Speaker
Description
With ever-increasing luminosity at the LHC, optimum online data selection is getting more and more important. While in the case of some experiments (LHCb/ALICE) this task is being completely transferred to computer farms, the others - ATLAS/CMS - will not be able to do this in the medium-term future for technological, detector-related reasons. Therefore, these experiments pursue the complementary approach of migrating more of the offline and high-level trigger intelligence into the trigger electronics. The presentation will illustrate how the Level-1 Trigger of the CMS experiment and in particular its concluding stage, the so-called “Global Trigger”, take up this challenge.
Summary
While during the early years of LHC operation the Level-1 trigger selection was dominated by single-object triggers whose thresholds (in particular, thresholds on transverse momentum or transverse energy) were gradually raised to limit trigger rates to manageable levels, this approach would now result in significant loss of good physics data and is not viable any more. So, while silicon-tracker information will still be available only at the high-level trigger for a number of years to come, the Level-1 trigger relies increasingly on topological combinations of trigger objects from calorimeters and muon systems and on physics quantities such as invariant mass or transverse mass. This requires using particle flight directions correctly extrapolated to the vertex rather than raw detector data distorted by the bending of charged-particle tracks in the magnetic field. The enormous progress in digital electronics, in particular with regard to the computing power of Field-Programmable Gate Arrays (FPGAs) makes it possible to use input data of higher resolution and to perform complex calculations (using Digital Signal Processors (DSPs) as well as large lookup tables) for large numbers of candidate objects resulting in challenging combinatorics. The available resources also allow us to calculate in parallel certain quantities such as missing transverse energy in different ways (including or excluding the electromagnetic or hadronic calorimeter, varying the angular range with regard to the beam direction (pseudorapidity range) etc.) and use for each physics channel the most adequate version of this quantity. As jets, electron/photon candidates and “tau jets” (narrow jets from hadronic tau lepton decays) can be discriminated in the calorimeters only to a limited extent, attention must be paid to avoid double counting of such objects (“overlap removal”). By better reflecting high-level trigger resolution and algorithms in the Level-1 trigger electronics, data losses can be reduced and more useful physics events can be obtained within the available bandwidth.
