SLHC Muon Trigger Simulation s

1. SLHC Muon Trigger Simulations Vadim Khotilovich Alexei Safonov Texas A&M University CMS Upgrade Workshop, CERN, May 13-15, 2009

2. Outline • Goals and roadmap • Challenges of simulations at SLHC luminosity • Simulation framework with high pile-up • Current status and improvements • Implementation of detector and trigger upgrades in simulation • Current status and plans for Phases I and II • Infrastructure status: • Code status and plans, outstanding issues • Samples and documentation

3. Goals and Road Map • Framework capable of realistically simulating incredibly busy SLHC environment • Identify sources of deficiencies of the CMS detector and trigger at SLHC luminosities • Provides basis for designing electronics/hardware upgrades to address these deficiencies • Implement proposed upgrade ideas in simulation: • Optimize to maximize performance of the system and trigger, iterate as needed • Provide simulation framework describing performance of the future CMS detector and trigger at SLHC luminosities

4. Simulation: Challenges at High PU • High occupancies in SLHC environment cause memory exhaustion • Already at very moderate pile-up • More pronounced for full simulation • Simplifications in fast simulation make detailed muon simulations unreliable • Solutions will require substantial software work: • Short term fixes developed to allow necessary estimates • A long term solution still needed • Simulation misses effects that may become key at SLHC: • Combinatorial effects of neutron backgrounds • Scattering from forward detectors and beam halo

5. Neutron Backgrounds • Could be very important for SLHC: • Fairly high hit rates, handled by pre-triggers • However, collective effects may turn things around and make it the dominant problem • Neutron backgrounds do not come out “naturally” from simulation: • Long lived (up to 1s) and so come from crossings many BXs ago • Technical solution is understood, but requires work: • In ORCA neutrons were simulated (although only added in chambers with a real muon) • Improved treatment ( proposed by Rick): • Create a sample of simulated neutrons “compressed” into the 0-25ns window • Treat as an additional source to MixingModule • Really need to get it going, but missing qualified manpower

6. Signal: Pile-Up Events: Initial Sample Event Generator FilterModule FilterModule Stripped Sample Standard MixingModule Digi Simulation L1 Emulation Some Reco FullSim Solution for High PU • Understanding occupancy and BX related effects and backgrounds make Full Geant simulation the only viable option • Custom solution by TAMU: • Full simulation including out of time pile-up in a fraction of the detector • Drop unnecessary data to minimize memory problems and event size • Enabled studies of muon and calorimeter triggers • Stable and reliable for estimating efficiencies and single object rates • But still a temporary solution • More on it in twiki: SLHCFullSim

7. Efficiency for existing trigger/detector PGun Muons with Pileup 400 • Observations: • Efficiency drop:the cause is ME1/1b • Holes:due to missingME4/2 • No triggeringin ME1/1a • for more details about efficiencies studies see Alexei’s talk!

8. Next slides: Caveats and their Fixes • Found some inaccuracies in emulation • Being fixed (together with experts) • No effect at low luminosity, but significant for high luminosity • Many thanks to Slava, Jay and Darin for a lot of help in our understanding of CSC triggering!

9. Improved Hardware Emulation Example • Default emulator efficiencies in PU400 • Actual matching window used in the Trigger Mother Board is not implemented in emulation • A fix to implement proper matching CLCT-ALCT reveals substantial inefficiency in stub finding • -1 ≤DBX ≤+1, as it is in Hardware

10. CSC Track Finder Fixes • Implemented corrected CSC TF lookup tables for ME4/2 (thanks to Joe Gartner – U of Florida) • Better efficiency for high quality TF tracks # of stubs in TF track in PU400 events: • Fixed stub assignments in the event record: • A problem in track-stub association • Fix from Khristian Kotov (Florida) • Implemented in 31X After fix Before fix • For SLHC studies in 22X • Force track to use stubs only from the same BX • A shortcut, but a pretty good solution

11. Next Slide:Leading Cause of Inefficiency • Next slide illustrates origins of CLCT dead-time in TMB, leading cause of inefficiency

12. Where TMB Dead-time comes From? ALCT BX CLCT BX • Examples of BX distributions for ALCT and CLCT with default emulator • Pileup is simulated in -5 ≤ BX ≤ 3 • A CLCT in a previousBX precludes any other CLCTs in this chamber for 6 BXs: • CLCT “dead-time” per chamber • Early CLCT in PU are highly probable • Mismatched CLCTs lead to an arbitrary LCT or to no LCT at all SIGNAL NO PU MINBIAS PU400 SIGNAL with PU400

13. Phase I: Implemented and Planned Improvements • ME4/2 : new CSC station • Restoring Local TP reconstruction: • Addressing dead time • Other possible options • Restoring Triggering Beyond |h|>2.1

14. New Features: Station ME4/2 • Efficiency gaps for good quality TF tracks disappear with addition of ME4/2 • ME4/2 will be included by default in 31X • Back-porting to 22X took a considerable amount of effort • Thanks to the experts: Rick Wilkinson, Tim Cox, Oana Boeriu and Slava Valuev!

15. Addressing dead-time problem • Default trigger performance • Current TMB logic • “Eliminate” CLCT deadtime • “Brute force” implementation, but clearly it solves the problem • Actual solution requires new TMB firmware

16. Further Possible Improvements • Improving timing resolution • Especially for CLCTs • Do not allow highly bent CLCTs: • Will affect lower pT (below 10 GeV) triggering • But good suppression of rates and occupancies • Requires caution: • CSC readout is trigger driven • If no trigger, CFEB is not read out • May need to use pre-triggers to ensure readout of low pT muon hits CLCT Bend Pattern in 10<Pt<100 Signal (no PU) PU400 Minbias

17. Current ME1/1 Situation ME1/1 • In current hardware ME1/1 = ME1/a + ME1/b • Wire groups • Continuous 1/a+1/b coverage • Some (#11-16) cross the border • It complicates treating 1/a and 1/b as separate logical chambers • Strips: • ME1/a and 1/b have separate sets • ME1/a strips: • 48 strips ganged into sets of 3 = 16 channels • These 16 channels are sent as channels #65-80 of ME1/1 • Not safe for high L ME1/b ME1/a

18. ME1/1a Occupancies CLCT Strips • Strip and wire group ME1/1 occupancies: • Linear growth of occupancies with luminosity, but no shape change • ME1/1a stub rate is x1.5 that of ME1/1b • One of Slava’s recent improvements to the emulation ME1/b ME1/a ALCT Wire groups ME1/a ME1/b Wires Crossing the border

19. Restoring Trigger in |h|>2.1 • Several scenarios. One option is using ME1/a and ME1/b as separate “logical chambers” for triggering • Strips in ME1/a need to be un-ganged : 1-16 1-48 • What is ready: • Geometry: • A special switch disables strip ganging • Digitization and Conditions DB: • Special tables with unganged strip channels created (Oana) • CSCIndexer: a special modification (Tim) • CSCChannelTranslator: a special modification (Vadim) • Work in progress: • Local TP emulator modification (Vadim) • Partially working, implementing full solution • To do: • Track finder modification (Darin’s group)

20. Phase II Tracker Addition

21. Phase II: Adding Tracking Trigger • Important for studies of improving DT and CSC track finding with track trigger Pileup 200 in half-dedtector • It took a lot of effort to make Long Barrel geometry working with FullSim (thanks to Mike Weinberger!) • We incorporated the new track trigger into high PU simulation framework • Laura Fields helped with validation • Note: middle of upper layers still is not removed Example plots

22. Logistics • Sample planning • Code and documentation status

23. SLHC Muons Samples Planning • Phase I: pileup hits digitized only for muon systems • Muon-only samples are ready to be produced in bulk • Small samples available by request • Phase II: pileup hits digitized for track trigger and muons • High PU: slow, high memory and disk space consumption • CSC: up to half-detector with 400 PU • DT: can’t do full DT h range with high PU, bottleneck – tracker • When mid-layers in small eta will be removed in geometry, it may work • Separate samples for DT and CSC studies is the best option at the moment • TWiki: SLHCFullSim • Dave Newbold started using it for work on centralized production and storing of samples with different PU

24. Code Status • Adopted policy: • cvs branching for modifying existing files/packages • New functionalities not present now go in SLHCUpgradeSimulation or SimDataFormats/SLHC • Code for high PU FullSim Machinery (scripts and modules) • Available from SLHCUpgradeSimulation, documented in SLHCFullSim twiki • Includes track trigger and primitives, examples of configuration files for LongBarrel geometry • Modified CSC TP Emulator • A branch of L1Trigger/CSCTriggerPrimitives package named slhc_branch_2_2_X is created and maintained • Includes some incremental SLHC related changes • New code for ME1/a and ME1/b splitting will start larger scale modifications

25. Documentation • SLHC muon trigger simulations:https://twiki.cern.ch/twiki/bin/view/CMS/SLHCMuonTriggerSimulations • Twiki for high pileup machinery https://twiki.cern.ch/twiki/bin/view/CMS/SLHCFullSim • An example of handling cvs branching for SLHC development (thanks to Harry Cheung for help with this!) https://twiki.cern.ch/twiki/bin/view/CMS/SLHCMuonTriggerCVSBranching • Planned: SLHCCSCTriggerPrimitives twiki • Will include details on SLHC related modifications in L1Trigger/CSCTriggerPrimitives package • Will create it after committing functional ‘step 0’ for ME1/a

26. Conclusions • Major progress over the past year: • Functional full simulation machinery with high pileup • Major result is the discovery and identification of the shortcomings of the existing system • New features are being added to emulate future upgrades: • New station ME4/2 • Studying good leads to improve the trigger • Developing algorithms for high eta triggering (ME1/1a) • Recently incorporated new tracking trigger into the high luminosity full simulation framework • Paves the way for reliable Phase II muon upgrade design work • But many new challenges • Entering era of many code developers and users, coordination will become essential • Significant lack of dedicated (and qualified!) manpower • One glaring problem is lack of neutron background simulation, but more help is needed

27. Backup

28. Restoring Triggering in |h|>2.1 • I am modifying the local TP emulator • Added a special class and switches for TMB emulation in ME1/1 • Step 0: naïve ME1/a and ME1/b separation • Separate wires based on geometric location info available in MC digis • There’s some cheating for wires that cross the 1/a - 1/b boundary • But even the official emulator did naïve ME1/b treatment up until recently • Run two separate ALCT finders for ME1/a and ME1/b wires • Not very realistic, but it’s “step 0” • CLCT are found and correlated with ALCTs independently in 1/a and 1/b • Current status: close to being done • Might be useful for starting TF upgrade studies • Next planned steps: • More realistic ALCT treatment • Problem: ME11 has single ALCT board per chamber; is it possible to get more than two ALCTs out of it? • Intelligent classification of ALCTs while building correlated LCTs