TreeSearch Track Reconstruction for GEMs

Transcription

1 Track Reconstruction for GEMs Ole Hansen Jefferson Lab SBS Collaboration Meeting 19 March 2010 Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

2 Introduction Motivation Existing tracking simulation very basic January 2010 Technical Review: We recommend to provide a full simulation of the tracking algorithm Realistically simulate occupancy of chambers (esp. front trackers) Demonstrate efficient track reconstruction for high-occupancy conditions in GEMs Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

3 Introduction Motivation Existing tracking simulation very basic January 2010 Technical Review: We recommend to provide a full simulation of the tracking algorithm Realistically simulate occupancy of chambers (esp. front trackers) Demonstrate efficient track reconstruction for high-occupancy conditions in GEMs Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

4 Introduction Motivation Existing tracking simulation very basic January 2010 Technical Review: We recommend to provide a full simulation of the tracking algorithm Realistically simulate occupancy of chambers (esp. front trackers) Demonstrate efficient track reconstruction for high-occupancy conditions in GEMs Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

5 Introduction Motivation Existing tracking simulation very basic January 2010 Technical Review: We recommend to provide a full simulation of the tracking algorithm Realistically simulate occupancy of chambers (esp. front trackers) Demonstrate efficient track reconstruction for high-occupancy conditions in GEMs Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

6 Tree Search Suggested by Dell orso et al., NIM A 287, 436 (1990) Recursive template matching Fast and efficient (speed and memory) Proven at HERMES with various drift chambers Used by Qweak (based on HERMES code) for HDCs and VDCs Well-debugged code exists for Hall A BigBite MWDCs Appears suitable for SBS front tracking system Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

7 Tree Search Suggested by Dell orso et al., NIM A 287, 436 (1990) Recursive template matching Fast and efficient (speed and memory) Proven at HERMES with various drift chambers Used by Qweak (based on HERMES code) for HDCs and VDCs Well-debugged code exists for Hall A BigBite MWDCs Appears suitable for SBS front tracking system Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

8 Tree Search Suggested by Dell orso et al., NIM A 287, 436 (1990) Recursive template matching Fast and efficient (speed and memory) Proven at HERMES with various drift chambers Used by Qweak (based on HERMES code) for HDCs and VDCs Well-debugged code exists for Hall A BigBite MWDCs Appears suitable for SBS front tracking system Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

9 Tree Search Suggested by Dell orso et al., NIM A 287, 436 (1990) Recursive template matching Fast and efficient (speed and memory) Proven at HERMES with various drift chambers Used by Qweak (based on HERMES code) for HDCs and VDCs Well-debugged code exists for Hall A BigBite MWDCs Appears suitable for SBS front tracking system Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

10 Tree Search Suggested by Dell orso et al., NIM A 287, 436 (1990) Recursive template matching Fast and efficient (speed and memory) Proven at HERMES with various drift chambers Used by Qweak (based on HERMES code) for HDCs and VDCs Well-debugged code exists for Hall A BigBite MWDCs Appears suitable for SBS front tracking system Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

11 Tree Search Suggested by Dell orso et al., NIM A 287, 436 (1990) Recursive template matching Fast and efficient (speed and memory) Proven at HERMES with various drift chambers Used by Qweak (based on HERMES code) for HDCs and VDCs Well-debugged code exists for Hall A BigBite MWDCs Appears suitable for SBS front tracking system Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

12 Tree Search Suggested by Dell orso et al., NIM A 287, 436 (1990) Recursive template matching Fast and efficient (speed and memory) Proven at HERMES with various drift chambers Used by Qweak (based on HERMES code) for HDCs and VDCs Well-debugged code exists for Hall A BigBite MWDCs Appears suitable for SBS front tracking system Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

13 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

14 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

15 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

16 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

17 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

18 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

19 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

20 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

21 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

22 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

23 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

24 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

25 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

26 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

27 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

28 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

29 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

30 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

31 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

32 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

33 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

34 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

35 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

36 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

37 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

38 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

39 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

40 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

41 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

42 Successive Approximation Method Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

43 Key Advantages Tree-like search process allows fast template lookup: O(log N bins ) Symmetry and self-referential properties of the templates allow efficient storage. Only base patterns need to be stored: O(1MB). Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

44 Key Advantages Tree-like search process allows fast template lookup: O(log N bins ) Symmetry and self-referential properties of the templates allow efficient storage. Only base patterns need to be stored: O(1MB). Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

45 3D Processing 3D Matching Chamber front view good matchvalue poor matchvalue Repeat for each chamber group along z Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

46 3D Processing 3D Track Fitting Fit the linear equations A ˆβ = y using the coordinates y i of the best 2D fits in all planes i, ( ) ( ) x + mx z y i = i cos αi, y + m y z i sin α i where x, m x, y, m y are the parameters to be fitted (β k ). The fit is done by Cholesky decomposition of the normal equation (A T WA) ˆβ = (A T W)y, where W is the weight matrix (cf. ROOT s TLinearFitter). NB: Each plane can be at arbitrary z Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

47 Requirements Requirements Straight tracks 3+ planes per projection (coordinate) Normally: 3+ projections (but see discussion later) Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

48 Requirements Requirements Straight tracks 3+ planes per projection (coordinate) Normally: 3+ projections (but see discussion later) Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

49 Requirements Requirements Straight tracks 3+ planes per projection (coordinate) Normally: 3+ projections (but see discussion later) Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

50 Existing BigBite MWDC Code General BigBite MWDC Code Written in 2008 for E & Transversity Used in production replay Well tested and debugged by now Parallelized Event display available Very good analysis speed Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

51 Existing BigBite MWDC Code Data Residuals (E online results!) BB.mwdc.x1p.coord.3Dresid 300 h1 Entries 5930 Mean RMS Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

52 Existing BigBite MWDC Code Data Track y vs phi (E online results!) NB: approx. point target BB.tr.y:BB.tr.ph BB.tr.y BB.tr.ph Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

53 Existing BigBite MWDC Code Data Track χ 2 (E online results!) BB.tr.chi2 {BB.tr.n==1&&BB.tr.chi2<40&&BB.tr.ndof==8} h1 Entries 3742 Mean RMS Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

54 Analyzing GEM Trackers Required Code Modifications Required Modifications to BigBite Code for GEMs Analyzing GEM trackers is easier: No L-R ambiguity Additional correlations Required software changes: Add weighted averaging of strip signals Disable/remove handling of L-R ambiguities Exploit additional information from GEMs for 3D matching: Amplitude correlation Time correlation First two items already finished for PREX GEM analysis Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

55 Analyzing GEM Trackers Required Code Modifications Required Modifications to BigBite Code for GEMs Analyzing GEM trackers is easier: No L-R ambiguity Additional correlations Required software changes: Add weighted averaging of strip signals Disable/remove handling of L-R ambiguities Exploit additional information from GEMs for 3D matching: Amplitude correlation Time correlation First two items already finished for PREX GEM analysis Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

56 Analyzing GEM Trackers Required Code Modifications Required Modifications to BigBite Code for GEMs Analyzing GEM trackers is easier: No L-R ambiguity Additional correlations Required software changes: Add weighted averaging of strip signals Disable/remove handling of L-R ambiguities Exploit additional information from GEMs for 3D matching: Amplitude correlation Time correlation First two items already finished for PREX GEM analysis Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

57 Analyzing GEM Trackers Required Code Modifications Required Modifications to BigBite Code for GEMs Analyzing GEM trackers is easier: No L-R ambiguity Additional correlations Required software changes: Add weighted averaging of strip signals Disable/remove handling of L-R ambiguities Exploit additional information from GEMs for 3D matching: Amplitude correlation Time correlation First two items already finished for PREX GEM analysis Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

58 Analyzing GEM Trackers Required Code Modifications Required Modifications to BigBite Code for GEMs Analyzing GEM trackers is easier: No L-R ambiguity Additional correlations Required software changes: Add weighted averaging of strip signals Disable/remove handling of L-R ambiguities Exploit additional information from GEMs for 3D matching: Amplitude correlation Time correlation First two items already finished for PREX GEM analysis Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

59 Analyzing GEM Trackers Required Code Modifications Required Modifications to BigBite Code for GEMs Analyzing GEM trackers is easier: No L-R ambiguity Additional correlations Required software changes: Add weighted averaging of strip signals Disable/remove handling of L-R ambiguities Exploit additional information from GEMs for 3D matching: Amplitude correlation Time correlation First two items already finished for PREX GEM analysis Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

60 Analyzing GEM Trackers Required Code Modifications Required Modifications to BigBite Code for GEMs Analyzing GEM trackers is easier: No L-R ambiguity Additional correlations Required software changes: Add weighted averaging of strip signals Disable/remove handling of L-R ambiguities Exploit additional information from GEMs for 3D matching: Amplitude correlation Time correlation First two items already finished for PREX GEM analysis Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

61 Analyzing GEM Trackers U-V Chambers needed? Do we need U-V Chambers? Most X-Y correlations expected to be random U-V information may not sufficiently disambiguate because it is similarly random More like planes may be better than more coordinates (since TreeSearch is more effective with more planes) Unlike drift chambers, amplitude and timing correlations available Need conclusive answers from Monte Carlo Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

62 Analyzing GEM Trackers U-V Chambers needed? Do we need U-V Chambers? Most X-Y correlations expected to be random U-V information may not sufficiently disambiguate because it is similarly random More like planes may be better than more coordinates (since TreeSearch is more effective with more planes) Unlike drift chambers, amplitude and timing correlations available Need conclusive answers from Monte Carlo Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

63 Analyzing GEM Trackers U-V Chambers needed? Do we need U-V Chambers? Most X-Y correlations expected to be random U-V information may not sufficiently disambiguate because it is similarly random More like planes may be better than more coordinates (since TreeSearch is more effective with more planes) Unlike drift chambers, amplitude and timing correlations available Need conclusive answers from Monte Carlo Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

64 Analyzing GEM Trackers U-V Chambers needed? Do we need U-V Chambers? Most X-Y correlations expected to be random U-V information may not sufficiently disambiguate because it is similarly random More like planes may be better than more coordinates (since TreeSearch is more effective with more planes) Unlike drift chambers, amplitude and timing correlations available Need conclusive answers from Monte Carlo Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

65 Planned Monte Carlo Studies Tracking Monte Carlo Event Generator (e, photons) Geant (digitization) File INFN JLab ROOT file Analyzer / GEM TreeSearch Simulation Decoder Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

66 Summary Summary TreeSearch algorithm appears well suited for SBS Front Tracker reconstruction Successfully used by numerous other experiments To be tested with GEM data in PREX Preparations for SBS Tracking Monte Carlo in progress Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

67 Summary Summary TreeSearch algorithm appears well suited for SBS Front Tracker reconstruction Successfully used by numerous other experiments To be tested with GEM data in PREX Preparations for SBS Tracking Monte Carlo in progress Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

68 Summary Summary TreeSearch algorithm appears well suited for SBS Front Tracker reconstruction Successfully used by numerous other experiments To be tested with GEM data in PREX Preparations for SBS Tracking Monte Carlo in progress Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16

69 Summary Summary TreeSearch algorithm appears well suited for SBS Front Tracker reconstruction Successfully used by numerous other experiments To be tested with GEM data in PREX Preparations for SBS Tracking Monte Carlo in progress Ole Hansen (Jefferson Lab) TreeSearch Track Reconstruction for GEMs 19 March / 16