fraction functions (Elton)
  • Plot Nshower per event for fine and course data samples (Andrei)
  • Send calibration scripts with instructions to David (Andrei)
  • Implement time distribution of signals at photodetectors (David)
  • Verify noise term is zero in resolution fits (David)
  • Investigate effect of TRMS threshold used to split clusters in KLOE code (David)