Status of Fast Tracking Algorithm Mdc Hough Guowei

  • Slides: 14
Download presentation
Status of Fast Tracking Algorithm Mdc. Hough Guowei YU 8 th March 2006

Status of Fast Tracking Algorithm Mdc. Hough Guowei YU 8 th March 2006

Outline Introduction Mdc. Hough Algorithm Results and Discussions Summary

Outline Introduction Mdc. Hough Algorithm Results and Discussions Summary

Introduction q Algorithm Developments in MDC Reconstruction Developer S. L. Zang Y. Zhang X.

Introduction q Algorithm Developments in MDC Reconstruction Developer S. L. Zang Y. Zhang X. M. Zhang G. W. YU q Offline Algorithm Mdc Tracking Mdc. Pat. Rec Events filter Algorithm Mdc. Fast. Trk Mdc. Hough Presented by W. D. Li , Migrated from ATLAS

q Ø Ø Ø Purpose Efficient track finding Nice transverse momentum resolution High efficiency

q Ø Ø Ø Purpose Efficient track finding Nice transverse momentum resolution High efficiency of track finding at high noise level

Mdc. Hough Algorithm Cosθ=0. 83 Cosθ=0. 93 Interaction point ü 43 layers, 19 axial

Mdc. Hough Algorithm Cosθ=0. 83 Cosθ=0. 93 Interaction point ü 43 layers, 19 axial type ü|cos |<0. 93 üCell is near square ~8. 1 mm

Flow of Mdc. Hough Initial track finding Mdc. Hough Local maximum finding Hits PT

Flow of Mdc. Hough Initial track finding Mdc. Hough Local maximum finding Hits PT Track selection and Merging Track fitting

Flow of Mdc. Hough q Ø Ø Initial track finding (use a LUT-base Hough

Flow of Mdc. Hough q Ø Ø Initial track finding (use a LUT-base Hough Transform) (R, ) ( , 1/p. T) [ (0~2 ) p. T (400 Me. V~ )] q. CTR=sin ( – 0) CT= 0. 3/p. T Build a wire-ordered look-up table ( 1/p. T= 300 100) . wire n+1 wire n active wire n-1 wire. . Bin 1 Bin 2 … Bin 100 . wire-ordered LUT

Flow of Mdc. Hough q Local maximum finding (select good track candidates by wiredoreded

Flow of Mdc. Hough q Local maximum finding (select good track candidates by wiredoreded LUT) q Ø Ø Track selection and Merging Nhit > 15 Merge some tracks sharing more than 9 hits 15 15 16 16 18 17 13 14 14

Flow of Mdc. Hough q Ø Ø Track fitting Obtain hits from Bin-ordered LUT

Flow of Mdc. Hough q Ø Ø Track fitting Obtain hits from Bin-ordered LUT Fitting track to get PT by using lpav tool. . . bin n+1 bin n-1 number. . wire 1 wire 2 … wire 19 . bin-ordered LUT

Results and Discussion q q Track Reconstruction CPU Time ~ 1 ms/1 track Resolution

Results and Discussion q q Track Reconstruction CPU Time ~ 1 ms/1 track Resolution of PT(1. 0 Ge. V ) p=8. 0 Me. V q q Generate (PT : 1 Ge. V) by Fixpt Efficiency of Reconstruction ( ) VS cos (polar angular)

q Efficiency of Reconstruction VS PT( e p) q Momentum resolution VS PT (μ,

q Efficiency of Reconstruction VS PT( e p) q Momentum resolution VS PT (μ, e, π, p) Double Gauss Fit

Noise level type 0: = C type 1: q Efficiency vs noise 1/r type

Noise level type 0: = C type 1: q Efficiency vs noise 1/r type 2: 1/r 2 (PT: 1. 0 Ge. V ) q Resolution VS noise

Summary ü It costs about 1 ms to reconstruct 1 track ü Efficiency of

Summary ü It costs about 1 ms to reconstruct 1 track ü Efficiency of reconstruction( ) : v >99% (PT>300 Me. V) for single track v >99% when noise level are 5%, 10% , 15% and 20% v decrease quickly when polar angular more than 0. 8 ü Resolution of momentum( p): v PT < 1. 0 Ge. V p of proton is more than others v PT > 1. 0 Ge. V p keeps about same value for all particles v p turns bad at noise level is more than 10% in type “ 0” ü Same results by adding wires shift; Ø Further work is to enhance near polar angular and test the Algorithm in adjusted magnetic field

Thank !

Thank !