acts: bug: interpolation out of bounds with ATLAS B-field map

Using the propagator with the ATLAS B-field map can give error messages:

TrackFinding ERROR Step failed with MagneticFieldError:1: Interpolation out of bounds was requested

This can give a lot of messages: ~259 MagneticFieldErrors per ttbar+μ=200 event, each of which produces 2 errors, and 4 warnings (1.6 GB logfile for 1000 events)!

This error is due to InterpolatedBFieldMap::getField() called for a position outside the mapped ATLAS B-field

  • The ATLAS B-field is mapped in (-10000,-10000,-15000) to (10000,10000,15000)
  • but I see, calls for e.g. getField({-2849, -2486, 54450})
    • These are presumably for extension of seeds that will eventually be cut or rejected

#1467 removes these error messages by trying to return the edge-bin values instead of giving an error. That is still WiP because I couldn’t yet get it to pass the unit tests: seems to work with B=(0,0,0) at the edges (as for ATLAS), but not otherwise.

Even if that can be resolved, the question remains: why is getField() called for points so far outside the detector? @Corentin-Allaire suggested (ACTS-ITk meeting) that the world volume is too large. If so, it may be an ATLAS-specific error, but I hope it’s OK to discuss here.

About this issue

  • Original URL
  • State: closed
  • Created 2 years ago
  • Comments: 22 (20 by maintainers)

Commits related to this issue

Most upvoted comments

Looked into this today. The underlying bug looks quite significant to me and makes me wonder how the CKF produced any results at all…

Will create a fix and link it here.

https://github.com/acts-project/acts/pull/2299

I have removed the workaround for this issue from #1467 and confirmed that the errors still occur with the latest main (ITk ttbar+PU200):

19:34:43    Propagator     ERROR     Step failed with MagneticFieldError:1: Interpolation out of bounds was requested

So back to square one. What next?