Missing Space Points from Cluster3D

[bear-is-asleep]

I'm submitting a ...

• bug report

Space points are produced for SBND in SPINE using cluster3d. This issue shows how we're missing space points at this stage.

Checklist

• I've added at least 1 label under Labels
• Assigned a reviewer under Assignees

Expected Behavior

Expect all space points to be made out of cluster3D.

Current Behavior

We see that MIP tracks have poor completeness. We study this for two versions of code to observe the sudden drop. The plot shows the completeness for these two versions.

Visual example of missing space points in the middle of a track below.

Possible Solution (optional)

• Including bad channels into cluster3d in sbncode PR #490
• ...

How to reproduce

• Run a job up to reco1 (we used MPVMPR but anything with MIP muons will do) - how to do this
• Parse larcv.root file through SPINE, which will store the completeness per particle

ana:
  overwrite: true
  track_completeness:
    run_mode: truth
    time_window: [0, 2000]

[SFBayLaser]

One potential issue was that bad channels were not being accounted for. The space point building tool is able to check for bad channels though the original version was written in the MicroBooNE era and expected different information returned from the channel status service. This has been updated and now the bad channel code works with the "simple" channel status service. However, this does not really fix the problem reported above, but at least we have that implemented to take it off the table.