Minutes of Detector Characterization Teleconference
(June 25, 2004)
Present:
Caltech: Laura Cadonati, Szabi Marka, Peter
Shawhan, John Zweizig
CSUDH: Ken Ganezer
Florida: Sergey Klimenko, Andre Sazonov
LHO: Mike Landry, Daniel
Sigg
LLO: Marc Cenac, Brian O'Reilly
LSU: Gaby Gonzalez
Michigan: Keith Riles
MIT: Erik Katsavounidis
Oregon: Emelie Harstad,
Robert Schofield
Penn State: Mike Ashley
UWM: Xavier Siemens
S2 / S3 Investigation Reports
-
Calibration:
-
Status and plans (Gaby Gonzalez, Mike Landry, Brian
O'Reilly)
Mike reported that version 2 S3 calibration files
have been posted, including vetted frame files containing alpha and beta
coefficients. Gaby is working with Xavi on updating the data quality flags
for calibration line injections. The conservative flags chosen in version
1 break the data into many segments. She hopes that refinement of the flags
will lead to less fragmentation. KR wondered whether the flags would be
tied to 60-second intervals with good alpha/beta measurements, given that
Xavi has second-by-second measures of line strength. Gaby confirmed that
for now the 60-second discreteness will be preserved, given the way analysis
programs (including simulation jobs) obtain the calibration info. One technical
obstacle has been recreating SenseMon alpha/beta values to align with the
60-second boundaries, since the original SenseMon runs had somewhat arbitrary
start times. John suggested that SenseMon should use a new synchronization
feature in the DMT to avoid the problem in the future.
The final document on S2 calibration errors is
near completion. It will include studies of fast fluctuations (see recent
presentations by Gaby and Xavi), along with documentation of the determination
of the sign of the calibration.
-
Time domain update (Xavi Siemens)
Xavi is working with Gaby to resolve a 5% discrepancy
in alpha/beta determinations between frequency and time domain calibration.
He has been holding off on generating full S2 h(t) frames until that is
understood. KR suggested not waiting much longer to make the h(t) files,
even if not perfect, in order to give analysts some experience working
with them. Similarly, in getting ready for S4, we may not have final reference
functions at the start of the run (although the calibration team will try
to have them); so it's reasonable to put out the best current estimate
of h(t) quickly. Xavi believes the time domain calibration should be no
more susceptible to imperfect reference functions than is the frequency
domain calibration. A look at the S3 data shows much improvement in the
presence of calibration lines, but a handful of the odd 1/3-second dropouts
seen in S2 occurred in S3 too. (See plots Xavi provided after the
meeting for H1 , H2
and L1 .)
-
Data quality - Investigations: (
S2 , S3
) Segment repository (S2 , S3
) (John Zweizig, Keith Riles)
John has updated his DAQ
readout error web page based on new info from Ben Johnson. As John
showed at the Tufts LSC meeting, Ben has mapped the data buffer used for
LHO and LLO in S3 and has verified that AS_Q should nominally not be affected
by the low-order-byte corruption problem seen in other channels due to
a rogue VME clock card. Except for timing jitter, the times of corruption
and therefore the channels affected during block transfers are known. The
evidence seen in the data channels seems to bear this out in that very
high rates of data-valid-word corruptions occur in the right channels,
and very few occur in other channels (including AS_Q). The data-valid word
comes at the start of each channel block, giving a sampled measure of corruptions.
KR wondered whether the three S2 AS_Q channels were similarly only occasionally
affected. John hasn't received S2 info from Ben yet, but believes he is
working on it. (KR followed up with Ben later and verified he is producing
the S2 channel maps.) John has also been looking at differences between
version 3 and 4 S2 "no_calib" flags to make sure they are understood. In
a few cases, the most recent version 5 flags do not fall on clean 60-second
boundaries.
KR urged all investigators to get ready any info
they have on bad S3 data intervals for inclusion in the upcoming version
2 release.
-
Timing (Szabi Marka, Daniel Sigg)
Szabi reported that after a recent calibration
at LHO, the atomic clock's drift has been reduced to below 10 ns/day. Also,
he will produce in the next week data quality flag info for bad L1 timing
during S3.
-
Glitches
/ Burst/Inspiral search group vetoes (Laura Cadonati, Erik Katsavounidis)
Laura summarized
a great deal of activitity in the glitch group. In brief:
-
Online trigger studies are well underway, using both
DMT and inspiral triggers generated during the run
-
Several DMT monitors are being re-run offline to
generate triggers, with systematic collegion of results. ETG production
of playground triggers has also begun.
-
Glitch-finding algorithms are being validated, both
via simulations and via cross-comparisons
-
In addition, PEM injection studies and glitch characterization
by segment have begun. A new database table has been defined for DMT triggers.
-
The group is also stepping back to re-examine veto
strategy and is working on a standardized set of figures of merit to allow
quick turnaround analysis during the S4 run.
-
BurstMon
development and S2 non-stationarity (Sergey Klimenko, Andre Sazonov)
Sergey provided a detailed description (with fabulous documentation!
- KR) of a new DMT monitor called BurstMon which will provide figures of
merit during the S4 run that are relevant to the burst search. Using part
of the algorithm developed for the WaveBurst ETG, the new monitor will
report 1) rates of cluster reconstruction for a nominal black-pixel probability;
2) strain sensitivity (hrss at 50% efficiency), based on injected simulated
waveforms; and 3) noise variability in selected frequency bands. Although
the monitor is not yet complete, it has already produced interesting results
on noise variability during the S2 run. The noise variability looked
reasonable for H1, but the variation was much larger for H2 and even larger
for L1, with variability dominated by the microseism and peaks in the few
Hz region (stack resonances?). It is assumed that most of the L1 effect
comes from true noise variation, not calibration fluctuations, since Gaby's
and Xavi's past studies haven't found such large fluctuations. Just to
be sure, though, KR asked Xavi to check his second-by-second reconstructed
alpha's for one of Sergey's sample intervals. Sergey and Andre plan to
finish BurstMon by the August LSC meeting, in time for shakedown before
the S4 run in the fall.
John Zweizig wondered whether it would be better to report an astrophysical
range for an assumed waveform/model. He suggested that FOM's which increase
with good behavior would be less confusing than one that minimizes hrss
at 50% efficiency. KR felt that operators are accustomed to minimizing
noise and might find it useful to have an astrophysical FOM with units
identical to the familiar strain density; variation of the ratio of hrss
over the minimum strain density might be a good indicator of serious non-linear
noise to track down. It was agreed to try both kinds of FOM's and get some
experience with them.
-
Violin modes (Sergey Klimenko)
No update.
-
Interchannel
correlations (plus info
on 250 Hz lines ) (Nelson Christensen)
Nelson was unable to make the telecon, but provided links to new tables
compiling known significant coherences between AS_Q and PEM channels. These
will be used by the pulsar group for vetoing instrumental lines. Nelson
has also looked into possible coherences between the three IFO AS_Q's and
PEM channels at 250 Hz, since both the stochastic and pulsar groups ran
into artifacts at that frequency.
-
Bilinear couplings (Steve Penn)
No report.
-
Correlated inter-site environmental transients &
local environmental disturbances (Robert Schofield)
Robert gave updates on several ongoing efforts to reduce environmental
coupling into AS_Q at Hanford:
-
Concerning dust glitches, they are now pricing air
filters and looking at a scheme where a plexiglass enclosure around the
dark port beam path would be purged whenever a person is present.
-
It has been decided not to buy acoustic enclosures
for H1 electronics racks. The racks will be moved out of the LVEA completely
after S4. For H2, the racks will be moved farther from optical tables but
still in the LVEA, and acoustic enclosures will be installed after S4.
Robert believes that if other noise sources above 100 Hz are beaten down
to design level on H1 for S4, there will be acoustic limitations in some
bands.
-
Below 100 Hz there is a bigger acoustic problem,
which will be addressed by floating optical tables after S4.
-
Regarding magnetic pulsing (leading to sidebands
on 60 Hz), new temperature control boards are being tested now. Also, a
new smaller-area heating coil element has been tested, which reduces 60
Hz fields by about 3 (measured near the heater). In addition, an experiment
is being carried out to use continuous heating with cooling water.
-
Regarding magnetic couplings, the earlier hopes that
the couplings were to readout electronics instead of to the test masses
have been greatly dimmed. It looks as if the couplings are directly to
the masses (due to imperfect dipole cancellation of the magnets glued to
the mirrors). One possible solution is wrapping mu-metal around the chambers
-
Regarding dewar glitches, further tipping of the
dewar (to clear a bumper) did not succeed. John Worden has approval now
to try insulating the dewars. Robert's calculations suggest the insulation
should reduce temperature fluctuations by about 4. New Oregon graduate
student Emelie Harstad will be writing a new DMT monitor to detect
dewar glitches.
KR asked whether Robert could put together a list
of times during S3 when there were intrusions at the dark ports of the
three IFO's, from going through the elog, so that new DQ flags can be defined.
Robert ageed to do so.
-
S2 & S3 AS_Q spectrum and
line compilation - preview (Keith Riles)
KR gave a preview of a compilation of strain noise density estimates
from H1 for all of S3. The densities are based on weighted averages of
30-minute FFT's (Hanning with 50% overlap) provided by Vladimir Dergachev,
using an algorithm developed for the all-sky, incoherent pulsar search.
The 0.5 mHz binning and hundreds of averages allow many lines to be observed
which affect the incoherent pulsar searches. KR is working on a line-detection
algorithm for automated compilation and for cross-checking against the
list of coherence peaks found by Nelson Christensen (see above) and students.
The hope is to allow vetoing of instrumental lines in the pulsar analysis.
In the meantime, Vladimir will be generating the strain estimates for H1
and L1 for S3 and for all three IFO's for S2. John requested that the reporting
of lines include estimates of strain power, not just SNR, to make it easy
to determine which lines to notch out in DMT monitoring of non-stationary
noise.
A.O.B.
-
Next detchar telecon: Friday July 2 at 1:30
p.m. EDT (DMT software)