GArSoft Tracking Update Tom Junk HPGTPC Meeting November 2, 2018
Updates Since Last Week • Justo gave me a drift velocity as well as longitudinal and transverse diffusion coefficients for 10 bars of P10 gas operated at 400 V/cm. • There was a factor of sqrt(2) in the definition different from LArSoft – now understood. • Diffusion is about half of what I had been using before. Hits now have half the scatter. Pattern recognition is now much easier. • We don't scale diffusion with sqrt(drift time), but with sqrt(distance) (according to Justo's numbers). Can go back to scaling with sqrt(time) if need be, e.g. if we need to adjust the drift voltage. • I increased the sampling of the Gaussians to at least 10 points ("clusters"), and fewer drifting electrons per sample. These integrals had been tuned for liquid. • I did not tune the hit clustering yet. 2 Nov 2, 2018 Tom Junk | GArSoft Update
Updates Since Last Week • I tried to match reconstructed tracks to MC truth and found several bugs. • Fitter was actually okay. I exported track parameters and hits and plotted them in ROOT and they matched up. • Problem was in downstream representations: I miscalculated px, py, and pz from track parameters in a few places. • Exporting data to ROOT is clumsy – better to use the built-in event display for scanning many events. So I updated that to draw the helices from track parameters. • Event displays now show two track helices per track. One for the track parameters at one end, one for the other end. 3 Nov 2, 2018 Tom Junk | GArSoft Update
The GENIE Sample • New TPC Center is at (0, 218.196,1124.02) cm • Shoot 2 GeV numu and nue's in the +Z direction starting at (10,200,1000) cm. • Require one neutrino interaction in the TPC Gas in each event • Genie in "mono" mode ignores the beam width. You get an infinitely narrow beam. • I had originally shot it along the cathode plane Eldwan put in there. Thanks for the detail! And I only got carbon, hydrogen, and nitrogen interactions • Sample size = 1000 events including CC and NC interactions. • So far I plotted only the numuCC interaction variables. nueCC are expected to be similar given that electrons don't shower (they brem though). 4 Nov 2, 2018 Tom Junk | GArSoft Update
Some New Event Displays 5 Nov 2, 2018 Tom Junk | GArSoft Update
Some New Event Displays A busy event 6 Nov 2, 2018 Tom Junk | GArSoft Update
Some New Event Displays 7 Nov 2, 2018 Tom Junk | GArSoft Update
Some New Event Displays 8 Nov 2, 2018 Tom Junk | GArSoft Update
Some New Event Displays 9 Nov 2, 2018 Tom Junk | GArSoft Update
Issues Seen in Event Displays • "Unnecessarily" split tracks A delta ray or a scatter can make the pattern recognition split a track in pieces. On the to-do list to join these. • Joined tracks! A track going away from the primary vertex may be joined with one also going away but in the opposite direction. - This is going to be tough in any case – an event may appear to have just one track in it when in fact it has two. - Scattering means events with genuinely one track may look like they have two. - Sometimes the charge may appear to change sign along the length of a track. Not yet part of the pattern recognition to detect that. • Kalman filter slavishly follows little wiggles especially at ends of tracks – can tune parameters to help with this. 10 Nov 2, 2018 Tom Junk | GArSoft Update
GENIE Sample MC numuCC Distributions 2 GeV neutrinos traveling along the +Z direction Blue=true Red=true quantity, but with a reco match (only first two plots_ 11 Nov 2, 2018 Tom Junk | GArSoft Update
GENIE Sample Tracking Distributions Two entries in track r, pt, and px plots – one for each end of the track 12 Nov 2, 2018 Tom Junk | GArSoft Update
GENIE Sample Vertexing Many more events have vertices now compared to last week. Some events just have the muon track reconstructed. 13 Nov 2, 2018 Tom Junk | GArSoft Update
The To-Do List From Last Week • Associate true and reco tracks in analysis tree plotter - Much better now – simple bugs fixed. Still work to do, however • Make tracking efficiency & resolution plots vs Pt, R, Px, track length - Still need to do some work before these are believable • Fit t0 for tracks crossing the cathode plane • Revisit Kalman filter stepping to make sure we are starting off each track fit with the best guess - Getting more confident about this • Revisit second-pass hit assignment. • Lower min number of hits requirement on second pass, or make a separate cleanup module to find short tracks near primary vertex • Put in electronics response 14 Nov 2, 2018 Tom Junk | GArSoft Update
Recommend
More recommend
