You can find the intial version of this page, with Glauber and Fritiof simulations, and results and rejection factors using the old version of the muid_shield, here. All the results here are with the new muid_shield implemented.

The expected peak rate for dAu is about 92 kHz. The DAQ group's goal is to have a Lvl1-accept rate of 4.5 kHz and an archive rate of 1.2 kHz. This means that an overall rejection of about 20 at Lvl1 and 75 at Lvl1*Lvl2 is needed. The muon triggers can not take all the bandwidth and the estimated needed muon trigger rejection factors are about 50 at Lvl1 and 200 at Lvl2.

The collision-related rejection factors have been estimated using min. bias dAu Hijing events. With the new MUID shielding, the rejections factors at Lvl1 seems to be sufficient already with the Blue-Logic Trigger (BLT) for single deep muons.
There is good overlap between the single deep muon acceptance for the three available triggers (BLT, MUID LL1 and L2SingleMuon). The dimuon correlation is not as clear between BLT vs LL1 and L2.

A Lvl2 single deep muon provides just enough rejection to meet the goals at Lvl2, i.e. if the simulations describe reality and there will not be any significant non-collision-related background, and the quotient between the achievement/goal factors for the RHIC operators and the DAQ group will be below 1, we should not need to downscale single deep muons.

Mainly dAu simulations from S.C. Johnson et al. (LLNL). File locations
The list of ok/completed runs (126 runs with 1000 events each) can be found here.

There are 3 trigger acronyms in use throughout this page:

LL1 = (MUID) Local Level 1
BLT = Blue Logic Trigger
L2 = L2SingleMuonTrigger

The L2 code, with the Lvl2/software versions of LL1/BLT (the software is supposed to mimic the hardware performance), can be found in CVS under lvl2_distribution/algorithms/:
L2MuiLL1Trigger.C
L2MuiPseudoTrigger.C
L2MuSingleMuonTrigger.C

The L2MuiLL1Trigger is only available in the l2_run3_dev branch.

For the record, the BLT code used the mui_pseudotrigmap_deep_mh2.dat for the deep roads. This gives less rejection than the mh1 version.

For 126000 min. bias dAu events:

Auto correlations (per definition) are marked with an X in the table. Only half of the symmetric correlation matrix is shown. The format of the entries is Perc (MinFired), where Perc is the percentage of the time that the triggers overlapped and MinFired is the number of triggers for the trigger that fired/was accepted least, and is supposed to give you a sense of the statistical uncertainty.
As an example, take the overlap between 'D BLT South' and 'D LL1 South'. 'D BLT South' fired for 1403 events, 'D LL1 South' for 396 events and they were both accepted/fired in 391 events. The overlap frequency, Perc is thus 391/396 or about 99% and MinFired is 396.

For 126000 min. bias dAu events: North and South together

It's rather hard to digest the table with South and North mixed together, so I also give them separately.

For 126000 min. bias dAu events: South only

For 126000 min. bias dAu events: North only

Only the South arm was included. As a first test, a pure J/psi sample (1000 events with 2 muons within ~10-30 degrees) was processed. The rejection factors were all rounded off to 1, so I just list the acceptance numbers.

The correlations were all in the 98-100 % range.

Only the South arm was included. The same pure J/psi sample as above (1000 events with 2 muons within ~10-30 degrees) was processed, but now merged with one (the first ok file: 001) min. bias dAu Hijing file.

The correlations were all in the 98-100 % range. As you see, the numbers are close to identical to the ones obtained with J/Psi's alone. (For the North, there was a bigger relative difference, since the dAu Hijing files contained particles going into the North also, while all J/Psi's I used were headed South).

/Again, only the South arm was included, in this section./ A perfect trigger has a very high rejection _and_ and a very high efficiency. As long as the rejection is high enough to fulfill bandwidth goals, it makes sense to just chose the one with the highest efficiency. That would lead to chosing D BLT on LVL1 and D L2 in Lvl2.

For a comparison between the triggers, I introduce a quantity I call quality and define this as the product of the embedded J/Psi efficiency and the d-Au rejection, normalized to the value of S BLT. (S = shallow single). This is then a measure of well the triggers find the J/Psi's in the background.

As you can see, the dimuon versions are much more discriminatory than the singles, but one thing not shown in this table (however, it's in the one above) is that the dimuons versions do have a lower efficicency. BLT does well overall in this comparison to L2, and LL1.

Hits, for min. bias events: South.

Hits, for min. bias events: North.

The muTr hit numbers are divided by the number of gaps in each station, so the numbers are really supposed to be number of particles passing thru the stations per event.

A simple parametrization was done, to estimate the momentum from:
mom = Norm/(dphi23) * 1/(theta -t) + Const;
where dphi23 is the difference in phi between the track's hits in stations 2 and 3. Theta is the angle from the beam of the track's hits; it's very similar in stations 2 and 3.
When theta and dphi23 are in degrees, the constants can be chosen as t = 10., Norm = 76.5, Const = 1.14. The momentum error one makes (estimated by looking at simulated single muons and comparing the estimated and the original momenta is) can be seen, plotted as a profile, i.e. showing mean and rms, on momdiff.gif. The relative error: dp/p vs p, is plotted on momres.gif(2D box version).

The profile plot is actually somewhat deceiving and the dp/p turns out to be roughly 5%. The used files, were produced, without multiple scattering and with perfect position resolution. With multiple scattering the dp/p values increase to about 20%, which can be somewhat reduced if one uses the projections or stubs at stations 2 and 3.

To be able to reconstruct the mass of a pair, one also needs to estimate the original phi and theta angles of the tracks. A good approximation is to assume that theta is constant, since the bend occurs in the phi direction. Since theta determines the strength of the field and therefore magnitude of the bend, and the momentum also affects how much the particle bends, one would expect that the phi difference between the vertex and station 2 (dphivtx2, inversely proportional to the momentum) is proportional to the difference in phi between station 2 and 3 (dphi23) and proportional to theta. In phicorrtheta.gif, I have plotted the ratio between dphivtx2 and dphi23 vs theta and a clear correlation, that can be parametrized as p0 + p1*theta (p0 ~= -0.5, p1 ~= 0.09), is seen. Thus, knowing theta and dphi23, one can estimate the original phi.

Only the South arm was included, and the muTr primitives from Jason and Doug et al. were used when studying the same J/Psi sample as for the efficiency part above.

HIJING (dAu):
/phenix/data07/johnson/runs/

PISA (dAu):
/phenix/data38/phnxreco/PISA_Run03/simProject28/pisa_out

PRDF (dAu): 
/phenix/data38/phnxreco/dAu_simulation/simProject28/prdf/

L2TestFramework - libraries, macros etc:
/phenix/u/silvermy/lvl2_distribution/

Single muon files:
/phenix/workarea/zhangc/sandbox/mutoo/analysis/lvl2_trig/

pro.30 version + private libraries.