Star-fst-l mailing list

[Daniel Brandenburg <dbrandenburg.ufl AT gmail.com>]

Hi All,

Flemming has spent quite some time helping me to pinpoint the problem with the FST geometry and to solve the extreme slowdown that the full detail geometry caused in GENFIT.

Flemming found the root issue and we have demonstrated that the full detail FST model can now be used with GENFIT (no slowdown) with only a few small changes (wrt the version in CVS now) to the GDML definition. We are working with Jason to make sure the updated geometry is put into CVS asap so that it becomes the default for "dev2021".

I will still need some time to study the performance (momentum res and charge mis-id) but the crucial progress is that the new geometry runs at the expected speed (<100ms / track) with essentially all track fits converging cleanly.

Here is a summary of the changes Flemming made and some notes:

-- The cvs version that Te-Chuan had made had some issues that I fixed.

This primarily relates to the support ring and has unlikely had much
effect on the tracking as it was mostly outside the 2.4 to 4 eta range..
-- After we included the assembly=true for the FSTW volume GENFIT ran fine in terms of time/track
-- I do not believe there are any overlaps and should have worked
-- I believe this model was generated from an older solid works model of
the FST disks, but differences to the latest one is small. As an example
the Peek supports seems to be roughly 1/2 of the latest model.

Best,

Daniel