Hit Finder Validation & Prospects for Purity Measurement Matthew Thiesse 7 September 2016 35-ton Sim/Reco Meeting 1
Overview ● Recent progress has been good, some interesting plots ahead ● The priority to fjnd a “good event” list has been narrowed to fjnding a “good wire” list since the levels of signal and noise vary so much over time and throughout the detector ● MC studies have shown what should be possible from the hit reconstruction, does it still apply for real data??? Is calorimetric measurement even possible? 2
Code changes ● Committed new reconstruction fhicl fjles and new Hit Reconstruction module: – dune/HitFinderDUNE/RobustHitFinder_module.cc – dune/fcl/dune35t/reco/standard_reco_dune35tdata_robusthit.fcl – dune/fcl/dune35t/reco/standard_reco_dune35tsim_robusthit.fcl ● Fixed some existing muon counter code: – dune/DetSim/SimCounter35t_module.cc ● Now gives accurate timing information – dune/T0Reco/t0reco.fcl ● Use @local::dune35t_t0counter for real data ● Use @local::dune35t_t0countersim for simulated data ● Not sure if people are using T0Counter_module.cc to fjnd counter coincidences and anab::T0 objects, but you should! It works! – Now works for simulation too, with the appropriate fhicl tag 3
Last week ● Validating RobustHitFinder ● For real data, situation looks bleak ● Red line ----> – Essentially hit fjnding effjciency vs. drift distance – Should be a fmat line – hit fjnding effjciency should not depend on drift distance Drift distance (cm) – i.e. the hit fjnding threshold Red Line: Average number of should be low enough that hits per event. Divide this by even the lowest charge hits number of good wires to get are recorded 4 an efficiency.
MC Study ● Did MCC-scale simulation of opposite-counter East-west muons ● Test RobustHitFinder “effjciency” ● Using threshold of 2*RMS, no loss of hit-fjnding ● 336 total collection plane wires possible for track parallel to APA effjciency with drift distance ● Roughly 290 wires not already ignored by was observed. channelstatus_dune.fcl ● Simulated noise: 2.5ADC ● So, divide 315-ish (red points) by 290 is better than 100% effjciency ● e- lifetime: 3ms ● This means more than one hit is recorded per event per wire on average (deltas) 5
MC Study ● Need to confjrm that it is consistent on a channel-by-channel basis since the threshold is determined on a channel-by-channel basis ● I apologize for the messy plot, need to fjnd a better way to present this info… ● Each line represents the same thing as the red lines in slides 4 and 5, except one black line per channel ● Confjrms expectation: no loss of effjciency with drift distance, all wires have more-or-less fmat effjciency. 6
MC Study ● What if the hit fjnding threshold changes? Or equivalently, the signal-to-noise is reduced? ● Make an artifjcial cut on hit amplitude... Amplitude: (ADC counts) 7
MC Study ● Make a cut at 25ADC amplitude ● Drastic loss of effjciency with drift distance for almost all wires 8
Real 35-ton Data ● For real data, situation is grim ● For clarity, straight line fjt is shown in red, rather than jagged TGraph lines ● Hit fjnding threshold: 2*RMS ● Interesting, some wires have very low effjciency, consistently. – TPC3, not yet corrected for whether wires are “active” or not during the run 9
Real 35-ton Data “REAL” HITS Threshold = 2.0*RMS 10
Real 35-ton Data “NOISE” HITS *Note: Effective track length on wire Threshold = 2.0*RMS is undefined if a hit is not on a track. Use wire pitch (0.449cm) instead. 11
Real 35-ton Data NOISE HITS REAL HITS Threshold: 1.5*RMS 12
Real 35-ton Data NOISE HITS REAL HITS Threshold: 2.5*RMS 13
Moving Forward ● Purity measurement is ● The RobustHitFinder seems to be validated for low-noise data (i.e. still not out of reach! MC) ● With enough events, ● No conclusion yet on real 35-ton even 1 good wire should data ● Need to make extremely tight be able to give a decent cuts on channel/wire noise RMS measurement of purity! ● Lots of checking with event ● The more consistently display good wires, the better! ● Possibly found more “Bad” wires to add to the list, will probably update the list before the end of this week. 14
Recommend
More recommend