Prop plane

[cerati]

PR description:

Implement propagation to plane and kalman update to plane.
Propagation to plane is used for propagation to the hit, while propagation to r and z are used for propagation to the layer center. Propagation to the hit starts from the state at the previous hit (by re-calling the mkfndr->inputTracksAndHitIdx(...) ).
This also updates the material effects, as developed by @slava77 in pr #106.

PR validation:

http://uaf-10.t2.ucsd.edu/~cerati/prop2plane-PR/

[slava77]

This also updates the material effects, as developed by @slava77 in pr #106.

we were discussing how to interpret the results, since there are some regressions

in the initialStep

please clarify if the difference between red and black is only prop to plane or if there is more, like some other differences in how material is handled or something else.

[cerati]

Red does not have new material nor p2plane. It just changes whether the material effects are applied before or after the error propagation.

[slava77]

Red does not have new material nor p2plane. It just changes whether the material effects are applied before or after the error propagation.

Is it possible have the new material separately to decouple the prop2plane itself. It may be just a matter of adding another reference from the earlier PR

[cerati]

I set PROP_TO_PLANE = false in MkFinder.cc, which should disable all p2plane changes and keep only changes to the material. Here is the validation: http://uaf-10.t2.ucsd.edu/~cerati/prop2plane-PR-checkPR106/

[mmasciov]

I set PROP_TO_PLANE = false in MkFinder.cc, which should disable all p2plane changes and keep only changes to the material. Here is the validation: http://uaf-10.t2.ucsd.edu/~cerati/prop2plane-PR-checkPR106/

Comparing black to orange, and orange to reference, then this would imply that the regression comes from PR #106, and that the additional changes actually reduce such a regression?

[slava77]

Comparing black to orange, and orange to reference, then this would imply that the regression comes from PR #106, and that the additional changes actually reduce such a regression?

this seems to apply to the endcap part.
The barrel part (e.g. loss of hits) seems to be from prop to plane.

[slava77]

in a follow up to our chat about free vs constrained Jacobian, it would be good to check muons, I wanted to suggest to check low pt, smth like 0.2-0.6 range so that the incidence angle in the barrel is generally relatively large

[cerati]

agreed. But it would also be good to separately check larger pt muons. If there are samples this is easy enough for me to do

[slava77]

agreed. But it would also be good to separately check larger pt muons. If there are samples this is easy enough for me to do

are you working with CMSSW or with bin files? Which release?
We discussed yesterday that @leonardogiannini may help with making samples.

[cerati]

Thanks. I am working with CMSSW, release CMSSW_13_3_0_pre2

[slava77]

Thanks. I am working with CMSSW, release CMSSW_13_3_0_pre2

from the earlier discussion, I guess we could benefit from some 10K events each, 10 muon events (AddAntiParticle = False to avoid continuous tracks) for

• pt 0.1 - 1 GeV
• pt 1 - 1000 GeV

[leonardogiannini]

Thanks. I am working with CMSSW, release CMSSW_13_3_0_pre2

from the earlier discussion, I guess we could benefit from some 10K events each, 10 muon events (AddAntiParticle = False to avoid continuous tracks) for

• pt 0.1 - 1 GeV
• pt 1 - 1000 GeV

despite some delay, here are the 10 muon samples for testing the tracking step:

/ceph/cms/store/user/legianni/sample-generation-CMSSW_13_3_0_pre2/step2_10muon_0p1to1_alt.root
/ceph/cms/store/user/legianni/sample-generation-CMSSW_13_3_0_pre2/step2_10muon_1to1000_alt.root

[cerati]

Thanks Leonardo. Here are the results:
http://uaf-10.t2.ucsd.edu/~cerati/p2plane-MuLowPt/
http://uaf-10.t2.ucsd.edu/~cerati/p2plane-MuHiPt/
Looks like the problem at high pt is macroscopic, so hopefully I can find a good friend to spend time with:
http://uaf-10.t2.ucsd.edu/~cerati/p2plane-MuHiPt/plots_building_initialStep/hitsLayers.pdf

[cerati]

The commit above fixes the high pT issue: http://uaf-10.t2.ucsd.edu/~cerati/p2plane-MuHiPt-fix/
There is still a loss of hits at low pT in the barrel. Not sure this is a problem or it just requires a new tuning of chi2 etc.

[osschar]

Recreated as #136 ... which is in turn cms-sw#43146. Closing.