Pixel seeding in phase2 is worse with CPE generic than with fast

[sroychow]

Reported by TrkPOG experts during Phase2 simulation meeting and later in CMS Talk thread. On 100 events of workflow 20834.1, D88 geometry; in CMSSW_13_0_0_pre4, the generic CPE perfroms worse than the fast.

effandfakePtEtaPhi-1340.pdf

More plots here:- http://uaf-10.t2.ucsd.edu/~slava77/figs/lst/CMSSW_13_0_0_pre4-lst12/mtv-oob-fast/plots_initialStep.html

[sroychow]

I opened this to book-keep discussions in a single place.

[mmusich]

Confirmed at the level of rechit resolution:

10 events of TTbar, noPU, D88 in CMSSW_13_1_X (wf 20834.0):

• red = vanilla release
• blue = process.siPixelRecHits.CPE = "PixelCPEFastPhase2"

Would be very curious how phase1 looks like

[mmusich]

for the curious ones:

• 100 evts of 12434.0 in CMSSW_13_1_X_2023-03-05-0000: http://musich.web.cern.ch/musich/display/CPE_check_13_1_X_phase1/
• 100 evts of 20834.0 in CMSSW_13_1_X_2023-03-03-1100: http://musich.web.cern.ch/musich/display/CPE_check_13_1_X_phase2/

[slava77]

100 evts of 12434.0 in CMSSW_13_1_X_2023-03-05-0000: http://musich.web.cern.ch/musich/display/CPE_check_13_1_X_phase1/

how does the performance look at the higher level, is seeding efficiency about the same with fast and generic or better?

[mmusich]

how does the performance look at the higher level, is seeding efficiency about the same with fast and generic or better?

In phase-1, CPE fast looks marginally better at low pT:

[mmusich]

@sroychow any feedback on this?