| Run II Readiness "Trigger and Dataset Update" Rob Snihur and Joey Huston |
|
March 23, 2001
|
|
Summary: Dijet trigger: The main problem may be at L1. According to Mel, there are still a large number of single object triggers at L1 but no extra di-object triggers. Evidently the adjustment from single object to double object triggers can be done, but it's not very simple. There already exists a DT5,5 L1 trigger being used for the di-tau trigger, but this is restricted to the central region. This could be expanded to the entire detector but with a resultant increase in rate. Presumably the di-tau trigger people would not want an increase in the 5 GeV threshold, if that was required to extend the reach past the central region. There was some nervousness on the part of Mel and David about the L1 rate that the DT5,5 trigger would take, if extended past the central region. We need to approach the problem of the dijet trigger from a different angle. First, we need to identify the physics that we are aiming for. In this case, the physics is fairly specific: dijet cross sections with both jets on the same side of the detector, and on opposite sides of the detector, with a specific interest in those cases where there is a rapidity gap between the jets. These configurations are a small fraction of the total dijet cross section and so it may pay to design a specific trigger for them These configurations are a small fraction of the total dijet cross section and so it may pay to design a specific trigger for them with lower thresholds. Then we need to identify what ET range we are interested in, how many events we need for our analyses, how many events we would obtain from existing triggers and what we would gain from adding a new trigger. We should also state/estimate what trigger efficiencies we can live with; for example, if we are just interested in the percentage of dijet events with rapidity gaps, then we are not as concerned if the trigger efficiency is not close to 100%. So, for example, can we estimate the event statistics for dijet events (as a function of rapidity and ET) that we would get from the existing inclusive jet triggers? And what we would get from the proposed dijet trigger. This is not so different from the existing analysis that Christina is doing. Christina, can you provide these estimates? Currently, trigger towers 1 and 24 are not in the trigger and will not be placed in the trigger until the rates are understood. If the QCD group does not make a case for the addition of these towers, then they will not be added. The addition of these towers will be crucial for any dijet study with a rapidity gap between them. It is also crucial for any measurement of the inclusive jet cross section as a function of rapidity as well as a sum ET analysis. If the L1 rate for TT1, 24 is too high with a threshold of 10 GeV, then we can add an additional (higher) L1 threshold that includes these two trigger towers. Such a higher threshold trigger would include the central region as well (because of the way the trigger is set up) but this would have no impact since the ST10 trigger is not prescaled. Use of kT algorithm at L3: It is safe to say that this proposal was greeted with some skepticism. Again, we need to demonstrate the specific need for the use of the kT algorithm at L3. The largest component deciding the turn-on of any jet trigger, either cone or kT-based is the difference between the jet definition at L2 and the jet definition offline. The second largest component (my guess) is the use of Z=0 as the interaction point at L3. I would imagine that the turnon curve introduced by the use of a cone algorithm at L3 and a kT algorithm offline is smaller than the above 2 effects. This is something that we should check, however. Can we try looking at a sample of jets, simulating the L2 and the L3 triggers, and then reconstructing the jet ET's using the offline algorithm? We could examine the effect of the use of the kT algorithm at L3 versus the use of a cone algorithm at L3 as a function of the offline reconstructed ET. We could establish to what extent we would miss jets because we did not use the kT algorithm at L3, in the region where the trigger efficiency was close enough to 100% so that we could justify using those events. Would this entail running TRGSIM to obtain this information? One point that Kirsten made was that a separate trigger path would need to be established for an event depending on whether it was triggered with kT or cone at L3. We could write out the events to the same stream, so practically speaking I think this just means more bookeeping. I had asked Mel and David a question as to whether a detailed optimization of the L2 clustering had been performed. Initially the answer was no, but then David remembered that Xin Wu had given a presentation to the TDSWG in 1999. That presentation can be found at: http://www-cdf.fnal.gov/internal/upgrades/daq_trig/twg/level_2_jets_3_99.ps The main recommendation (now implemented) was to add EM+HAD towers. I think that we shouldn't go back to the TDSWG meeting until we have firmer answers to the above questions. How about if we revisit the situation on Friday to see if there has been any progress? |
| Sorry this speaker has not yet provided and electronic version of his talk. |
| Also see Rob's talk on March 9, 2001. |