Level1 Trigger Menus ClaudiaElisabeth Wulz Institute for High
Level-1 Trigger Menus Claudia-Elisabeth Wulz Institute for High Energy Physics Vienna CMS Week Mumbai, Dec. 2000
Introduction Definition of Trigger Menu: Set of algorithms running concurrently in the Global Trigger. There may be different sets for different run conditions (Bphysics at low luminosity, heavy ion runs, discovery physics at high luminosity, calibration etc. ). Run control must record the used menu. Remember: If an event does not pass Level-1, it is gone forever and will never make it to a physics publication! Claudia-Elisabeth Wulz 2 Mumbai, Dec. 2000
Examples of Trigger Conditions Trigger Examples of explorable physics channels 1 m 2 m m+e/g m+jet(s) m+ETm 1 e/g 2 jets e/g+jet(s) m+t e/g+t t+jets+ETm HSM, H, A, H±, W, W’, t, B-physics channels HSM, h, H, A, Z, Z’, V, , LQ, Bs 0 ->2 m, , ’’ HSM, H, A, t, WW, WZ, Wg, , V HSM, h, H, A, , LQ, t t, , LQ, WW, WZ, Wg HSM, h, H, A, W, W’, t, B-physics channels HSM, h, H, A, Z, Z’, WW, WZ, Wg, , LQ QCD HSM, h, H, A, , LQ, QCD (g +jets, W+jets) HSM, H, A, H± Claudia-Elisabeth Wulz 3 Mumbai, Dec. 2000
L 1 Menu Working Group Established during Tri. Das Week 9 Nov. 2000. Every interested person is invited to join and to provide his or her ideas! Presentations at initial meeting: • • • Introduction Global Trigger overview Calorimeter Trigger overview Muon Trigger overview Trigger menu requirements from physics point of view • Trigger menu requirements from HLT point of view • Trigger menu requirements from DAQ point of view Claudia-Elisabeth Wulz 4 W. Smith C. -E. Wulz S. Dasu G. Wrochna M. Dittmar P. Sphicas S. Cittolin Mumbai, Dec. 2000
L 1 Menu Working Group Tasks l Provide initial trigger menus to capture the interesting physics. Menus for calibration etc. should also be established. Menus should not be considered fixed once and for all, but will evolve with experience gained. The flexibility and special features of the CMS L 1 trigger should be optimally used. l Check that trigger design is capable of handling all physics and technical requirements. l Provide corresponding suitable trigger parameters (at level of global trigger and at regional and perhaps local levels). l Allocate suitable bandwidths for categories of algorithms. Claudia-Elisabeth Wulz 5 Mumbai, Dec. 2000
CMS Level-1 Trigger Claudia-Elisabeth Wulz 6 Mumbai, Dec. 2000
Basic Principles of the L 1 Trigger For most other comparable experiments the trigger is based on counting objects exceeding thresholds. Only summary information is available. This implies applying thresholds at local or regional levels. In CMS, only the Global Trigger takes decisions, i. e. no cuts (except inherent thresholds for defining a jet, isolation criteria etc. ) are applied by lower level trigger systems. The trigger decision is based on detailed information about a trigger object, which includes not only p. T or ET, but also location. For muons, quality information and charge are also available. This enables selecting specific event topologies. The objects are ordered by rank. An algorithm is a combination of trigger objects satisfying defined threshold, topology and quality conditions. There are 128 trigger algorithms running in parallel. The resulting bits are available in the trigger data record. The Global Trigger runs dead-time free by principle, i. e. a L 1 Accept/Reject decision is issued with every bunch crossing. The Trigger Throttle System may, however, inhibit a L 1 A in case of e. g. buffer overflow warning. For each algorithm a rate counter and a programmable prescale factor (up to 16 bits) are available. The L 1 decision is taken by a Final OR of which up to 8 are available for physics. Claudia-Elisabeth Wulz 7 Mumbai, Dec. 2000
Global Trigger For physics running the Global Trigger uses only input from the calorimeters and the muon system. Trigger specific sub-detector data are used. The high resolution data are used by the Higher Level Triggers. Apart from the trigger data, special signals from all sub-systems may be used for calibration, synchronization and testing purposes (technical triggers). The TTC System is an optical distribution tree that is used for the transfer of the Level-1 Accept signal and timing information (LHC clock etc. ) between the trigger and the detector front-ends. The Trigger Control System controls the delivery of L 1 A signals and issues bunch crossing zero and bunch counter reset commands. There is a facility to throttle the trigger rate in case of buffers approaching overflow conditions. The Event Manager controls the Higher Level Triggers and the Data Acquisition. Global Trigger Environment Claudia-Elisabeth Wulz 8 Mumbai, Dec. 2000
Input to Global Trigger Best 4 isolated electrons/photons Best 4 non-isolated electrons/photons Best 4 central jets (|h| £ 3) Best 4 forward jets (3 < |h| < 5) Best 4 t- jets Total ET Missing ET 6 jet counts (central jets) 2 jet counts (forward jets) Best 4 muons ET, h, f ET, h, f SET ETmiss, f(ETmiss) p. T, sign, f, h, quality, MIP, ISO 4 inputs (approximately 100 bits) are still free. Claudia-Elisabeth Wulz 9 Mumbai, Dec. 2000
Features and Flexibility of Global Trigger The Global Trigger logic is largely programmable. Particle energy or momentum thresholds and h (or f) windows can be set separately for each object. Different thresholds for central and forward regions are therefore possible. Templates for muon quality, including MIP, isolation and charge information can be selected. Space correlations are possible between all objects, but restricted to “close” and “opposite/far”. Jets are actually separated into central and forward jets. There also 8 jet multiplicities, 2 of which are reserved for the forward jets. Claudia-Elisabeth Wulz 10 Mumbai, Dec. 2000
Basic Trigger Setup In the stable phase of the experiment the trigger is set up via Run Control using predefined menus which include reasonable thresholds for different luminosities. These thresholds may be changed by the physicist, without reconfiguring the logic chips. Most of the 128 algorithms are available for physics running. The basic rule is to keep the trigger menus as simple as possible. If not all interesting physics processes can be caught with these, more sophisticated logic may be used, but careful studies of trigger efficiencies have to be made. If a new algorithm (i. e. one not already present on the chips) becomes necessary, the chips can be reprogrammed by experts. The timescale for this is a few hours, but it should not happen too often. Claudia-Elisabeth Wulz 11 Mumbai, Dec. 2000
Predefined Algorithms Claudia-Elisabeth Wulz 12 Mumbai, Dec. 2000
2 Forward Jets in opposite h-hemispheres Claudia-Elisabeth Wulz 13 Mumbai, Dec. 2000
4 muons with template conditions Claudia-Elisabeth Wulz 14 Mumbai, Dec. 2000
2 muons with space and charge correlations Claudia-Elisabeth Wulz 15 Mumbai, Dec. 2000
Features and Flexibility of Calorimeter Trigger Primitives Fine grain veto: max ET in h-strip pair vs total trigger tower ET Trigger Towers Separate ET cutoffs for e/g and t/jet/ET triggers H/E veto: ECAL vs HCAL ET ratio, can be non-linear Active tower definition: programmable ET cut to adjust for pileup 4 x 4 trigger tower region level for jets: ET cut for pileup suppression, cut on active tower count for t veto t /jet candidate level: h-dependent center region threshold and ET lookup Possible additional Global Calorimeter Trigger algorithms: SET of jets, missing ET of jets Claudia-Elisabeth Wulz 16 Mumbai, Dec. 2000
Features and Flexibility of Muon Trigger DT, CSC, RPC p. T scale of all 3 systems (DT, CSC, RPC) programmable, can in principle be different for all 3, but Global Muon Trigger must convert to a common scale. Global Muon Trigger Implementation of the matching scheme, many tunable parameters DT, CSC TRACO: LUTS for correlation of BTIs, filters for ghost suppression Track Finder: Extrapolation windows, assignment LUTs, filters for ghosts (also in Global Muon Sorter) RPC patterns, gate (noise suppression) Claudia-Elisabeth Wulz 17 Mumbai, Dec. 2000
HLT, DAQ and L 1 interplay Check decision-making and filtering of Level-1 Disentangle detector and trigger malfunctions, monitor rejected events Check seed-generator function of Level-1 Regional reconstruction of HLT depends on what L 1 sends! Need to reconstruct also objects that did not fire the L 1 trigger. Storage of L 1 Trigger parameters Need database, pointer to it is major element of “run number”. Event is not simply identified by run and event number, but by data structure containing run conditions. Run = time between fill start and end. Minimum Bias events Need full reconstruction. Example of use: check events that systematically fail e/g trigger, but fire the jet trigger. Claudia-Elisabeth Wulz 18 Mumbai, Dec. 2000
HLT, DAQ and L 1 interplay Importance of prescaling Automatic or fixed. For automatic prescaling need good and traceable luminosity measurement! Check trigger efficiencies at lower thresholds than in main trigger menu, flag events for calibration. Dynamic changes of trigger menu Relax thresholds and optionally change algorithms as luminosity drops. Trigger type Option in the Global Trigger, no strong demand yet. . . Trigger menu recording Should be in database accessible both by the Global Trigger Processor and the HLT farm. DAQ issues Local event filter rate 1000 Gbit/s, event storage 5 Gbit/s. Recording rate can be greater than 100 Hz. Claudia-Elisabeth Wulz 19 Mumbai, Dec. 2000
Physics considerations Hardware Trigger(< 100 k. Hz) raw calibration subdetectors with coarse detector segmentation single (isolated) objects with p T cuts multiple object triggers h-f correlations satisfy physics needs with hardware HLT (< 100 Hz) almost final calibration almost full detector fine segmentation and combination of tracks calo and m system verify trigger object(s) object matching with tracks mass of clusters + tracks satisfy physics requirements with software Analysis (10 6 -7 events? ) best calibration full detector available signal optimization accurate track matching precision mass calculations complicated h, f and p. T selection criteria be "undisturbed" by trigger conditions Trigger cuts should be softer than physics selection criteria, but some rates will be too high! Need compromise. High Q 2: Exciting, possibly exotic physics Medium Q 2 and low x physics: New domain of strong interactions Low Q 2: b and c physics with unprecedented statistics Claudia-Elisabeth Wulz 20 Mumbai, Dec. 2000
Physics considerations Physicists have many different points of view. Examples: can(not) be measured Rapidity gap events are interesting/boring Physicists want redundancy. Example: high mass Drell-Yan lepton pairs After discovery physicists want to explore. Need to study more difficult signatures. Claudia-Elisabeth Wulz 21 Mumbai, Dec. 2000
Statistics considerations 107 events/day at rate of 100 Hz Accuracy for cross-section measurements: ± 1% -> 105 accepted signal events -> ± 0. 3% statistical error Claudia-Elisabeth Wulz 22 Mumbai, Dec. 2000
Large and low cross section measurements Low cross section • Important not to lose any event Example: Bso -> m+ m - (BR @ 3. 5 x 10 -9) • After discovery can accept worse S/B ratio for BR measurements Large cross section (s x BR > 100 nb) • Prescaling • Accept 1 Hz rate for a few days and make analysis • Use luminosity lifetime: use free rate near end of fill • Combination of all three above Claudia-Elisabeth Wulz 23 Mumbai, Dec. 2000
Simulation studies Physics goals • Observation, evidence, exclusion • Cross section measurement • BR measurement • Measurement of trigger effiency (control channels, lower thresholds, etc. ), background Simulation studies to be done • SM Higgs, SUSY, higher dimensions, exotica, … • b and t physics • QCD and other SM physics • New signatures as we go along Claudia-Elisabeth Wulz 24 Mumbai, Dec. 2000
Conclusions Baseline trigger design suitable and flexible enough for most imagined signatures Reasonable thresholds can be set Redundancy for high Q 2 processes Rates for most triggers ok Rates for many single object triggers relatively high, multiobject triggers should be made use of Topology and quality options available Need scenarios for use Please join L 1 Menu Working Group! Claudia-Elisabeth Wulz 25 Mumbai, Dec. 2000
- Slides: 25