Talks

• Studies of data taken with global_CMT-8.xx:
  • Update (with higher statistics than available last week) on single electron triggers: Ulla Blumenschein.
    From a set of runs taken with global_CMT-8.10, Ulla has used 1720 events triggered at L1 with CEM(1,10) and mark&passed by L3 to check rejection rates for the new set of 5 single electron filter scripts. The rejections for each individual script agree pretty well with the online numbers. The rejection for the .or. of the 5 filter scripts is 14 (plus/minus about 1). This study was performed using the information on fired trigger names in 300,000 events read from the reco_analyze roottuples. (Roll-on the inclusive streaming of monitor data!)
    Action item: It is interesting to note that there is an anti-correlation between the events passed by the calorimeter-based electron filters and the track-based filter for CEM(1,10). The hypothesis that the calorimeter-based filters pick up high energy pizeros, which naturally do not have high p_t tracks pointing to them, needs to be checked with MC.
  • Update on L3 tracking: Daniel Whiteson.
    Daniel showed plots of the p_t distribution of leading L3 tracks and the expected rejection rates as a function of p_t. This study was performed using the debug_info data on mark&pass events triggered by CEM(1,10) or mu1ptxwtxx_fz. The data were from run 160588, which was taken with global_CMT-8.00. The expected rejections appeared to be lower than those seen online.
    Action item: Daniel will check for consistency between the information on the found tracks and the decision of the track filter.
    N.B. to see basic track quality plots from these data see Daniel's talk at last week's meeting.
    Other stuff, not directly related to global_CMT-8.xx:
    • Daniel has provided cookbook-level documentation for general users of how to run the L3 tracking on data and Monte Carlo. This can be found at: http://www-d0.fnal.gov/~danielw/tracking/L3TrackingFAQ.html. The info. has been test-driven already by a few users and found to be very useful.
      We need similar "how to" level contributions from other experts and experienced users. Well Volunteer!
    • Work is in progress on the efficiency measurement for L3 global tracking using Z->ee events.
    • Work is continuing also on trying to understand the peculiarities observed in the track parameter pull distributions. Studies using a toy MC suggest that even when guaranteed-to-be-correct hit errors are provided, the track fitter returns incorrect track parameter errors.
  • Muon geometry with the L3 geometry manager: Martijn Mulders.
    Starting with global_CMT-8.00 we are running online the L3 geometry manager, which was provided by Robert Illingworth. This means that the "offline" geometry for muons is now available in L3. On-/off-line comparisons of local muon track-segments suggest that the correct geometry is, indeed, being used. Notes:
    • The hit information in L3DebugMuoInfo is in local muon coordinates rather than global DØ coordinates and so cannot be used for this purpose.
    • Unexpected differences between the segment parameters in debug and Maxopt were seen, apparently only for the first processed event. Under investigation.
• Update on planning/implementation of L3 Thumbnail: Peter Tamburello.
  Peter has been working on understanding the information stored in the L3Chunk by the various scripts and filters. He had a number of nitty gritty detailed questions that we couldn't completely resolve during the meeting, but need to get sorted out. As a first step Peter has implemented a minimalist approach of storing on the Thumbnail the p_t, eta, phi of the L3 physics objects. This will allow the L3 objects to be matched to objects at other levels (L1/L2/reco/MC). There was general agreement (I think - scribe) that we need to store all of the quantities upon which cuts may be made in a filter. Obviously costs space, but difficult to see serious trigger studies being possible without this information. There were more diverse opinions expressed on the need for information that identifies which physics_objects caused which filters to fire online. We need more discussion of this to reach a conclusion.

Discussion Items

• Han Do has started work on timing studies. He will use both the stat_manager information and profiling. He needs a modification to d0tools to allow him to use runD0TrigSim with the -profile option. Jon Hays will help with this. Two levels of profiling are available. The first returns the cpu time used by each function. The second provides a detailed tree, differentiating between calls of a particular function from different places.
• News: Update on L3 Filter Monitoring.
  l3LumServer is no more; long live l3fmon!
• Jon mentioned:
  • p11.10.01 may be run online. A few basic checks are needed to make sure nothing has been broken due to changes out of our control. Martijn/Ulla volunteered to do this.
  • p11.11.00 l3fanalyze progress. The change to ensure shorter names has been implemented. This means that in order to sure to avoid ambiguities, when you use a variable you should specify block name + variable name. The crash on termination has been fixed. As a short term solution for reco farm production, Jon will run the tool list generator and copy the list to the run_data area. This will have to be updated every time the farm processes data corresponding to different trigger lists. Perhaps the long term solution will never exist?

Scribe: Terry Wyatt.