Meeting solid ec

From solidwiki
Jump to: navigation, search

ECal weekly July 2025

  • Weekly ECal July 25, 2025 (Friday)
    • Mohhamed and Darren looked more into ML PID pion rejection with momentum cuts. Findings:
      • Darren repeated the classical PID work (previously done by UVA student Spencer), found that the pion rejection performance drops as the momentum increases, for classic PID as well, see Classical_vs_ML.png Our previous confusion was that the pion rejection "without momentum cut" would go up as we increase the "slope cut". It's possible that some of the higher momentum pions produce secondary particles and start to become similar to electron signals.
      • Mohhamed's update: ML_20250725_Rafi.pdf. First, here is a short summary of ML PID study:
        • So far, everything is using simulated data, with electrons from eAll and pions from bggen. (Here the particle type is as produced at the target.)
        • For training, the cuts applied include TS2 trigger cut, LASPD>0, ShowerSum>0.
        • For testing, momentum bin was applied. Also, the same LASPD and SC-C cuts as classical PID were applied to allow a more equal comparison.
        • The Cherenkov Sum cut used in the classical PID pion rejection was not applied to the testing sample, as we can't study electron efficiency with this cut (and ML method does not allow different cuts for electron vs. pion performance). On the other hand, we could apply CerSum cut to the testing results for the pion performance alone after the performance has been extracted. The only concern now is statistics.
      • To do items (Mohhamed):
        • Now we need to reconcile the classical pion rejection without vs. with momentum cuts. Hope that's straightforward to do.
        • We need to use all available simulated data for higher statistics.
        • We need to add "no background" ML performance to the table.
        • Producing confusion matrices with the momentum bins would be nice.
        • Still need to apply model to beamtest data and produce plots.
    • Simulation: the latest status is Lorenzo needs to set up the new magnet and geometry in his Fluka code for the Preshower radiation dose study. We believe Jianping will talk to him about setting aside some time for this work.
    • Bench testing: Michael L. - everything is set up in Carl's lab. Observed that high rate (~2 MHz) discrete pulse produces similar gain shift to constant background. Observed 2x gain shift with 60uA anode current. Talked about checking current is the same in ammeter and O-scope, setting HV to 860V instead of 875V, characterizing gain shift with discrete pulses. See slides: Benchtest_7_25_25.pdf. Additional notes below:
      • The quoted anode current on slide 2 was estimated using the baseline ship on the oscilloscope, which may not be precise enough. Mike mentioned it may be possible to store the spectra digitally. In fact, a 3mV shift on the scope should correspond to a 60uA anode current, which is where the gain just starts to shift according to Ben's plot. The observed "signal" pulse increased from 25 to 35mV (slide 2) indicates that we might be already at 100uA anode current or 5mV baseline shift.
      • For a 2MHz pulsed LED (with about 30ns width and 25mV height), the equivalent estimated baseline shift is 1.5mV or 30uA anode current. If the signal height starts to become larger, it means a high-rate pulsed signal (such as charged pions for SoLID) would have similar effect as a constant, low-amplitude background (such as the beam-on-target background for the beamtest). It means that the radiation background for the SoLID case (dominated by charged pions) will age the PMT and cause its gain to shift if the base is not designed properly. We will need to measure the PMT's response to pulsed LED precisely.
      • Michael also managed to "reproduce" the gain shift by making the 10th dynode of the PMT circuit much larger gain than dynodes 1-9. Will continue working on the code and apply the same "model" to the THU base.
      • Note that to extrapolate the LED amplitude (or expected anode current assuming perfectly linear) vs. potentiameter value (last slide), one can simply use: I = a*V/(R_LED + R_pot), with V the power supply voltage, R_LED a fitting parameter, and "a" also a fitting param that related the light intensity of the LED to its current.
    • Xiaochao showed calculation of "FASPD2" (a thicker, 2nd FASPD aiming for timing resolution), see:
      • Faspd2_dx0_tdc0_bc408_timing_mc.png. The "design" of this FASPD2 is long, single-piece scintillator bar from inner to outer radii of the SIDIS FA geometry, and with 240 segments in the azimuthal direction. The readout is light-guide directly coupling to a PMT (could be MCP-PMT) on the outer edge, no WLS fiber is used. The calculation was done with a perfect TDC (0ns resolution), 0ps PMT TTS, and 0cm position uncertainty. It shows the timing resolution when the particle is hitting on the edge furthest from the PMT readout side is already 100pm, purely due to the range in photon travel path within the scintillator bar.
      • As a comparison, the same calculation for the LASPD (60cm long) can be found at Laspd_dx0_tdc0_bc408_timing_mc.png. With a 60ns TDC and a fast PMT, the observed best performance of the LASPD was 150 ps. This means realistic factors would increase the timing solution by a non-negligible amount, and so if our calculated best resolution is 100ps for a 2nd FASPD2, we may not get 100ps in reality at all.
  • Weekly ECal July 17, 2025
  • Weekly ECal July 3, 2025
    • Mohhamed presented ML PID progress: still need to add smearing to match simulated spectra to data; need to separate charged pions from pi0's.
    • Ye contacted Lorenzo about preshower radiation dose from Fluka. Lorenzo need to run the code to get this information
    • Bench testing Shower PMT: Michael L. showed a few pictures of the laser and LED setup

ECal weekly June 2025

  • Weekly ECal June 27, 2025 (Friday)
    • Ye showed simulated radiation dose for 82 deg, it's dominated by BeamOnTarget "total including hadrons", which is ~10 times higher than pi+/- and pi0.
    • ML PID: we are still confused by how to mix background with signal -> after off-line discussions, we (MR, DUP, XZ) realized the background mixing meaning adding the signal in real time. So for 40-ns timing window, the ratio of bg/signal is simply 40ns times the background total rate. We got the total integrated rate from Ye, see... (slides?)
    • Michael L. discussed bench testing progress.
  • Weekly ECal June 5, 2025
    • Sort of like the "summer kickoff" meeting. We will have Mohhamed working on ML PID and Michael L. working on PMT bench-testing.
    • To follow up on questions from the last meeting, the dosimeter reading at GEM0 frame (front-most) was in the 70's to 100's of krad.

ECal weekly May 2025

  • Weekly ECal May 22, 2025
    • Xiaochao showed a summary of known radiation hardness information of scintillators and fibers, see slides: Summary_radhard_20250522.ppt or Summary_radhard_20250522.pdf
    • PMT bench testing: setting up FADC would take time, we can start with just QDC (and scope)
    • It was brought up that Zhongling at JLab is setting up FADC/CODA in ESB for DAQ test, to support LAPPD test of Sanghwa.
    • Need to email Carl Zone about bench testing
    • What was the total radiation of the beam test?
  • Weekly ECal May 8, 2025
    • Ye's slides on comparison of preCDR vs. latest simulation. Note that the preCDR plot (on 2D plane energy vs. R, reproduced by Ye) is for e.g. "pi- from pi-" while the latest simulation is "everything from pi-". Even though the plots are for different things, it seems to contradict the table that the latest simulation gives much higher radiation background.
    • Zhiwen will dig into how Jin extracted radiation dose from the particle rate plots.
    • We discussed the labor cost of large-angle ECal and SPD assembly and testing. Here is Richard's estimate, consistent with 2023 estimate of 744 PW scaled by area of 440/1800.
  • Weekly ECal May 1, 2025
    • Ye worked out a way to extract the z-dependence of Edep in the Preshower and Shower. Counting the whole preshower as a layer, the radiation is pretty high. On the other hand, the radiation in the Shower layers does not look much worse than preCDR, though the exact z-dependence is different. See her presentation: PVDIS_simulation_configuration_comparison_04302025.pdf. Note that the radiation dose of the preshower needs to be integrated over the preshower's Z (slides 2 left, 8, and 9 left).
    • Jianping suggested re-visting low-energy pion comparison of bggen vs. wiser, i.e. DVCS and MARATHON data analyzed by Ye Tian (IMP) and Zhihong. -- followup on 5/8: Ye dig out the old slide but that study stopped at 2GeV for the pion momentum (minimum).
    • Mike and Richard talked to Carl about benchtesting the Shower PMTs, to be followed by MAPMT and MCP-PMT tests.
    • Darren gave a presentation SoLID ECal ML PID on the latest status (from Nov. 2024) of AI/ML PID study. Recap:
      • the study was done using simulation without background, then applied to data
      • e vs. pi0: ML output could reject pi0's (since probability of pi0 is always < 0.5)
      • applying ML to data, we see pi+, pi- and e ID in the output
      • ML-ID'ed e- and pi+ or pi- preshower, ShSum from data agree with simulation very well
      • ECal resolution: had to add by hand about 35% for PSH, 10% for SH
      • one result is ML PID charged pion rejection vs. e- efficiency -- something comparable to classical PID
      • the only thing missing was background simulation!

ECal weekly April 2025

  • Weekly ECal April 10, 2025
    • LHCb tracker upgrade, Table 3.4. (Note 1Gy = 100rad)

ECal weekly March 2025

  • Weekly ECal March 27, 2025
    • Ye showed comparison of Jin's preCDR and latest study. Checking particle entering ECal. Large difference already between Jin's and "THis study" using short endcap and Babar baffle. This means the difference is in the generator (or the interpretation), not the endcap and magnet.

ECal weekly February 20, 2025

  • Mike's PID update. Some discussions:
    • JP: what are the uncertainties in the preshower and Shower calibration? -> Ye and Mike need to find the uncertainty. Ben's plot may be more precise
    • Ye: there is also a 65uA (4409) run
    • JP: there should also be a Preshower cut
  • Question arose how the preshower radiation dose is 3 times higher now (SOLID PVDIS) vs. preCDR value...!
    • JP: there was a study of Fluka va. GEANT by Lorenzo Zana before the Science Review, need to dig this out
    • We need to vet the statements of *3 and *5 estimate (how much SoLID PVDIS is higher than the beam test in terms of radiation dose in the ECal modules, currently in the report)
  • Comments on the report draft:
    • FADC timing should not be in the Shower section

ECal weekly March 2024 to December 2024

  • Weekly ECal October 31, 2024
    • Ben started analyzing the THU base. Here is his Calculation on all three PMT bases
    • Zhihong shared pictures of the super-module being assembled at THU: pic0, pic1, pic2; He will also send a new PMT base design diagram (for the super module) so we can study its possible gain shift in advance
    • Zhihong will look if there are old PMTs with the same base as the 2023 Hall C beamtest, and will do a bench test to study the gain shift if any such PMT/base is found.
    • Xiaochao shared a rough draft of the beam test report. Still need to wrap up some calculation, and need someone to double check all numbers.

ECal weekly March 2023 to February 2024

FY2022 beam test summary

ECal weekly January 2023 to March 2023

ECal weekly January 2013 to October 2022

Retrieved from "https://solid.jlab.org/wiki/index.php?title=Meeting_solid_ec&oldid=2128"