Minutes of Detector Characterization Teleconference
(October 18, 2002)
Present:
Caltech: D'Ambrosio,Marka,
Zweizig
Florida: Klimenko
LHO: Gustafson, Landry, Raab,
Sigg
LLO: Chickarmane, Gonzalez,
O'Reilly
Michigan: Riles
Oregon: Frey, Leonor, Rahkola
Penn State: Sutton
Syracuse: Saulson
S1 Reports: (S1
Web Page
)
- Calibration Status (G. Gonzalez)
Gaby gave a status report
on the effort to track calibration drifts in S1. A simulink model put together
by Rana allows prediction of the L1 calibration function for different overall
gain factors that vary due to misalignment of the arms or of the recycling
cavity. The model includes all known transfer functions that should affect
loop gain, including digital filters and agrees well with open loop gain measurements
up to about 1 kHz (time delays not included in the model may lead to discrepancies
seen there). Efforts to correct open loop gain for observed drifts in detected
strength of the injected calibration lines have been reasonably successful,
although the signal-to-noise ratio for that detection requires integrations
of minutes to achieve good stability. So far studies have been based on measured
amplitude response, but it would be desirable and will be pursued soon to
monitor the complex response of AS_Q to the excitation channel. Sergey confirmed
that LineMon can now provide such info. Calibration drifts are now handled
with relative corrections to baseline strengths. Daniel and Fred suggested
using absolute strength instead, given knowledge of the transfer function.
KR remarked that it would be highly desirable to have well before the S2
run a model for the LHO IFO's similar to what Rana has done for L1 and to
have autocalibration scripts working as LLO, as they do already at LHO. Daniel
wondered what precision on calibration is needed by the upper limits groups.
Gaby reported that the inspiral group needs about 15% precision on amplitude
and about 10 degrees in phase near 100 Hz. (KR note: Gaby later gave a
LIGO seminar
at Caltech on this subject.)
- Results from SenseMonitor (P. Sutton)
Patrick provided a set of plots (
tar ball
) showing calibration-corrected binary inspiral ranges (averaged over orientations).
He is now writing up a report, but his preliminary executive summary on the
H1 data is in the 235 hours of science mode running, there was an average
of 16 kpc for a binary with 1.4 solar mass neutron stars, giving a cumulative
integrated volume-time of 4 million kpc3-hours. In determining
the calibration line strength for correcting the raw noise-based range, he
used a windowed PSD. He was able to perform some consistency checks on his
correction algorithm by looking for jumps between inferred transfer functions
and daily autocalibration results. Generally, the discontinuities were observed
to be small. The frequency range from which the inspiral range is computed
is wider than it was online in S1. The lower limit has been lowered from
300 Hz to 20 Hz. KR asked Patrick to doublecheck how much the range increases
with the frequency range change, since the rescaling algorithm used here
isn't expected to work well below 100-200 Hz. Patrick will soon be looking
at L1 data from S1 and will set up a web page with a summary of results.
- Data Quality (J. Zweizig)
John gave a followup report on what he presented on October 4. He hasn't
yet finished rerunning the data quality monitor on all the data, for technical
reasons, but he has written a preliminary
draft summary
based on what he has seen so far. Briefly, he finds a number of potentially
important pathologies:
- Periodic noise bursts in the 39-56 Hz band of H1 AS_Q correlated
with similar bursts in REFL_I. These bursts occur roughly every 20 minutes
and may be connected with automated discrete actions in the PSL control system.
Fred suggested looking at PSL EPICS channels for source glitches.
- A 100 Hz "bulge" in H1 AS_Q on August 26 (blrms increased 2 orders
of magnitude for more than 2 hours). A similar bulge was seen briefly again
on August 27.
- Occasional H2 lock stretches with anomalously high glitch rates,
mostly seen in REFL_I
- Periodic glitches in L1 MC_F in 150-261 Hz range, correlated with
glitches in REFL_I
John is creating new trend data from the data quality monitor
and will eventually replace those now at the site (generated online during
S1) with the revised versions. In order to allow different upper limits analysis
groups to define their own "good data" criteria, John plans to set up software
infrastructure for choosing among candidate veto criteria determined from
these studies.
A.O.B
- Fred expressed concerned that more configuration control be put
in place on trends produced by DMT monitors, given the snafu in S1 when one
monitor's trend channels were renamed, leading to malfunction of one of the
frame builders. John offered to compile a list of all known DMT trend channels
to serve as a baseline.
- Next meeting: November 8 telecon at 1:30 p.m. EDT.
Schedule
for S1 investigation reports over next several meetings