kalfitalg: recent developments and tests
DESCRIPTION
KalFitAlg: recent developments and tests. WANG Dayong Jan.18,2006. Outline. Recent developments and some tests Status and usage Present performances Existing problems and possible solutions The next to do. A retrospect: what we achieved for Boss5.0.0(Nov,2005). - PowerPoint PPT PresentationTRANSCRIPT
KalFitAlg: recent developments an
d tests
WANG DayongJan.18,2006
Outline
Recent developments and some testsStatus and usagePresent performancesExisting problems and possible solutionsThe next to do
A retrospect: what we achieved for Boss5.0.0(Nov,2005)
Event model in both TDS and DST are fixed, root file persistified, the loop closed for KalTrack
Previous memory leakage problem extensively inspected and cured by a new model with improved track-hit relation after discussion
Check and use nearly the same material and geometry parameters with simulation
Run smoothly with both Pattern Recognition algs and nearly the same or a little better helix parameters and momentum as PR algs obtained, the tendency of error matrix variation is correct with 5 particle hypotheses.
0.4GeV/c Proton, theta=90degree
Mean -0.002364
Sigma 0.01272
Mean -0.002364
Sigma 0.01272
From wangdy’s report at software meeting Nov.3
Recent developments since Dec
Backward fitting with different mass hypotheses Flight length and TOF calculation Sampling length calculation Extraction of control flags and parameters Codes adjustment and comment to make it more clear a Many checks and tests for multiple scattering, energy
loss treatment, inner wall treatment, choices for cuts, etc (see the following)
Tests and debugWith and without material effectDifferent LR decisionAll hits or used hitsDrift distance and tof correctionComparison of different mass hypothesesComparison of different matrix initializationDifferent cuts for Δχ2
Status and usage
Helix and error matrix at IP: 5 particle hyp. directly for physics analysis
Helix and error matrix at MDC innermost layer: 5 particle hyp. for 2nd vertex fitting
Helix and error matrix at MDC outermost layer: 5 particle hyp. (optional), for track extrapolation
Flight length from IP to outermost layer and time of flight: different particle hyp. optional, for TOF recon.
Sampling length at each layer: for dE/dx recon
Control flags and parameters muls: multiple scattering 0:close 1: open loss: energy loss 0:close 1:open lr : how to set lr ambig 0:independendtly 1:use lr from PR matrixg: the factor to use in matrix initialization debug: debug verbose 0:close 1--4: lower to maximum ntuple: control of ntuple using bitmap FEDCBA
A:n101 B:n103 C:n102/n104 D:n105 E:mctruth in n101 F:error matrix activeonly: whether use active hits only 0:all 1:active hits only matfile, cylfile: dir and name of data file currently used deltachi2_cutf , deltachi2_cutfs: Δχ2 cuts pe_cut , pmu_cut , ppi_cut , pk_cut , pp_cut: mom cut in filter back :backward fitting 0:close 1:open i_back: mass hyp. used in backward fitting <0 :only pion Pt: pt cut for backward fitting tofflag : re-calculate the drift distance with tof correction tof_hyp: mass hypothesis when calculate TOF
Choice of cuts for Δχ2
One cut is not good enough, so we set two, different for forward and backward fitting
The criteria is the distribution of chisq and probability
The cut is tested with kk2f, dimu, and different tester samples
At present, the cut is set as 7.0 and 9.0 respectively, still need optimization
1GeV/c Pion Δχ2=7,9e
mu
pi
K
p
trk
p
K
pi
mu
e
length tof
Filter(forward)
smoother
Psi’’ inclusive Δχ2=7,9e
mu
pi
K
p
trk
p
K
pi
mu
e
length tof
Filter(forward)
smoother
Two track resolution(1)
Two track resolution(2)
Two track resolution(3)
Helix parameters:1GeV pion
Helix parameters:1GeV pion
Helix parameters:1GeV pion
Helix parameters:1GeV pion
Helix parameters:1GeV pion
Helix parameters:1GeV pion
Helix parameters:1GeV pion
Helix parameters:0.5GeV proton
Helix parameters:0.5GeV proton
J/Psirho pi To test the error matrix with the help of kine
matic fitThe algorithm is provided by Kanglin He4C efficiency is lower, so some problem mus
t existCareful check and test is still going on
J/Psirho pi :test
MdcTrkRecon
After KalFit
TrkReco After KalFit
event 1000 1000 1000 1000
2 good trks 459 459 663 663
2 pion 375 375 618 618
>=2 photon 351 351 564 564
VtxFit 350 350 562 563
4C Fit 319 223 483 315
5C Fit 315 222 475 309
J/Psirho pi :test
without wall
Inverse lay dire
Inverse wir direc
Inverse bck-fwd
Inverse lr defin
VtxFit 350 350 350 350 350
4C Fit 219 218 213 221 216
5C Fit 218 218 212 219 215
Matrix gain=1
Matrix gain4000
Diag Matrix
withoutNUMF
No mat effect
VtxFit 350 349 350 350 349
4C Fit 226 223 216 223 212
5C Fit 224 222 215 222 212
Possiblities• Multiple hits in a single layer
• Sign conventions
• Others???
Other issues
Non Uniform Magnetic Field:Wire sagita: the framework for this is there;
c++ implementation of the correction function by Jike Wang
Wire shift, x-t relation, wire resolution: from calibration
The next to do
Find out the origin of the problem rho pi and improve
Optimization of all the parameters and cuts with different physics events
More error matrix check with other physics channels
Documentation and notes