1 / 23

TT in Trigger Part III

TT in Trigger Part III. The Issues. Momentum (p T ) resolution B∙d l Pattern Recognition Efficiency (want 100%) vs Ghosts (want 0%) Low occupancy Timing Quick decisions Reject low momentum tracks (say p<3 GeV) Select high pT tracks (say p T >1.25 GeV)

brasen
Download Presentation

TT in Trigger Part III

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. TT in TriggerPart III

  2. The Issues • Momentum (pT) resolution • B∙dl • Pattern Recognition • Efficiency (want 100%) vs Ghosts (want 0%) • Low occupancy • Timing • Quick decisions • Reject low momentum tracks (say p<3 GeV) • Select high pT tracks (say pT>1.25 GeV) • Optimal pitch? Smaller pitch near beamline? Detection element size? • Material between VELO and TT • Correlation with calorimeter high energy shower deposits?

  3. TT Detector Review • Four layers of silicon • X1, U(+5o), V(-5o), X2 • ~200 um pitch • Initially intended for “L1” trigger • Idea abandoned (efficiency, ghosts, resolution) • Instead use CALO clusters & large IP Velo tracksas seed to do local PR in T-stations. • More efficient, better p determination, lowerghosts (presumably more CPU intensive though)

  4. TT Layers TTa – X1 TTa – X2 TTa – U • Sensor (LxW): 110 mm x 780 mm • Regions 1,2 and 3 have stripsof length 1L, 2L and 3L, resp. • 330 mm length at large Y • Minimal Y information 3 readout sectors 4 readout sectors

  5. Field in Velo-TT Region Z of TT stations: 232  265 m Bymax ~ 0.15 T Not very large

  6. VeloTT Tracking Strategy X projection • Search window 1/pz • Different tune for online (not used anymore) and offline • In offline, we are mainly interested in low momentum tracks • Additional tolerance window – hits in TT should be nearly on a line (could be many candidate tracks within the search window)

  7. VeloTT TrackingY Search Window 33 cm long strips 22 cm long strips 11 cm long strips Large Y window necessary A X,U,V hit by itself carries no Y information

  8. Performance at 2x1032 cm-2 s-1 4 GeV 2 GeV 1 GeV 0.5 GeV Efficiency Ghosts • P>4 GeV: • Eff ~ 90%, Ghosts ~10% • Not great if to be used in trigger • Rapid increase in ghosts asmomentum is reduced • Large X search window req.

  9. But, on the good side …Momentum Resolution Does improvemomentum resolution(at least in MC) Dp/p TT hits not used in track fit TT hits used in track fit p (GeV) M. Needham 2007-144

  10. pT resolution Using only straight linefits to hits in VELO andhits in TT (Was meant to be fast for (now defunct) L1 trigger) Trigger TDR • For tracks we’d like to trigger on, s(pT)/pT ~ 20% - 35% • Improvement possible with Kalman fit, but here you’d wantaccess to the detector material description (costly in CPU) • See next slide

  11. Y. XieLHCb 2003-100 Offline Eff What room is there to be gained, if CPU were not an issue pT(GeV) Ghost rate ~20% pT resolution in theregion we care about… compared To ~25% for no Kalman (see previous slide) pT(GeV)

  12. pT Resolution LHCb 2003-110, H. Dijkstra, et al pT(true)=1 GeV sp/p = 30% Large fraction of Trackswhere pT is significantlyoverestimated

  13. The backgrounds -- MinBias 2e32 • Use MCHits after full simulation. No Pattern Recognition • Look at DX between linear projection of Velo MCHit and all TT hits within a given Y search window. P range (X range for S/B): P<5 GeV: (-50.0, 5.0) 5<P<20 GeV (-10.0, 2.0) P>20 GeV (-2.5, 0.5) |DY|  5 mm(see Slide 17) S/B~23 S/B~98 S/B~16 Hits within readoutsector pointed to by track. If close to boundary,also take a secondreadout sector. Meant to mimic currentVeloTT algorithm. Tolerance window willimprove S/B, but still haveto cope with combinatorics S/B~0.4 S/B~4 S/B~0.6

  14. The backgrounds -- MinBias 2e33 P range (X range for S/B): P<5 GeV: (-50.0, 5.0) 5<P<20 GeV (-10.0, 2.0) P>20 GeV (-2.5, 0.5) S/B~9.3 S/B~44 S/B~3.8 S/B~1.6 S/B~1.4 S/B~0.15

  15. By Region – 2e33 Region 2 Region 3 Region 1 P<5 GeV 5<P<20 GeV P>20 GeV

  16. Looking by Occupancies – 2e33 • To translate to occupancy, • Divide by: • 100 events (n1=100) • 1 mm bin  0.198 mm strip • n2 = 5 All 4 planes grouped (n3=4) So, overall, divide by 2000. This is an event average. Clearly significant event-by-event variations, as the #intis Poisson distributed. Reasonably good agreementwith M. Needham’s talk atupgrade mtg, Jan 12, 2007. 10% Hits/mm Region 1 5% Hits/mm Region 2 3.5% Hits/mm Region 3

  17. BsDsK Signal – 2e32 P range (X range for S/B): P<5 GeV: (-50.0, 5.0) 5<P<20 GeV (-10.0, 2.0) P>20 GeV (-2.5, 0.5) Roughly S/B a factorof two worse than MinBias.(B events have~twice as manytracks/event) Low stats inbackground forp>5 GeV (DY<5 mm) S/B~22 S/B~136 S/B~9 S/B~0.8 S/B~2.6 S/B~0.3

  18. DY Distributions Using only TT Hits associated with track P<5 GeV 5<P<20 GeV P>20 GeV

  19. DX due to By and Multiple Scattering MS - back of envelope B deflection from observed deflectionsin MC. At 1 GeV: Field deflection ~ 23 mm s(DX)MS ~7 mm At 3 GeV: Field deflection ~ 7.5 mm s(DX)MS ~2.5 mm At 5 GeV: Field deflection ~ 4.5 mm s(DX)MS ~1.4 mm For MS, multiply by. From BOVExpected ~ 1.5 mm Both scale as ~1/p Scale MS expectation by 1/sqrt(2) for projection onto X or Y axis s(1/P)/(1/P) s(DX)/DX ~ 20% just from MS (maybe slightly overestimated)

  20. Improving pT Resolution via B∙dl Current LHCb Field (about ~0.7xField043) 0.2X 0.4X 0.6X 0.8X LHCb 2003-110, H. Dijkstra, et al 1X nominal B∙dl 1.2X 1.5X 2.0X 3.0X 5.0X Can gain in pT resolution with finer granularity as well

  21. Summary/Remarks • Increasing B∙dl sounds good, but will require a larger search window in X • Need to keep Y window “small” • Finer segmentation • Clearly, need better Y information ( strixel-TT ?) • Reducing pitch will improve p resolution, up to the point that MS in new TT starts to dominate. • E.g. 5 mm length x 100 um pitch ~2 million strixels/m2. (Current TT at ~1.5 m2) • Would need to look at what granularity is needed. • Should try to reduce material between VELO and TT if at all possible. • Can (upgraded) RICH1 help? • Toss low p tracks below threshold, confirmation… • Remove it – that’ll help also !

  22. Backup

  23. Previous Look from Matt

More Related