Talks

• Tau ID: Yann Coadou
  Selection statistics were presented for the data collected since p11.01.00 was installed online. Various strange inconsistencies were found between the online statistics (as given by the L3 monitor) and those Yann obtained offline. The online statistics suggest that the tau filter is behaving approximately as expected. However, it is clearly very worrying that we cannot get consistent results offline. Subsequent to the meeting Yann sat down with Suyong and resolved some of the inconsistencies (which were traced to a bug in vertex_examine), but work is still in progress.
• Shower Shape for Electron ID: Ulla Blumenschein
  Following on from previous presentations, Ulla showed the results of shower shape in real data. Conclusion, we should get this into p11.02.00 as a "bug fix" and start cutting on it. Getting a realistic idea of true rejection fraction for EM triggers was not possible, because Ulla was also having difficulty decoding the trigger words in the data!
• Electron ID and JET ID results on data collected since p11.01.00 was installed online were not shown, because Volker Buescher (who was at the physics/ID convenors' meeting) had been having similar problems to Yann.
• Muon ID: Prospects for p11 certification: Martijn Mulders and Martin Wegner
  We know that the p11.01.00 version of L3FMuon suffers from a number of problems:
  • Various bugs in the track finding have been found by Martin. N.B. These affect also the "certified" offline muon id.
  • There is a memory leak at the level of about 1 kB per event in the offline code that has been know about for months, but is not yet fixed.
  • The unpacker requires a fixed configuration (cfg.dat) file, which render it sensitive to changes in the muon readout configuration other than at the level of entire crates being removed from the readout. This was identified as an urgent problem last November, but there has been no visible progress since then.
  Inspite of these problems, there was a strong consensus (although not unanimity) within the L3 algorithms group that we should make all possible efforts to get the p11.01.00 version of L3FMuon certified for use online. In this context, "certification" means to a large extent making sure that L3FMuon would not screw up the already running calorimeter-based filtering. The consequences of the fixed configuration file and possible strategies for dealing with instabilityin the muon hardware need careful discussion with the muon hardware experts. Initially the plan would be to run L3FMuon in mark&pass mode. However, the hope would be to run in active filter mode (exclusively on single muon triggers in the first instance) as soon as possible.
  N.B.
  • The currently "efficiency" for single muon triggers is around 2% (since the prescale at high luminosity for mu1ptxatxx is 50).
  • Anyone who does not want to be affected by single muons filtered with an imperfectly understood L3FMuon if/when it runs in active filtering mode, merely have to select unbiassed mark&passed events on the basis of the L3 trigger bits. (We need to make it easy for people to to this.)
• CFT Tracking (and Unpacking): Robert Illingworth
  Testing of the CFT tracking for p11 has been progressing. It was agreed that Robert's first priority should be to get the CFT unpacker to use variable thresholds and the most recent offline cable maps.
• Global Tracking: Daniel Whiteson
  The global tracking has been very stable and the code is in excellent technical shape. The performace is very similar to that obtained by the standard offline tracking, in that it suffers from the known deficiencies of the detector: hardware, readout and alignment. That is, the number of tracks found is much less than in Monte Carlo, the phi distributions are far from flat, the matching efficiency between SMT and CFT is much lower than it should be. However, once Robert has put in the improvements for the CFT unpacker, we should be trying to get the tracking running online as soon as possible!
• Primary Vertex: Help is needed to test the performance of the tool with up-to-date release and data samples. Dan will contact Guilherme Lima, in particular with regard to incorporation of L3TPrVtx into the l3fanalyze framework.
• Secondary Vertex: We should take advantage of the presence of Robert Illingworth and Arnaud Duperrin to discuss plans for vertexing at IC and Marseille.
• Missing ET: Lee Sawyer will ask Diptansu Das to give a status report at the next meeting.
• Discussion on L3 monitoring:
  The L3 Monitor provides a very useful summary of L3 filtering rates and a way to spot major problems. (We have asked that this information is printed out two hours after the start of a new fill and at the end of every run. The DAQ shift instructions have been modified accordingly.) However, we need more than this. In the short term we need to overcome the problems we have with our offline decoding of the trigger information and get much faster turn-around with our offline checks of performance. In the near future, the l3fanalyze program needs to be run online as an "examine". There are a number of questions/options:
  • The ideal solution would be to use the ROOT GUI to make histograms from the rootuple already produced by l3fanalyze. However, it is not thought that it currently has this functionality (although we believe this to have been part of the original design).
  • An alternative would be for l3fanalyze to have a second mode of operation implemented, in which it would use histoscope to book and fill histograms, which could be displayed by the ROOT GUI. This is the approach we believe is taken in all of the other examines, although it seems a rather retrograde step.

There are a number of projects that would be a good way for new people to get involved in L3 algorithms work:

• Development of programs/rootuples/histograms to monitor L3
• Test the dynamic unpacker for muon data that would allow the muon code to run even when the muon readout configuration changes.
• Help develop and test the performance of the primary vertex tool.
• Develop an online version of NADA, which kills hot calorimeter cells. This would be useful for finding jets and absolutely indispensible for running filters based on missing ET.

For more information on Level 3 Algorithms work, please consult our web page or contact Dan Claes (dclaes@unlhep.unl.edu) and Terry Wyatt (twyatt@fnal.gov).

Scribe: Terry Wyatt.