First steps on optimising RICH analysis Outline: Improved ring-track matching for low momentum (W.Koenig) Pad multiplicity and comparison with simulation Summary and conclusion Christina Dritsa JLU Giessen XXV Hades Collaboration Meeting, GSI. Darmstadt 19-23 November
Improve ring track matching for low momenta W.Koenig
Improve ring track matching for low momenta W.Koenig
Improve ring track matching for low momenta W.Koenig
Improve ring track matching for low momenta W.Koenig
Pad multiplicity and comparison with simulation
RICH analysis - reminder 0. Hardware thresholds (pedestals, cut on 3xσ noise) 1. Cleaning: Removes 2 types of structures single pads (mainly firing due to unusually high noise) large structures with high pulse (from direct hits) 2. Labelling: flag areas to search for rings (increase speed of analysis) 3. Ring finding “Hough” transform Pattern matrix Question: Is there room for optimising these steps to enhance the ring finding efficiency without creating more fake rings?
Pedestals S0 S1 S2 S3 S4 S5 DATA, apr12 SIMULATION, apr12 Uniform distribution of thresholds over all the sector surface Same for simulation
Thresholds S0 S1 S2 S3 S4 S5 DATA, apr12 SIMULATION, apr12 Pattern may be explained by the different lengths of circuit paths connecting the pad and the amplifier. This leads to different capacitances and hence the noise of the channels vary slightly. Middle lines with high thresholds are due to missing CsI cathode coverage at the edges of the 3 sector segments.
RICH “anatomy” http://hades-wiki.gsi.de/pub/RICH/RichFrontEnds/sector.pdf Anton Johann Kastenmüller, PhD thesis
Same effect observed before (old electronics) Anton Johann Kastenmüller, PhD thesis (2000) Note that in simulation the noise is always represented by a Gaussian function. This is not the case in reality: noise distribution has “tails”!
Investigation of the pad multiplicity for different event multiplicity classes
Number of hits ToF+RPC
All firing pads, Sector 0, day 123 0-20 > Slight dependence on event multiplicity. > Inner part of sector has relatively high occupancy > Similar behaviour observed for all sectors 80-100 160-180
Single pads removed, sector 0, day 123 > No dependence on event multiplicity, as expected. > Similar behaviour observed for all sectors (not shown here)
Single pads remaining, sector 0, day 123 Slight dependence on event multiplicity. Inner part of sector has relatively high occupancy Are thresholds lower than needed? Enhancement of fake rings? Similar behaviour observed for all sectors
Comparison to simulation: Number of remaining pads as a function of theta DATA SIMULATION
Comparison to simulation: Number of remaining pads as a function of theta DATA SIMULATION
Comparison to simulation: Number of rings as function of theta Data, all sectors Simulation, all sectors 100k events 60k events Real data are seemingly dominated by high noise in the low theta region. This might lead to fake rings which pass the standard ring quality cuts! Discrepancy with simulation!
Ring-to-proton matching θrich S0 Select protons from ToF/RPC: β<0.7 charge==1 mom>700MeV Select Rings: AC>70 Centr>0 matched ring index Selected rings match to identified protons from ToF/RPC. Clear enhancement of the fake matches in the low theta region (RPC region)
Summary New method introduced by Wolfgang improves ring-track matching for low momentum tracks by a factor up to 6! Non uniform noise levels lead to non-uniform threshold settings: Low theta region seems to have much higher occupancy potentially due to too low threshold settings Problem with “good quality” fake rings matching to non-leptons for low theta Discrepancies between simulation and measurement for pad and ring multiplicities Outlook Need to bring simulation and measurement closer! Investigate different threshold settings (e.g. increase thresholds in inner theta region) Update noise in simulation: Replace Gauss by “Gauss+tails” taken by measurement Investigate close pair rejection (in progress by Wolfgang)
Improve ring track matching for low momenta W.Koenig
Same effect observed before (old electronics) Anton Johann Kastenmüller, PhD thesis (2000) Note that in simulation the noise is always represented by a Gaussian function. This is not the case in reality: noise distribution has “tails”! Effect smoothed out with improved grounding. Measurements done in laboratory with pulser.
close pair investigation W.Koenig