FCAL threshold simulation in mcsmear

[igjaegle]

The threshold depends of the data gains which for a typical GlueX run (e.g. 30496) varies between 0.5 and 10 i.e. roughly between 12 and 250 MeV. A more appropriate solution would be to set the same threshold for all FCAL channels and this value should be the highest (realistic) data energy threshold + 20 to 40% higher (for security margin) of a run range. It also means in data the same software threshold should be used for a detector to be included in a cluster.
@sdobbs @rjones30 @aaust @andrsmit @somovalex

If I am not mistaken this solution is already applied in BCAL. @markdalton could you confirm that and explain how BCAL_ADC_THRESHOLD_MEV was determined?

[sdobbs]

My 2 pfennig: If you're proposing to change the reconstruction of the data in the future, it would be helpful to show that such a change does not adversely affect the energy resolution of the reconstructed showers.

Side questions: For the channels with the gain factors ~10, are these in the standard GlueX fiducial region? Were they properly functioning during this run period?

[mashephe]

This simulation was checked and tuned on data. It was verified that it produce the desired result. While range between 0.5 and 10 may be true this isn't a fair representation of the typical variation. If we are going to start changing things like this, and also the new non-linear approach that is proposed, we really need to verify we are making improvements. So much effort goes into testing and validating the algorithms we can't just keep starting from scratch. The mechanism for applying the threshold in data is fundamentally different from MC. And I don't see the value in actually throwing away information in data by applying a higher threshold, when we can attempt to simulate the data threshold. True, our simulation may have problems for pathological channels with gains of 10 -- there aren't many in the FCAL -- but we should fix the the gains on those channels rather than throw the simulation algorithm. On May 12, 2020, at 11:04 PM, igjaegle <[email protected]<mailto:[email protected]>> wrote: The threshold depends of the data gains which for a typical GlueX run (e.g. 30496) varies between 0.5 and 10 i.e. roughly between 12 and 250 MeV. A more appropriate solution would be to set the same threshold for all FCAL channels and this value should be the highest (realistic) data energy threshold + 20 to 40% higher (for security margin) of a run range. It also means in data the same software threshold should be used for a detector to be included in a cluster. @sdobbs<https://github.com/sdobbs> @rjones30<https://github.com/rjones30> @aaust<https://github.com/aaust> @andrsmit<https://github.com/andrsmit> @somovalex<https://github.com/somovalex> — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub<#126>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ADG5L2FADOMMYSCBJAVUOUDRRIE2NANCNFSM4M7LOZ7Q>.

[igjaegle]

Matt, could you share the notes and/or publications describing these studies?

[mashephe]

See an email to you from April 24 quote below -- the shapes on the right side of slide 9 are driven largely by how the block level thresholds are simulated. Your proposal would introduce a hard cutoff in this distribution, which is easy to simulate, but I really don't like the idea of throwing away perfectly good data. Matt As I noted to Igal, this was studied and tuned in G4. It is particularly important to get low-energy response correct when writing the algorithm to reject hadron interactions. See slide 9 of this: https://halldweb.jlab.org/DocDB/0040/004084/001/MC_NeuralNet_Updates.pdf Also the checking of energy response in G4 is illustrated on slide 14 of that same talk. You can see the reconstruction of the photon gun works just fine. This suggests a discrepancy between Igal's plots and these photon gun results, perhaps somehow related to the use of a pi0? ...or somethin On May 13, 2020, at 9:21 AM, igjaegle <[email protected]<mailto:[email protected]>> wrote: Matt, could you share the notes and/or publications describing these studies? — You are receiving this because you commented. Reply to this email directly, view it on GitHub<#126 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ADG5L2BM3URKLYIGJSHBLJDRRKNEFANCNFSM4M7LOZ7Q>.

[igjaegle]

Slides, in general, do not allow reproducibility. When the note will be ready? Could your student/postdoc study for the slide 14 results,
1/ if there is any run-dependence? The gains ie the threshold are changed from time-to-time which is particularly important for the low energy part of the curve.
2/ if there is any polar-angle dependence
3/ Decompose the origins of the non-linearity? Shower leakage/threshold/transparency
Also, I am used to that all scripts/macro/simulation concerning calibration and/or simulation tuning are shared and described to allow reproducibility by someone else. Might all the relevant materials be shared?