120 likes | 151 Views
Identify and fix bug in Muon Seeding algorithm to enhance pT calculation accuracy and boost reconstruction efficiency in CMS standalone muon tracking. Investigate missing muon reconstructions, validate fixes, and collaborate with UC Riverside group for further improvements.
E N D
Standalone Muon Seeding Rick Wilkinson
One Problem Solved:Bad Seed Efficiency in Cracks ORCA PTDR Standalone Muon CMSSW
Efficiency Bug Found Single-segment seeds in MB1 and MB2 were accidentally being removed Bug was introduced in porting from ORCA
Endcap Seed pT Resolution • Endcap pT algorithm makes some strange assumptions: • Measures pT only from direction of “best” segment • Uses only local magnetic field, not field integral
Estimating Endcap Muon Seed pT • Should estimate pT from the bending between stations, not slope within a station • Field integrals look reasonably linear for a given chamber type. • Try to parametrize: • pT = (c1 + c2 * h) / Df
Endcap Seed pT algorithm • If there are segments in stations 1 & 2, get pT from the bend between them: • pT = (c1 + c2 * h) / Df • Use 10 GeV muons for parametrization • If not, try stations 1 & 3. • If not, then what? • Use old algorithm for pT? • or assign a straight track with errors up to min pT : • q/p = 0, • s(q/p) = 1/ (3GeV)?
Endcap Seed pT: Results Blue is new, red is old, for 100 GeV muons Overlap region is bad, but there should be some good barrel-only seeds there
Effect on 100 GeV standalone endcap m- • Helps prevent a few catastrophic failures • But some muons are missing! Needs to be investigated.
Effect on 1 TeV standalone endcap m- • Seems to help a lot! • Piotr showed a plot of bad TeV standalone reconstruction last meeting • Still too many misreconstructed. • Barrel also has a spike at low pT.
To-do next • Find why some muons aren’t reconstructed • Seed errors too small? • Get others to validate what I have, and release it into 1_3_1. • Committed to CVS • Parametrize errors as well • Seed errors could have a big effect on final reconstruction • Deal with barrel-endcap overlap region
Conclusion • A lot of the bad effects we’ve seen in muon reconstruction can be directly blamed on bad seeding. • The efficiency dips at |h|=0.2 were due to a seeding bug. • Endcap seed pT was wrong, and could cause bad standalone reconstruction. • Endcap seed pT can now be calculated from bend in phi between stations. • New seeding not quite ready: unexplained inefficiency • UC Riverside group is starting to rewrite the seeding, but are willing to use this pT calculation. • Still lots of work and tuning left to do.