Difference between revisions of "SoLID Ecal Weekly 20230727"

Revision as of 17:02, 28 July 2023

SPD Timing discussion

• Carter showed his slides from previous week. SC_C - SC_D seems reasonable. See comparison table at SPD page, 2nd table from top. Note that HRS S2 was 50mm thick and had ~300ps resolution. Scale by 1/sqrt(Npe) or 1/sqrt(thickness) we expect 474ps for the 2-cm thick scintillators we are using. (This also means our PMT might have better "Transient Time Spread(TTS)" than the S2 PMTs.
• To do for Carter:
  • For LASPD timing, should plot (T+B)/2 minus another scintillator.
  • Jianping mentioned a formula for timing resolution vs. Npe, we can look for it.
  • study the SPD page (above).
• Mike will post PMT information soon

Slides

Rate data/sim discussion

• From previous week, it was obvious that MIP for Shower shifted to higher values for high rate runs.
  • Paul commented raw FADC spectra did not show obvious baseline drift. Maybe the drift is subtle and can't be easily identified by eye? Pulse deconvolution might help
  • Mike will check with Jixie on the code and technical note availability. The note should contains the 2 simpler options (fixed vs. dynamic pedestal subtraction, no deconvolution). Perhaps co-writing the note?
• Ye showed 10uA run comparison throughout the 18-deg run period and the MIPs remained stable
• Jixie's pedestal study (on redmine) is without beam. Checking the run plan, it looks like on Jan.12th we talked about checking the pedestal shifts with beam on, but the run list doesn't show any run for this purpose. Thinking about it, checking the pedestal (baseline) shift with beam on would require careful analysis, and what we are doing now is (perhaps?) exactly what was planned to do.
  • Can wait for pulse-deconvolution to be ready to extract pulse height.
  • more raw FADC pulse is always useful.
  • To be discussed next time: (1) How do we extract baseline shift from data? (2) Can the same code output baseline shift??
• Ye showed SC-B pulses and it had significant higher rates and baseline shift than front scintillators. (SC-B had no beam-side shielding.) THis is okay if we want to cut on half-MIP, just need to shift the MIP position accordingly.
• For longer term we need to simulate pulse shape in SoLID full sim. This will give us: baseline shift, timing, etc. Maybe Ye will do this.

slides from Ye

PID classical method

• Spencer's slide using beamtest simulation, no beam-on-target (BOT) background, no momentum slice, with TS3 cut. Comments and to do:
  • Psh vs. Sh 2D cut slope was determined to be -0.8 (by using momentum-binned sim spectra)
  • e-eff starts at 95%. This is due to low energy electrons mixed in with pions (i.e. expected).
  • try mixing in background to "beam test full sim" at bg*1 and bg*10, to provide comparison points with ML method
  • double check that the trigger cuts (TS3) are the same as Darren's;
  • use either "beam test sim" or "beam test sim plus bg", check if classical method is multiplicative. That is, the e-eff for using both ECal and Cher is the e-eff's of ECal multiply that of Cher; and similarly for pion rejection.
  • Note: there were some discussion about momentum cuts/slices but I think with the TS cuts, that may not be necessary given the short time left.
  • Write the report but keep it concise -- the important part is to document all cuts used and the results (can I see your overleaf?).
• Combined with Darren's update below, we also need pi0 study with the classical method. To do list for Ye:
  • dust off the pi0 study, make sure cuts are the same as ML method so we can have a direct comparison
  • apply pi0 rejection to TS1 and TS3, both sim and data, and see how it goes -- this is to provide a classical comparison point to ML method.

PID ML method by Darren

• Darren's slides: Discussions and comments:
  • Slides showed bg*1, *10, and *100. Zhiwen + Ye said the bg for actual situation is simply about *1. We should focus on *1 and *10 in this case; Need more bg events.
  • TS3 results (using SC-A and SC-D at half MIP and ShowerSum.gt.1) gives very good 2x2 (electron vs. charged pion) confusion matrix.
  • Can we see ECal alone and Cher alone curves, just to confirm the point that ML method is not multiplicative?

• To do for Darren: run the "beam test + bg" with more stats, focus on bg*1 and bg*10, for all of TS3 (3x3), TS3 (2x2), and TS1 (2x2) [BTW after you left, JP commented that pi0 are very important]