Triggering at High Luminosity at CDF

1. Triggering at High Luminosity at CDF Tom Wright University of Michigan Fermilab All-Experimenters’ Meeting August 3, 2009

2. The CDF Trigger System • Three-level system • Synchronous Level 1 • 100% hardware • Guaranteed latency • Asynchronous Level 2 • Hardware/software hybrid • Limited buffer space • Can incur deadtime • Input from L1 too high • Output too high for DAQ • Level 3 • Fast offline reconstruction • No problem with capacity so far

3. Trigger Objects • Level 1 & 2 can use almost all subsystems of CDF • Tracking (fast lookup table) • Calorimetry (electrons, photons, jets, missing ET) • Muon stubs / scintillators • Silicon tracking (Level 2 only) • Level 1 has simple matching (muons, electrons) and can do coincidences • Level 2 allows more complex matching and correlations on the same inputs (plus silicon tracks) • All data delivered to a PC, can implemented any desired algorithm in software JDL 11/15/07 PJW 9/23/98

4. CDF Trigger Menus • Goal is to keep ‘signal’ triggers unprescaled • High-pT inclusive leptons, photons, missing ET, high-ET jet, multijet, etc • At lower luminosity, start enabling more triggers • Medium-pT leptons, b-physics leptons, etc • Looser versions of signal triggers • At lowest luminosity, switch on hadronic b-physics triggers • Variety of backup triggers on fixed prescales (jets, tracks, etc) • Low-bias samples for measuring ID fake rates / efficiencies • Monitor and control Level 1 and Level 2 rates to manage deadtime • Target is 5% • Trade-off between high efficiency and high utilization (events on tape)

5. Dynamic Prescaling (DPS) • Main tool for managing rates • DPS triggers start at a specified maximum prescale to keep rate low • As luminosity falls, prescales are reduced as bandwidth becomes available • Prescale adjustments based on feedback from total L1/L2 rates • Prioritization based on maximum prescale and individual rates • Lowest-rate trigger prescales are adjusted first

6. Trigger Rates Store 6832 Feb 2009 L = 335E30 hadronic b-physics low-pT dimuons bit over our target rate changing DPS to maintain rate

7. Total Deadtime Stick pretty close to our goal up to ~310E30 Somewhat higher above that (from high Level 2 output rate)

8. Data-Taking Efficiency average of store efficiencies versus store initial luminosity Amount of luminosity we see above 310E30 small (so far) Initial deadtime doesn’t have large impact on our efficiency

9. Trigger Menu Evolution • A sample of menus from the past year • Tighten / prescale at high luminosity to keep up with the Tevatron • Haven’t had to sacrifice a lot so far, trimmed backups or added new requirements with high efficiency • Continue this process if/when Tevatron performance improves further

10. Conclusion • The CDF trigger/DAQ system has worked well as luminosity increases • High-priority signal triggers have remained unprescaled • Triggers on dynamic prescale seeing 50-90% of available integrated luminosity • Continue to collect sufficient backup triggers • Gradual improvements to trigger menu have allowed us to keep up • Headroom exists to continue this process if necessary, without large impact on our core high-pT physics program • Increasing luminosity also impacts b-physics program, particularly hadronic triggers • Spending much less time at low luminosities where those triggers are active • Yields have dropped by ~25% for B mm, ~80% for Bhh over past few years • Focus of most of the recent work on the trigger menu

11. Backup

12. High-Rate Triggers • Most difficult is ‘CMX’ muons • 0.6 < |h| < 1.0 • Have had this on DPS in the past • Next is missing ET plus dijet • Higgs/diboson trigger • Should be able to keep these and other signal triggers live if luminosity increases further

13. Sample Store Profile