Review Committee Meeting Monday 14 January 2008 08:00 Pacific / 11:00 Eastern
Minutes: Monday 14 January 2008 08:00 Pacific / 11:00 Eastern
Agenda and Contact Info
- Review of BurstMDC Simulation
- Continue with last portion of graven (production of inverse-calibrated data) at line 347.
InterCall telecon service:
Phone: 1-866-380-5536, participant code: 435 672 9587 #
International callers ++1-816-249-4731 with same code
International access numbers (~10 cents/min surcharge for LIGO Lab):
Paris 017 080 7156 ;
Lyon 042 603 0036 ;
Germany toll-free 0800 182 1591 ;
Berlin 030 726 167 371 ;
Rome 00645 217 080 ;
UK toll-free 0808 234 7914 ;
London 0203 107 0293
Handy participant command: Press *6 ["*M"] to mute, #6 to unmute
Erik Katsavounidis, Brian O'Reilly, Norna Robertson, Peter Shawhan,
Patrick Sutton (minutes), Keith Thorne.
- Proposal for "equinox event":
Shawhan presented a proposal for bringing the Sept 21 burst candidate to the detection committee.
- This event may or may not be a blind injection.
- At the last LSC-Virgo meeting Dave Reitze and Benoit Mours presented a flowchart of the detection procedure. This includes bringing a paper/results statement to the detection committee and asking for feedback, including suggestions of further tests.
- Today's proposal is a modification to part of the flowchart, bringing the detection committee in earlier to ask whether all appropriate checklist-type questions have been addressed (i.e., is the info about the event "complete").
Shawhan: When the event goes to the collaboration, we'll have a paper draft in hand. It will be a multi-pipeline paper (cWB, Q, BN, ...?).
O'Reilly: When will a paper be presented? Shawhan: realistically after the March meeting, since the first year paper is not finished. Katsavoundis: We want to take the event to the DetComm earlier, since the paper may be a long way off but the detection checklist results probably won't change much by then. We will ask to the DetComm to vet the statement of 1 event in "X" years, and to look at tests and pipelines, and we'll ask if there is anything missing. We expect the event will survive in the zero-lag data set, and we want to move ahead with the DetComm now instead of waiting for full second-year analysis results.
Klimenko: What is the purpose of submitting to the detection committee? Do we claim a detection, or make no judgement? Shawhan: The purpose at this point is to ask the DetComm if we are missing anything in the follow-up checklist. Later, we will draft a paper that makes a statement and run that by the DetComm for evaluation.
(O'Reilly signs off).
- Question regarding Virgo MDCs:
- Klimenko asked if the format of Virgo information in the MDC log files has changed? Seeing "funny behaviour" with coherent WaveBurst. Thorne: No change. The frames have changed, in that the MDCs now have both 16kHz and 20kHz channels for Virgo, but the log files are the same.
- Review of BurstMDC Simulation:
Continued with review of graven (production of inverse-calibrated data), lines 347-409.
Thorne: This section does not affect h(t) data.
Both AS_Q and DARM_ERR timeseries are made. Set to zeros if skipCalibFlag == true. For pre-S4,
DARM_ERR is filled with AS_Q.
Action item: Correct internal documentation at lines 348, 349, which states uncalibrated data is set to raw data instead of to zero:
0347 % IF not calibrating
0348 % SET AS_Q to raw data
0349 % SET DARM_ERR to raw data
Rest of graven is straightforward; review to concentrate on the functions called between lines 347-409:
ndx2gps, and their child functions.
Reviewed function calibsimfd: straightforward. The actual calibration is done by h2asq_fd. This function merely checks for valid arguments and also applies zero-padding in the call to h2asq_fd to protect against ringing. Also, requires user set environment variable LIGO_CALIB_DATA to hold directory containing the calibration files. This allows the user to control the calibration version used without recompiling the code.
Possible failure point? Line 70,71: variable sigType not forced to uppercase before being compared to 'AS_Q', 'DARM_ERR'.
Reviewed function h2asq_fd: no issues seen. Need to review child functions getalphabeta, getopenloopgain, and getsensefunction.
Action item: Thorne will send links to direct tests of calibsimfd, h2asq_fd.
Reviewed function getalphabeta: uses "persistent" storage for cal data read from file - for speed.
Action item: Make names gpsTimes and gpsList at lines 4, 9 in help documentation consistent.
Action item: Many sub functions. Each looks simple enough, but there are enough of them that they should be read over carefully offline. These include the gps-utils functions and the hperasq functions readalphabetafile and getabfile.
Reviewed function getalphabeta and subfunctions (in the same .m file): no significant issues. This function gets calibration alpha, beta values given an IFO and GPS time.
Action item: The subfunction getgpsrows appears to give misleading help info for the case errCode==3:
3 - [gpsStart-gpsEnd] does not overlap list
Specifically, in this case gpsList is completely contained within [gpsStart-gpsEnd], in that gpsStart < gpsList_start and gpsList_End < gpsEnd.
Action item: The subfunction setalphabeta has a minor bug. The internal documentation says that for not-valid data the gps time is set to the first input value, while in the code it is set to the last value:
0274 % IF GPS rows are valid
0291 % ELSE
0292 % SET alpha,beta,gamma = 1
0293 % SET GPS to starting GPS time
0331 gpsList = gps2time(gpsVec(end));
Also, the help documentation for this function should explain what happens in this case.
Continue next week with h2asq_fd, line 85, and child functions
getopenloopgain and getsensefunction.
Reviewed function getabfile. No issues. This function gets a calibration file with alpha, beta values given an IFO and GPS time. It is useful for cases where the calibration changes during a run.
Action item: Did not review child functions.
- A.O.B.: None.
$Id: minutes-2008-01-14.html,v 1.5 2008/01/21 14:54:05 psutton Exp $