Track fit differences in DD4HEP versus DetDesc
Currently we use parametrized scattering covariance / corrections in the default track fit. For a direct geometry-based full track fits with e.g. TrackMasterFitter
we see large differences (e.g. in efficiency, momentum pulls, ...) both between DD4HEP and DetDesc and between different versions of the geometry within both DD4HEP and DetDesc.
Errors in the configurations of materials in the geometries have been spotted (especially in DD4HEP) that could lead to such differences. It clearly shows we should keep track of changes in the geometry and have a feedback handle for work on the geometry.
To do
-
add parametrized scatter parameters to ParamFiles, see Rec!3780, Rec!3808, lhcb-datapkg/ParamFiles!73, Rec!3686, Detector!471 -
add qmtest to track determination of scatter parameters, using above functionality -
recover efficiency in full direct geometry-based track fit (use above test for feedback to sub-detectors / simulation?)
Follow up from !2907 (merged), to keep track of issues on differences in DD4HEP/DetDesc versions
The following discussion from !2907 (merged) should be addressed:
-
@dovombru started a discussion: (+16 comments) @msaur do I understand correctly that you are in the process of following up on some test failures? I.e. it makes sense to take another look at the reconstruction changes maybe early next week?