d. jason koskinen fnal collaboration mtg. 10/2005 1 near detector efficiency
DESCRIPTION
D. Jason Koskinen FNAL Collaboration mtg. 10/ Track Efficiency – How many planes hit for individual tracks Plane Efficiency – record how many tracks traverse plane and how many hits are recorded XXXXXXXXXXXX Vertex End PlaneTRANSCRIPT
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 1
Near Detector Efficiency
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 2
Rehash
Examining the efficiency of the Near Detector to record muon hitsEfficiency defined as #hit planes expected/#hit planes seenR1.16 was used initiallyCuts– Pitt track quality(chi2/ndof<20,|beginU -beginV|<6,
pass tracker fit)– Through going spill muon– Fiducial: 0.3m<U<1.8 and -1.8<V<-0.3 for vertex and
track end-2 planes
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 3
Track Efficiency– How many planes hit for individual tracks
Plane Efficiency– record how many tracks traverse plane and how many
hits are recorded
X X X X X X X X X X X XVertex End Plane
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 4
Initial Results
Calorimeter Tracks (120 planes)– Partially Instrumented (PI)
was .8883 +/- .0136– Fully Instrumented (FI)
was .8783 +/- .0140
Inefficiency stems from – Tracking– Poisson(PMT)-
Landau(Scintillator) nature of light
– Other Stuff?
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 5
Quick Check
Study of the Far Detector put the tracking efficiency at ~.95 (John Marshall)
– Extracting the tracking efficiency from overall efficiency yields the inefficiency due to light
● 1-.88/.95=.074
Is the ~7% inefficiency that is not track related solely from light or is it systematic of another problem?– Examine ADC per plane to get idea of light
contribution
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 6
Top plot shows summed ADC for all spill muon hits (w/o cuts) in May– Landau fit for comparison
Bottom shows similar with a 7.4% increase in low ADC counts– Simulates 0% light
inefficiencyThe ~7.4% is not totally light dominated
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 7
Monte Carlo Comparison
R1.16 High Energy Monte Carlo– PI .9446 +/- .0172– FI .9485 +/- .0140
MC is 6-7% higher using same method/codeDiscrepancy is big
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 8
MC/Data Discrepancy causes
Low level ADC hits in MC, which are not seen in dataMore MC tracks staying in analysis region of detector – inside volume containing both FI and PI planes
Data efficiency being 'pulled' down handful of single tracks with low efficiencyReconstruction not finding hits in data that are part of track
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 9
Both MC (top) and Data(bottom) track efficiencies are 'smooth'No peak at low efficiency values
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 10
Data(red) MC (blue)MC and Data are similar in ADC level
<2% of events fall outside fidvery similar to MC
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 11
Tracking
Removed possible tracking issues– Used all hits in event, regardless of (U,V) position
Efficiency increase of ~1% in both MC and Data
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 12
Inefficiency not caused by:
Low level hits– MC and Data have similar ADC distribution
More Data tracks move out of fiducial region– MC and Data have similar hit distribution
Handful of tracks with low efficiency in data– No spike in Data track efficiency plots at low
efficiencyThe only bit left is that the reconstruction is not finding or allocating hits to the track
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 13
More possible causes
'Dead' bits of Near Detector– Strips or planes that may not respond as effectively as
expectedAlignment issues – Are muon tracks in Data sneaking through gaps
Actual hits are not being included in event
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 14
Uniform distribution for both methods
Two methods of establishing missing hit position– Neighbour – interpolate missed hit position from adjacent
hits– U/V – use hit positions in next similar plane
● ex. Missed hit is a U plane, two nearest hit U plane positions are used
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 15
GapsIdentified by missing hits near strip edges– Hits congregate near edge more in MC than data– Tracks not 'slipping' through gaps
R1.16 Data R1.16 MC
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 16
R1.18R1.18 Data matches R1.16 MC, but not Data
R1.16 R1.18 MC
Track hits
All hits
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 17
2005-06 R1.16 2005-06 R1.18 MCTrk_efficiency PI 0.8794 0.9457 0.9465All_efficiency PI 0.8867 0.9481 0.9492Trk_efficiency FI 0.8863 0.9370 0.9408All_efficiency FI 0.8726 0.9387 0.9431
Same set of data (06/2005) with different reconstructions (R1.16, R1.18) have different efficenciesSuggests that hits are not being included in events in R1.16, but are in R1.18
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 18
Look at same event in R1.16 and R1.18 when efficiencies differ– R1.16 has 69% track efficiency, R1.18 has 94%
R1.16
R1.16
R1.18
R1.18
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 19
End
Finished– The culprit of the Near Detector efficiency issue is
hits not being included in event● slicer?
– Old code sucks, new code goodOutstanding– Missing hit location between MC and Data
● Missing hit location is similar for R1.16 and R1.18 Data, but does not match MC
● I'm not doing this
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 20
D. Jason Koskinen FNAL Collaboration mtg. 10/2005 21