LSC-Virgo Burst Analysis Working Group
Review Committee Meeting Monday 24 July 2006 09:00 PST / 12:00 EST
Minutes: Monday 24 July 2006 09:00 PST / 12:00 EST
Agenda and Contact Info
- Continue discussion of draft of S4 untriggered burst search [ PDF ]
- Continue discussion of the S2-S3-S4 GRB-GWB search [ HTML ] including an update on modeling the redshift
distribution of GRBs [ PDF ]
- Continue discussion of the SGR search [ HTML ] including an update on the sensitivity as a function of
various waveforms [ HTML ]
AccuConference teleconferencing service:
Phone: 1-800-704-9896, participant code: 038621#
International callers ++1-404-920-6472 with same code
- S4 Untriggered
- JC's sensitivity estimate: Z_G=6.7 implies w=4.88 for 4km IFOs which
corresponds (?) to an expected sensitivity of hrss(det) ~ 3.5 sqrt[S_h(f0)]
or hrss ~ 8.5 sqrt[S_h(f0)] including a factor of sqrt(5) from
sky-averaging and a factor of 1.1 from calibration uncertainty. But this
is not what is seen: something around a factor of 2 greater for hrss50%.
Need to understand discrepancy. JC will post a write-up of
the calculation. Sergey will see if he can reproduce it.
- Code reviews: want to have a clear statement of what the scope of
the review is, i.e., exactly which portions of the pipeline have
been checked. Worry is that what is reviewed in WB review might not
mesh with where the CorrPower review picks up... there might be a script
that translates the output from WaveBurst to the input of CorrPower that
is not reviewed an that may have mistakes. If we have a clear statement
then we can identify holes in the current review.
- GRB: Soumya's document [ PDF ]
- Investigation of redshift distribution of GRBs based on SFR history.
Peak in GRB redshift of about 2. This corresponds to a luminosity distance
distribution in which effectively all the sources are farther than 2 Gpc.
But we've seen GRBs nearer than this! Q: Are the nearer ones
all short GRBs (model based on SFR is for long-GRBs)? Or is there something
- Is cutoff at 2 Gpc important? It seems to imply that the chances
of detecting anything associated with GRBs is negligible. We would not
expect the population study to yield any meaningful limits.... Can see
if the limits are robust against changes in the cutoff.
- Alternative: model the redshift distribution with a beta distribution
that (i) behaves like z^3 for small z and (ii) has a peak at z=2. This
will fully specify the distribution. But would it look like the
theoretical distribution? Soumya: can do this with luminosity distance
too -- instead of having a cutoff can make it so the nearby GRBs have
the appropriate power law.
- Goal would be to constrain the factor that converts luminosity
distance into snr given the population distribution parameters.
- (Isabel: expects to complete her investigation in the next few days.)
- SGR: Luca's investigation [ HTML ]
- Injected a variety of different signals types. SGs with varying Qs.
AM and phase-modulated signals with a variety of modulation frequencies.
- Results: sensitivity relatively independent of Q for Q larger than
10000 ; below this the signal tends to get vetoed! Is this a worry?
Q: For these small values of Q, is FWHM of signal larger than the EM
observations bound it to be (at least for the EM QPO)? If we're worried
about this, perhaps modification to the algorithm could be done, e.g.,
tuning the veto level or introducing a gap between the source band and
the reference bands.
- AM/Phase Modulations: sensitivity degraded as sidebands enter the
neighbouring "reference" bands; slightly improves again as the sidebands
move out of the reference bands (for large modulation frequencies). These
extreme examples seem to violate expected FWHM of signal QPO.
- Question about modulation depth. Also: what about a slow frequency
modulation in which the signal starts at, say, 96 Hz and then drifts
down to, say, 90 Hz over the 50 s signal duration?
- Ratio test doesn't seem to work. Not surprising in hindsight: in
order to get a reliable power estimate in a 1 Hz band, would want to
integrate for at least 10 seconds -- more really. So ratio test needs
either a longer integration time or a wider band in order to measure
the reference spectrum.
$Id: minutes-2006-07-24.html,v 1.4 2006/07/24 21:09:06 jolien Exp $