F. Crescioli, P. Giannetti, M. Dell'Orso, G. Punzi, G. Volpi FTKSIM Updates F. Crescioli, P. Giannetti, M. Dell'Orso, G. Punzi, G. Volpi
New Sectors/Constants/Banks Sectors, constants and patterns generated only in the quadrant 0 region (0-90) Used two samples Old 68M full detector (no rotation) New 30M specific 0-90 New sectors are 120k with 98.2% coverage* (before we had 170k/4 with ~94%) * coverage is computed after cutting all sectors with less than 15 tracks
New Banks New bank is 8.2M patterns Coverage reached 82% (rotation+q0 were 88%)
Efficiency Efficiency is over 92% for muons and over iPatRec for all other particles. This kind of effect (better performance of FTK over iPat for pions, kaons and others) was somehow forecast in the resolution tail study: at low pT the resolution tails for pions were much lower in FTK than in iPat.
Efficiency summary table Sector coverage was an understimated problem. Optimizing sectors with the full sample produced a great increase in the overall efficiency.
Fakes Fakes are well under control, below 10%. Ghosts are almost gone. Fakes in the majority tracks are peaked on first layer missing tracks.
IP Resolution IP Resolution vs pT, as computed during constants building for 7/7 tracks. We still have a worst resolution than iPat, expecially at low pT. FTK iPat
IP Resolution (majority) IP Resolution vs pT, for majority tracks is quite stable if layer 2 to 7 is missing, but it's critically worst if layer 1 is missing. Tracks with layer 1 missing should not be used for IP-based algorithms.
Conclusions and the future New sectors/constants/banks have a much greater efficiency than old sectors + rotation trick banks Sectors importance was understimated Rotation trick mostly added noise patterns Fake distribution and IP resolution vs pT shows that first layer missing tracks should be avoided With the next 30M tracks from Erik we could reach iPat in muon efficiency on quadrant 0 We could stop quadrant 0 only production and resume whole phi production for the 8 banks configuration