coherent WaveBurst Review Meeting 30 Apr 2009 09:00 Pacific / 12:00 Eastern
Agenda: Thursday 30 Apr 12:00 Eastern
- Updates on Action Items [web].
- Follow-ups of loud missed injections [web]
Dial-in number: 1 800 704 9896 International dial-in: +1 404 920 6472 Conference code: 5374 2349 #
AttendancePatrick (minutes), Francesco, Igor, Sergey, Virginia, Romain, Irena, Giovanni.
Follow-ups of loud missed injectionsRG led us through his web page [web].
- Looked at html table of loud missed injections (LMI) by IY. Contains 527 injections.
- Most LMI have F_+^H < 0.01 for linearly polarized signals.
FS: Why doesn't low F_x^H also caused missed injections?
IY: Since linearly polarized, h_x = 0 and only F_+ is important.
PJS: It must be that F_+, F_x are defined in the source frame?
IY: These numbers (F_+, F_x) are taken from GravEn.
Action Item: (PJS) Check definition of polarization in GravEn logs.
- IY: F_+^L not as important as F_+^H because H1H2 gets more weight in coherent combination than L alone. PJS: May be due to correlation between H1 and H2 allowing these events to pass consistency tests.
- IY: Some LMI may be due to injection falling on glitches that fail the consistency test.
- RG ran follow-ups on 32 LMIs -- the ones IY listed as having high IFO response (therefore no obvious problem with sky location). 32 / (2700 inj/waveform x 28 waveforms) = 0.04 %, so not a significant fraction of the injections.
- RG: 22 occurred during the bad H2 month.
- RG: H1_ETMY_COIL_OVERFLOW flag was on during or near (not necessarily in coincidence with) almost all LMI. This channel was not used in DQ cuts as it was found to not have significant efficiency. Speculation: Maybe it is a better veto during the H2 bad month.
- DARM_ERR Q scans typically very quiet, therefore loss not due to glitch.
SK: May be detected by pipeline but lost in consistency cuts. IY: Not checked.
Action Item: (IY) Check ROOT files to determine at what stage these injections are lost (netcc, penalty factor, elsewhere).
- PJS: Don't see anything on Q scans. SK: May be lost not due to glitch but due to some inconsistency in the reconstruction. E.g.: the regulator may affect the reconstruction. For large E the regulator goes to 0.01, which is comparable to |F_x|^2/|F_+|^2 for many sky locations.
- RG: Most of these 32 LMI not explained by Q scans or DQF. One suspicious thing is that a large number of them occur in one 10-day period of the bad H2 month, with the H1_ETMY_COIL_OVERFLOW flag (this seems inconsistent with the hypothesis of the losses being due to consistency cuts).
- SK: During this time the H2 noise was so high that we effectively had a 2-detector network. This may be affecting the regulator and how the analysis is performed.
- Next steps: Repeat for LMI in other livetimes. PJS: H1H2 most important.
- PJS: Asked about application of L1 DQF to triggers in H1-H2 only time. Brennan's email on this issue for the second year said that 245 injections in H1H2 times were lost due to cat 3 L1 flags.
- SK: If only 3-4 % effect then just make sure the same flags are applied to the injection and raw triggers. IY is "99%" certain they're the same.
IY: Only applied to zero-lag triggers if there exist zero-lag events. None in this
search, so application of flags does not affect detection results.
Action Item: IY, SK to check application of L1 DQF to H1-H2 only times. Want to confirm that L1 segment list was applied in post-production only and that it had no effect on the livetime.