svxonly am bank performances for forward/backward tracks
DESCRIPTION
SVXonly AM bank performances for forward/backward tracks. Laura. Outline. Two pattern banks tested with SVT simulation : Quality of the roads => phi0 resolution . This can be done matching the SVX-only roads with Level3 tracks - PowerPoint PPT PresentationTRANSCRIPT
SVXonly AM bank performances
for forward/backward tracks
Laura
Outline
• Two pattern banks tested with SVT simulation:
– Quality of the roads =>phi0 resolution. This can be done matching the SVX-only roads with Level3 tracks
– Number of roads coming out the AM: we used 2 dataset/trigger selections to estimate the latency.
– Bank Coverage
We tested different configurations:a) 128 kpattern road size (SuperStrip size 64457),
IP<500 um, Pt>10 GeV ,1 < || < 1.8 100 Kpatt => 95% Coverage
AM bank study (128 kpattern available)
b) 512 kpattern road size (ss size 43334), IP<500 um, Pt>10 GeV, 1 < || < 2.5
400 Kpatt => 95% Coverage Too large bank
c) 512 kpattern road size (SuperStrip size 43334), IP<300 um, Pt>20 GeV , 1.2 < || < 1.5
128 Kpatt => 95% Coverage
Two different Trigger:1) JPSI_CMUCMU2_PS2 (run 211554):
- used to test the larger roads (128kpattern road size: ss
64457) - Soft events
2) MET_L2S1K_L1_MET15_&_JET5_&_TWO_TRK2(from run 211200 to run 211877) :
- used to test the thinner roads (512kpattern road size: ss 43334)
- Harder events, no level2-level3 requests
Trigger selections used to evaluate system
performances
matching
• A road corresponds to a certain set of tracks that have similar , Pt and impact parameter
• The roads are now really thin• The rms of track’s phi in the road can be enough
small to allow association between the road and the muon/cal.
• For the study we use both 0 ( measured @ SVX) and 1 (0 extrapolated to the BMU radius).
BMU
road
0
1
1 = 0 + asin(cur x R_BMU)
cur=track curvatureR_BMU=radius of
BMU
information 6 electrical barrels
Z
Two SVT track types: 1. Tracks crossing at least 2 barrels.
We know the track => flag=0(Forward), 1 (Backward)
2. Tracks contained in asingle barrel; we don’t know => flag=2
=> It reduces the fakes.
Quality of the road: 0L3- 0SVT
Effect of the eta Flag Resolution = 5.6 10-3 rad
(large road):
Resolution =3.5 x 10-3
(thin road):
The required resolution ~ 2.5 degrees ~40x10-3 rad
Without Flag Info
With Flag Info
<Max number of road> per wedge
Thin roads and smaller bank reduce the number of roads coming out the AM, even if we used a more energetic trigger!!!
Latency for the extraction of roads from AM is about 2 clock cycle for each road @ 40 MhZ.100-120 roads -> 5-6 us
Large road with JPSI_CMUCMU2_PS2
Thin road with MET trigger
Without RW
With RW
Integral distribution over all luminosities of the #of roads per event found in the most crowded wedge.
Max number of road per wedge
10-5
10-4
Thin Roads a factor 10 less events @200 roads
Finally:a) 128 kpattern road size (ss size 64457),
IP<500 um, Pt>10 GeV, 1 < || < 1.8 100 Kpatt => 95% Coverage Larger roads TOO Many ROADS
TOO large latency
AM bank study
c) 512 kpattern road size (ss size 43334), IP<300 um, Pt>20 GeV 1.2 < || < 1.5
128 Kpatt => 95% Coverage Number of roads reduced ~ a factor 2 on
MET sample Quality of roads improved and very good!
Eta Coverage
The bank is efficient also in central region. => High probability to have central tracks sharing SVX ss with forw/backw tracks. Fakes due to central tracks will be removed by RW!!!
Pt Coverage
The bank is also efficient for low Pt, but the efficiency of the match in phi0 with lepton will favor the high Pt tracks.
IP CoverageIP<300 um
[um]
Conclusion
Next:- Matching efficiency- A further reduction of # roads will be
provided removing all the central tracks faking forward roads
- level 2 rates evaluation
=> Work on progress !!!!