status of besiii event reconstruction system zepu mao ihep 100049 besiii col. meeting 2006/01/12
TRANSCRIPT
Status of BESIII Event Reconstruction System
Zepu Mao IHEP 100049
BESIII Col. Meeting 2006/01/12
MDC Tracking
The events Reconstruction system
EMC Rec.
TOF Rec.
MU Rec.
1 MDC Fast track
2 Event Start Time
3 MDC-Tracking-14 MDC-Tracking-2
7 dE/dx PID6 Track Ext.5 Kalman fit
8 TOF Rec.
9 EMC Rec.
10 MuC Rec.
EMC Tracking
Have TOFInformation ?
Calculate Test by TOF andMDC Fast tracking Inf. Calculate Test by
Fast tracking Inf.
TOF NoTOF
MDC Fast Rec.
Information?
Calculate Test
by MDC Inf.
Yes No
Yes
Particle ID MDC Fast Tracking
dE/dx
No
Yes
The TEST is important for the momentum Res., spatial Res. of the charged particles, but it is uncertain in the online system due to: 1) The BESIII trigger system can’t separate every event up to beam bunch.2) The start time of charged event is determined by TOF_T, but the TOF reach time is different for different particle and different momentum. TEST should be calculated by offline system.
The Status of TEST (Event Start Time) Module (1)
developed by X. Ma
• Finished the programming and testing by TOF and MDC-fast tracking and got Pre. results.
• Continue to test the codes for cosmic events • continue to study the new method for MDC-fast tracking fail case.
by TOF(90.2%)* barrel 84.73% ,* end-cap 5.47%)* Test: 0.01ns* σ: 0.27ns
e-e+ ->hadrons, set TEST=0
by MDC( 4.8%): • Test: 0.75ns• σ: 1.17ns
Preliminary results of Code performance
• 95% events got exact TEST • judgment error: 0.88%* 5% need Cal. by MDC new method
By TOFTest: 1.998 ns σ: 0.30nsTest: 10.01 ns σ: 0.33nsTest: 18.02 ns σ: 0.29ns
preliminary result of Test performance “hadrons”, 10000, set Test to ”2”,”10”,”18”
by MDC Test: 2.6ns, σ: 1.4ns)Test: 10.9ns, σ: 1.3ns)Test: 18.7ns, σ: 1.1ns)
MDC Tracking Module(1) (based on Belle Lib. Developed by S.L. Zeng )
Init: Geometry-Surv., Cal.-Constant, Adjust-constant Get Hit Information ….
R- Tracking: Segment finding by conformal transformation and histogram method
Z Finding
3D Helix Fit
Circle &
Lin
e Fit
• Finished the coding and Pre. Checking, the Pre. Results are reasonable• More works need to do: increasing the efficient for low momentum tracks & add noise case.• Increasing the multiple tracks separation ability• optimizing the parameters of the codes
Segment Finding Algorithm• Conformal transformation. A circle which passes through the origin is transformed into a line.
Segments found by histogram of hit wires’ azimuthal angle Track linking from the outmost to the innermost layer with the directions of segments and apply the circle fit.
S-Z calculation in Z finding
The Pre. Performance of MDC Tracking(1)
σ(dr)~179μm σz ~1.9mm
σxy~ 139μm σP~5.7MeV/c
xy vs momentum
Eff vs momentum
pt/ppt vs momentum
Nhits vs momentum
from 1 GeV -
Results from e, , proton, J/, J/ee
1GeV; cos(θ): -0.8~0.8.
M0(MeV)
Eff(%)
dr(μm)
dz(mm)
σxy (μm) σP
(MeV)Chi2 nHits
E 0.51 99.7 180 1.9 139 6.1 57.2 44.5
105.7 99.8 179 1.9 139 5.7 57.7 44.5
Proton 938.3 100.6 193 1.9 148 6.8 64.3 44.5
J/ 3096.9 99.5 151 1.9 144 11.6 57.7 43.9
J/ee 99.2 148 1.8 145 11.9 58.6 44.0
MDC Tracking Module(2) (based on Babar Lib. Developed by Yao Zhang )
Init: Geometry-Sur., Cal.-Constant, Adjust-constant Get Hit Information ….
Link segments to 2-D tracks,
3D Helix Fit
Line &
Circle F
it
Tracking:
Segment finding:Search segments in each super-layer using a pattern look-up table.
Add stereo segments
• Finished the lots of work of migration, Geometry changes, Stereo track finding algorithm improved, system conversion: (track parameters)…,
• work well in V5 and used by physics study
• The Pre. performance of the codes is: tracking efficiency > 98% , spatial resolution about 110 μm, momentum resolution δp/p = 0.4%.
• Testing further with MDC calibration constants
• Further checking to increase the efficiency in low momentum and add noise case.
• optimizing for the parameters of the codes …..
MDC SegFinder Sequence
For every superlayer and every wire in 2nd layerform a group of 8 wires
Try 4 hit pattern for this groupCall tryPatterns()
Try 3 hit pattern for this groupCall tryPatterns()
Set massage of segment list
Sequence of segFinder::createSegs()
For every group
4 3
2
1 0
7 6 5
10101001
01234567
Wires No. 0-7
Set one word for a group of 8 wire, each bit for a wire. Set “1” for a hit wire, others “0” This octal value used for its group No.
The Pre. Performance of Tracking Mode(2)
e- at pt = 1GeV/cSpatial resolutionxy 110μm
e- at pt = 1GeV/cMomentum resolutionp 0.40%
Efficiency Vs Pt (e,μ,π,p)Average efficiency >98 %
z0
0.8 mmd0
0.1 mm
Nhits used in tracks comparison with MC truth
dE/dx codes developed successfully, released for physics study. More reconstruction algorithms studied to get best performance Particle ID is tested with MC samples, dE/dx resolution,
distributions, PID efficiency is reasonable.
Overview of the sof tware
M dcDedxReconDed xCalibAlg
Calibrationconst
CalibDataSvc
D edxCorrecSvc
E ventDataSvc
T ransient DataStore
T ransient C alibData Sto re
DST
converter
converter
<<uses>>
< <uses>>
<<uses>>
<<uses>>
< <uses>>
M dcGeo m Svc
<<uses>> <<uses>>
Status of dE/dx Module (Developed by D.Y. Wang)
6%
χ distribution for Kaon sample Prob ( K ) distribution for Kaon sample
Pre. performance of dE/dx(1)
dxdE
mea dEdxdxdE
/
exp )/(
Distribution of is nearly a normal N(0,1)distribution
Distributions of probability function are flat
dE/dx seperation for 5 particles(MC) seperation power with dE/dx
Pre. Performance of dE/dx(2)
Good particle seperation in a wide range for different particles
The π/K seperation(3 σ ) reach about 800 MeV/c
Particle identification efficiency is more than 90% with MC samples
Track Extrapolation Module
(based on Belle Lib. developed by L. L Wang) Function: Extrapolate MDC tracks to Outer detectors: TOF, EMC and MUC Algorithm: based on GEANT4, Energy loss and Magnetic field. Multiple scattering effect put into error matrix.
TOF Reconstruction FlowTOF Reconstruction Flow
Initialize()
finalize()
execute()
TofGeomSvc
TofCalibSvc
Get Data
Get Tracks
Make Hits
Match Hitsand Tracks
ReconstructedTof Tracks
Data base
TDS
TDSTof Tracks
MDC Tracks
Tof Digits
Calibration constants
Geometry data
Unpacked
Reconstruct Tracks
(MDC Reconstracted)
TDS
Data Tracks
Find hits Find Tracks
Match Hits and Tracks
The Status of the TOF Rec. Module(based on Belle Lib. developed by Linli Jiang)
TOF Rec. package has been developed, and run well on V5TOF Rec. package has been developed, and run well on V5 The Pre. performance is:
• The Efficiency is about 99.8% with 1 GeV single e LAttenuation LAttenuation 3.2m, , Time Res. 62 ps
Pre. Performance of TOF Rec. code
Beta vs MomentumBeta vs Momentum
Mass distributionMass distribution
Checked by “LUND” MC data sample Checked by “LUND” MC data sample
Mass distributionMass distribution
The Status of EMC Rec.(developed by Wang Zhe and He Miao )
Digit Hit Cluster Shower
Emc Digit
Emc Mc Hit
from TDS
Input:
Emc Rec Shower
to TDS or Root
Output:
Data Flow• The Code EmcRec has finished the design, programming for barrel and
end-cap and work well in V5.0.
• Performance:
Eff: 100%
E 2 % for gamma with 1 GeV
• e /π separation studied
5mm, 5mm in barrelE 2 .5%, 6mm,
6mm for endcap
Energy Resolution vs EBy single gamma
Pre. Performances of EMC Barrel 1GeV gamma
theta
Position Resolution vs E
phi
Position Resolution vs E
E 2 % for gamma with 1 GeV
5mm
5mm
Pre. Performance of EMC Endcap
1GeV gamma
west end east end
σE/E vs layer
Res. (cm) vs
layer
Res. (cm) vs
layer
E 2 .5%, 6mm, 6mm
Monte Carlo Data
0.5Gev
1.0Gev
1.5Gev
Extrapolated
MDC track
Reconstructed
MDC track
Muc Reco. By Ext MDC tracks Alg.
Use Ext Track from Mdc as seed;
Searching hits gap by gap;
Searching on Barrel first, then EndCap;
Also Searching on neighboring segments;
Line Fit with hits on track; Ext track
Fired strips
Window
Barrel
End Cap
MUC Reconstruction Module (developed by Z.Y You)
•Finished the programming and testing by “Extrapolating MDC Tracks”and “2D/3D Road” tracking method.
* The ID method also studied. • “2D/3D Road” : Tracking EF: 99%
• “Extrapolating” : cosθ(-0.9~0.9) Eff=98.9% for 1.GeV Eff=93.3% for 0.5GeV
Muon ID Efficiency
GeV/c
%
0.5 0.75 1.0
Mu Efficiency
87.64 93.42 91.18
Pi
Fake rate18.96 13.04 6.92
Mu Efficiency
85.00 91.22 89.42
Pi
Fake rate18.92 12.78 6.24
MuID Alg
Global :MucRec + MuID
For all tracks with|Cos θ| < 0.9
Mis-identified muon includes : 1. Lost hits by Acceptance; 2. Lost hits by muc reconstruction;3. All hits found, but mis-identified as pion by MuonID Algorithm;
Helix parameters and error matrix looks reasonable. 5 Parameters of a track is reasonable comparison with MDC-Tracking’s. (but not improvement too more.) lots of work need to do, Parameters need further optimizing.
Output Pull check
Input Helix parameters from FastTrkAlg
Output parameters from KalFitAlg
proton events 1GeV/c
Package is migrated and work on V5, Five particle hypotheses is made; the materials is updated with BESIII’s.
Kalman filter track fitting package(based on BELLE Lib. Developed by D.Y. Wang )
Summary(1)
Module Name Pre. Results by general checking 1 MDC Fast track Trk-Eff 99 %, x160um, z4mm , p 10MeV
2 EvTime 95% events got exact TEST , err less 1%
3 MDC-Track-1 Trk-Eff 99%, p 6 MeV , x140m, z0 2 mm
4 MDC-Track-2 Trk-Eff 99%, p 4 MeV, x110m, z0 0.8 mm
5 dE/dx PID 6%, π/K 3 seperation about 800MeV/c
6 Trk-Ext Work well on V57 Klmfit Work well on V58 TOF Rec. Trk-Eff 98 %, T62ps, LALA 3.2m
9 EMC Rec. Trk-Eff 100 % , E2%, special resolution about 5mm
10 MuC Rec. Trk-Eff 93-99 %
Summary(2)• We finished lots of works: code design, programming, code
migration, code test, algorithm study, code optimization …
• Almost all sub-system work well on the V5, got the Pre.
Performance by the simple data sample, and released for
physics study.
• The results is different from two MDC Tracking codes,
due to the data sample is different. We will check the codes
and data(new MDC geometry) later.
• More works need to do to make each code work in the best
status by closely the data to real data case step by step.
谢谢Thanks
MDC Fast Tracking Module(based on Belle Lib. developed by xiaomei zhang)
x160 m
z6mm
p10MeV
Trk Eff99%
x160m z6mm
p10MeV
data unpacking
axial segments finding stereo segments finding
2D track finding
2D track fitting
3D track finding
S_Z track fitting
Test Calculation by TOF
A
B
TOF
PMT
Z=0
ZTOFCharged particle
scintillatorD
elcPMTprotofev ttttt
C
A->B: By MDC fast tracking
B->C: By information of TOF & fast track
constants
Test = TDCM(tof) -tev
C->D by MDC
information
Sense wire
Field wireDrift cell
A (test)
B
CD
elcwpdriftfev ttttt
Test Calculation by MDC
B->C by MDC track &
InformationA->B by MDC fast track
constants
Test = TDCM(MDC) - tev
Li sts of axi al hi ts andstereo hi ts
Conformal Transformation
Fi nd segments for each superl ayer(SL) wi th hi stogram of hi t
wi res’ azi muthal angle
Li nk segments
Parti al l y sol ve LR ambi gui ty of al l hi twi res usi ng nei ghbor wi res’ i nfomati on
Fi nd the best segment l i ststart wi th a seed i n outmost SL
Ci rcl e Fi t
Remove bad segmentCi rcl e fi t agai n
Sol ve LRCi rcl e fi t agai nCheck the qual i ty
ok
no
ok
no
Li st of R-phi Tracks
Sal vage other axi alsegments and hi ts
Ci rcl e fi tAppend to 2D track l i st
Assign shared segment or hi t
Make l i st of stereo segments usinghi stogram
Roughly sel ect segments wi thcross points of
r-phi track through SL
Cal culate (s,z) of hi tsand append them to zLi st
Take one s-z l i ne as a seed
Make possibl e s-z l i nes wi th hi ts i n zLi st(fi rst make the l onger)
ok
no
ok
no
Li st of s-z l i nes
Take a hi t i n zLi st
Whether hi t belong toseed
Append to seed
Select one of r-phiTrack Candidates
Qual i ty check for seed l i ne
Take the l ongestLine fi t
Cal culate (s,z) of hi ts
3D Hel i x Fi t
Fig.6 Flow of R-phi track finding(left) and z finding(right)
Structure of TOF• Aim: particle identification (PID)
Barrel TOF Endcap
TOF