search

eic / epic

DD4hep Geometry Description of the ePIC Experiment
https://eic.github.io/epic
GNU Lesser General Public License v3.0
24 stars 43 forks source link

missing MPGD hits #774

Open ShujieL opened 2 weeks ago

ShujieL commented 2 weeks ago

Environment: (where does this bug occur, have you tried other environments)

I generated some 10 GeV pi+ events at theta between 70 to 110 degrees with epic 24.08.1, and noticed that more than 10% events has no hits on inner MPGD.

I also checked a simulation campaign file and saw similar behaviors EPIC/RECO/24.07.0/epic_craterlake/SINGLE/pi+/10GeV/45to135deg/pi+_10GeV_45to135deg.0000.eicrecon.tree.edm4eic.root'

Steps to reproduce: (give a step by step account of how to trigger the bug)

  1. open one of the single event sample as mentioned about, and extract the branch "MPGDBarrelRawHits" with uproot
  2. check the number of events with at least one MPGDBarrel hits, e.g. len(br.entry.unique()), and compare the returned number to the total mc event number.

Expected Result: (what do you expect when you execute the steps above)

At 10 GeV and central angle, almost every single pi+ particle should trigger inner MPGD.

Actual Result: (what do you get when you execute the steps above)

Less than 90% events have inner MPGD hits. This number is even lower for outer MPGD

wdconinc commented 2 weeks ago

@ShujieL Can you post some more diagnostics here since you already started looking into it?

ShujieL commented 2 weeks ago

@wdconinc, With my 20k 10GeV pi+ sample at 70<theta<110 degrees, I got the following number of events with at least one hit at given radius range. The events missed MPGD seem to have no preference in theta and phi.

Hit collection name / rmin / rmax [mm] / # of events SiBarrelVertexRecHits 30 42 20000 SiBarrelVertexRecHits 46 60 20000 SiBarrelVertexRecHits 115 130 20000 SiBarrelTrackerRecHits 265 280 20000 SiBarrelTrackerRecHits 415 450 20000 MPGDBarrelRecHits 540 600 16353 TOFBarrelHits 610 650 20000 OuterMPGDBarrelRecHits 700 750 13875

wdconinc commented 2 weeks ago

Can you rerun one specific event with the track reconstruction in debug log level and analyze whether the hit even gets turned into a measurement? We know that for slanted tracks through the 3mm thick MPGD layers there is a non-negligible chance that there is a geant4 step that ends in the middle of the gas volume. That leads to two tracker hits then and each is going to be a bit displaced from the actual location of the center plane crossing. Can you rule that out based on debug output in the tracker to measurement converter?

ShujieL commented 2 weeks ago

Can you rerun one specific event with the track reconstruction in debug log level and analyze whether the hit even gets turned into a measurement? We know that for slanted tracks through the 3mm thick MPGD layers there is a non-negligible chance that there is a geant4 step that ends in the middle of the gas volume. That leads to two tracker hits then and each is going to be a bit displaced from the actual location of the center plane crossing. Can you rule that out based on debug output in the tracker to measurement converter?

those numbers are from RecHits which are converted from sim hits after digitization, before the tracker measurement conversion. Also, MPGDBarrelRawHits and MPGDBarrelRecHits have the same number of entries. So the hits are very likely missing in sim hits (geant4).

wdconinc commented 2 weeks ago

Interesting. Please post more as you find out more.

ShujieL commented 2 weeks ago

Interesting. Please post more as you find out more.

I will check the impact of missing MPGD hits in tracking. Meanwhile, I'd suggest

  1. @mposik1983 can probably check if there are acceptance gaps on MPGD with track projection.
  2. @rahmans1 @Chao1009 does this have anything to do with some threshold in Geant4 ?
mposik1983 commented 2 weeks ago

@ShujieL Thanks for already starting to look into this. I can look into the acceptance a bit more and I will also bring this up during today's MPGD meeting. One thing to keep in mind is that we know there are acceptance gaps in the outer MPGD barrel by design (follows DIRC bar geometry), so having less Sim hits in the outer MPGD barrel compared to the inner MPGD barrel should be expected.

ShujieL commented 1 week ago

It seems that the hits are missing from the sim hits to raw hits transition, i.e. digitization. See https://indico.bnl.gov/event/24735/contributions/96326/attachments/57062/97925/missing%20hits.pdf

wdconinc commented 19 hours ago

Because the SiliconTrackerDigi is used for the MPGDs, the only reasons why sim hits may not turn into sim hits are:

wdconinc commented 17 hours ago

With a threshold of 0.25 keV, that is the main reason for cutting MPGD hits here. Below is a plot of the energy deposition for simulated hits, in x-axis units of GeV, so keV if you ignore the 10^-6. A cut of 0.25 keV cuts a sizeable fraction of these events. @mposik1983 Are these thresholds correct?

image

mposik1983 commented 15 hours ago

@wdconinc for now the threshold values should stay as they are. From the discussion at the tracking meeting, there should be some follow up next week about a new threshold estimate

wdconinc commented 15 hours ago

Yeah, I didn't know you were going to present on this exact issue today :-)