Review Committee Meeting Monday 30 January 2006 09:00 PST / 12:00 EST
Minutes: Monday 30 January 2006 09:00 PST / 12:00 EST
- Start the review of the GRB GWB search
- Organization of the review.
- A combined paper is being prepared that will contain material from both Isabel's and Soumya's technical reports. A draft of the paper will be available on Wednesday at the burst F2F.
- APS meeting: again, combined results are to be presented. (Not yet clear who will be the speaker.) Hope to present the S2-S3-S4 results as well as S5 results.
- The particular results that are to be presented at the APS meeting should be identified to the Review Committee by 19 Feb. (A draft talk is not necessary -- just an indication of what particular sensitivities / results are to be shown.)
- The code is essentially entirely new and has not yet been subjected to a code review. It is MATLAB or C code compiled for MATLAB. Peter Kalmus will chair the code review committee. The target date for a substantial code review is the LSC meeting.
- We started to go through Isabel's technical document.
- GCN circulars: emailed to Isabel. In the past, data extracted by
hand from text message. Now automated. (Control room monitor uses
notices rather than circulars -- more false alarms.)
- S2 and S3: not always well localized (SWIFT not running). Only use H1-H2. Why not also use L1 and search over possible time-shifts?
- Could do, but requires a different pipeline.
- DQ flags. Used for both on-source and off-source.
- Sometimes hard to find 3 hours of off-source data near a GRB. Then use somewhat less than 3 hours.
- What is the minimum? Isabel will find out.
- Of GRBs that occured when at least two detectors were in science mode, how many were not analyzed due to DQ flags, and which DQ flags were responsible? Isabel will find out. (There were 46 GRBs during S2, S3, S4; 36 were analyzed; so 10 were "missed" probably mostly because of the Science Mode flag.)
- Data Conditioning: whitening.
- 1 second Tukey windowed PSDs used. Is there any spectral leakage from low frequencies across band? Isabel will check by comparing the power spectrum with a much higher resolution power spectrum of more data (say 16 seconds) using a Hann window. This will be done for data from S2, S3, and S4.
- Why not high-pass filter the data in the time domain before analyzing it?
- Whitening is done by dividing Fourier coefficient of current segment by magnitude of that coefficient in either the preceeding or the following segment. Is this a good idea? Usually spectral estimation involves doing some average to reduce the variance of the power estimate. A: don't want to do an average periodograms because this might not track non-stationarity well enough.
- Concern: there might be discontinuities of post-whitening time domain data. Isabel will produce plots of the whitened data stream and zoom in on the times around the joins to see if there are discontinuities.
- GCN circulars: emailed to Isabel. In the past, data extracted by hand from text message. Now automated. (Control room monitor uses notices rather than circulars -- more false alarms.)