Draft: Fix MM pileup overlay time distribution

Closed Alexandre Laurier requested to merge alaurier/athena:MM_DigitizationWindow into master

A MM digitization parameter setting a higher time limit in the digitization was incorrectly set too small and caused significant problems in the MM pileup and overlay time distribution of the strips.

This upper time limit would assign strip times to 400ns for all strips with t>400ns causing a massive peak in the time distribution and would allow these out-of-time strips to be used in the reconstruction. Not yet understood, but thoroughly tested, is the behaviour that this upper time limit would also create a population of strips with an erroneous time of zero.

Increasing the higher time limit by 300ns addresses all these issues and now properly returns the expected strip time distribution from MM presampling and overlay jobs. Attached are two figures comparing the problematic time distribution (peaked at 0ns and 400ns) and the time distribution post-fix.

mm_pileup_time mm_pileup_time

The issue of large memory usage in the reconstruction of overlayed samples is still being investigated but this MR will fix a crucial contributor.

However, for an unknown reason, increasing the upper time window also creates FPE warnings in some MM L1 trigger jobs. I will add comments and job commands in the comments below. I have tracked down where the FPE error stems from and I can offer some possible causes. The MM digi and trigger sim experts can take a better informed look.

@chchau @jojungge @rosati @pscholer @jchapman

Merge request reports