RUN 2B L1CAL MEETING MINUTES 12 December, 2002 Present o MSU: o Nevis: H.Evans, J.Mitrevski, J. Parsons o Northeastern: D.Wood o Saclay: J.Bisticky, D.Calvet Discussion of Peak Finding Feature ---------------------------------- o Attempt to address the issue of what information the ADF sends out for a TT in which it doesn't find a peak signal - Illustration of problem linked off agenda page - Briefly: in the current scheme the ADF sends out a nominal "zero" value for any TT in which the ADF's Peak Detector does not find a peak. . this leads to a bias toward positive noise fluctuations and decreases the effectiveness of isolation cuts for EM triggers and the resolution of missing-Et . for more information see discussion linked off of agenda - 2 Possible Solutions have been considered i) Send the raw output of the Digital Filter each BC as an 8-bit number and include a 9th bit indicating whether this is a peak or not ==> need to send 9-bit data from ADF to TAB ii) Continue using the current scheme - but send non-zero Et's only for those peaks that pass a small threshold. ==> note: the ADF already has a LUT after the peak-finder where this could be implemented o If noise has rms of 1 ADC count ==> very unlikely for noise to cause a peak o This also depends on transfer function - would have to simulate this to really understand the effect of noise on peak finding Jovan Mitrevski: Imposing Thresholds on TTs ---------------- o Affects mainly EM algo b/c of isolation cuts: both hadronic and em - but some study of jet algo also included o Noise Parameters used - noise = 0.25 GeV (1 ADC count) - this value was taken from an old L1Cal tsim RCP . also agrees with idea that LSB of ADC should correspond to the noise level o Signal(EM): WH --> ev bb 7.5 mbias Signal(Jet): ZH --> vv bb 7.5 mbias Bgrd: QCD 7.5 mbias o Study uses thresholds applied to individual TTs a) no threshold positive & negative Et's allowed what has been simulated up to now b) ET(tt) > 0 Gev like current peak-finder c) Et(tt) > 0.75 GeV sample threshold o EM Efficiency vs Et - drops at Et<20GeV for a) and b) o Isolation - c) produces less hadronic energy behind real electrons from W's o Eff vs Rate - compared c) with a) - appear to be rather similar - at 1 kHz rate eff(a)~0.80 eff(c)~0.83 at 0.1 kHz rate ~0.75 ~0.74 o Jet Algorithm - lower for c) than a) and b) . also rms(c) > rms(a,b) . E-frac gets worse for low Et jets for c) - Eff vs Rate . c) very similar to a) . loss of Et-fraction doesn't seem to have a big effect . needs more study o Remaining Questions - Is noise level simulated correctly? <== this is the main one - Can we treat BCs as independent? . correlation b/w 2 BCs is evident in trigger pickoffs at 396 ns > level = 15-20% - Is the gaussian noise scheme sufficient? - Check total-Et and missing-Et . should wait for p13 simulations for this: coming soon? - Understand noise model for analog chain. . this is complicated: see Denis' mail linked on the agenda page . LSB changes from 10-bit E to 8-bit Et . scale different depending on eta o Simulation Discussion - Note: Denis has a standalone TT-based noise simulation . only for single channels . correctly takes into account ADF-level noise . simple gaussian to model analog chain noise . no effects of previous BCs . digital filter coeff's not yet derived in final way . could conceivably compare these results to output of tsim . Atlas experience indicates that this will take a large amount of work. - Will we be able to make a reliable simulation in time to make decsions on 9- vs 8-bit sending? o 8- vs 9-bit - ADF: would have to change logic to run at BC*9 (from BC*8) . probably possible - but would take some work . this would distract from other efforts - Could conceivably send these bits in some of the "extra" words available on the Channel Link o Making the decision - only "9-bit" solution that we could make on the short time scale required to keep on schedule for prototype would be to fall back on sending "peak" bits in "spare" lines on ADF-to-TAB cable - BC*9 would require 2-3 weeks of work to implement . delays work on hardware ****************** * DECISION * ****************** o Baseline 8-bit data transfers with thresholds applied to TTS o Fallback send "peak" bits on "spare" cable pairs - need to study this some more o 9-bit xfer only if all else fails Status of Action Items from Saclay Workshop ------------------------------------------- 1) 10-bit TT raw E - working under the assumption that *no* 10-bit information will be sent from the ADFs to the TABs 2) Cabling the System - "patch panel" proposal from Denis would largely eliminate this worry - no reply from Dan yet - samples of required connectors (ERNI) on their way to Saclay 3) Parallel Commissioning - no news 4) Prototype Integration - no news 5) Trigger Terms from the GAB - no news 6) Cal-Track Match Data - Erich Varnes has started doing MC simulations for this at Arizona . no results yet - hopefully soon 7) Next Workshop - no news Any Other Business ------------------ o Connectors for cables from BLS patch panel to ADF - Denis has found a possibility: ERNI o Cables: ADF to TAB - Denis has found a shielded variety - 12-15 week leadtime - can they be ordered in the US? - should order cables for prototype test in Jan. . minimum order will probably be enough for final system o Analog Section of ADF - no comments received yet from Dan o Analog Splitter - final tests - output looks ok for wide range of signal ampl's - bandwidth: constant gain from 10kHz to >10MHz - Emmanuelle will bring splitter to Fermilab next week - put into D0 during Jan. shutdown o ADF PCB design - should be ready ~end-Jan - on track for integration in the summer o What is needed for Summer Integration Test? - ADF . will order 2nd crate after verification of 1st > ship directly to Fermilab: ~4 month lead time . backplane: will not really need this > something can be kludged here . Bit-3: will need another one at Fermilab > could bring the one from Saclay - TAB . probably will also not use final backplane in crate