[Halld-trigger] Minutes of L3 trigger meeting 5/19
davidl at jlab.org
Tue May 24 07:48:58 EDT 2016
Last Thursday we held an organizational meeting for the L3 trigger. Minutes of the meeting are
at the following link and copied below. For those who are interested, we will be meeting regularly,
every other Friday at 11:00am (JLab time) opposite the DIRC meetings. Next meeting is this
Friday. I’m sending this out to the general halld at jlab.org mailing list this time just to let people know
about it. Future announcements will be made only on the halld-trigger at jlab.org mailing list.
Attendees: David L., Adesh S., Justin S., Mike W., Cristiano F.
L3 is not a requirement for the Fall 2016 and Spring 2017 runs. We will schedule some commissioning time during these runs
Current schedule has no Hall-D running in Fall 2017
Expect to have fully deployed L3 system in use for Fall 2018 run
Primary goal of L3 is to reduce/eliminate events with Eγ < 7GeV
Justin briefly reviewed his slides from Oct. 2014 GlueX collaboration meeting to remind us what has been done and where we left it
Plan is to implement MVA (=multivariate analysis) with focus on BDT
We will start with the TMVA package that is integrated into ROOT
May eventually move to other algorithms developed specifically for this type of triggering
Detailed Monte Carlo where all truth information is available
Real data from Spring 2016 run where events can be flagged as "reconstructable"
Both will be pursued. Simulation will necessarily need to improve to meet physics goals of GlueX and L3 will benefit from that
Simulated data set being generated for GlueX ("sim1") will not be sufficient for L3 study since it has a lower bound of Eγ = 4GeV. We will need to generate another with energies down to π threshold
L3 simulated data set will match the conditions of the "first priority" data as determined by Analysis Working Group. Specifically, runs 11347-11555 (see here for details)
Adesh will put together wiki page with simulation specifications and status
Mike noted that another important piece of information will be the CPU cost of various algorithms that may be used as input to the BDT
Cristiano will start looking at this and documenting it.
This may prompt some optimizations of code where L3 would benefit so the benchmarking tools developed are expected to be reused multiple times
Some preliminary work in this area was shown at the May Collab. Meeting
We will have future meetings bi-weekly on Fridays at 11:00am. This is the same time as the bi-weekly DIRC meetings but out of phase by a week to make it convenient for those who attend both meetings
Next meeting will be Friday May 27
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Halld-trigger