Minutes of Detector Characterization Teleconference
(February 1, 2002)
Present:
Caltech: Drever, Lazzarini,
Marka, Matone, Shawhan, Whitcomb, Yamamoto, Zweizig
Carleton: Christensen
Florida: Klimenko
LHO: Landry, Raab, Schofield,
Sigg
LLO: Coles, Gonzalez,
Zotov
Michigan: Riles
Penn. State: Sutton
Syracuse: Penn
UT Brownsville: Romano
Wash. State: Bose
E7 Reports:
-
"Clean" Lock Statistics - G. Gonzalez
Gaby has put together a web
page with statistics on locked-IFO segments, keeping track of stretches
when no control parameters were changed and keeping track of double-coincidences,
triple-coincidences and anti-coincidences. (Dave Chin and Peter Shawhan
automated the selection of segments in which no controls changed.)
Although both 4K IFO's were locked slightly more than 70% of the time,
and the 2K more than 50%, the fraction of time over all of E7 for which
all three IFO's were locked for more than 15 minutes with no changing of
control settings was only 11% (46 hours). Gaby and Patrick Brady have also
identified "playground" data subsamples for use by the burst and inspiral
upper limit groups for initial exploratory analysis to define instrumental
vetoes.
-
Calibration Status - M. Landry
The E7
calibration page has been updated recently with information on the
three calibration runs carried out during E7 (beginning, near midpoint,
and at end) by Mike, Rana Adhikari, Szabi Marka, and Luca Matone). A preliminary
calibrated transfer function and implied sensitivity have been derived
for each interferometer. Results are not final, but should be in the right
ballpark. Mike is working on sorting out residual discrepancies among the
three calibrations. He is also working with Sergey Klimenko to track the
strength of injected lines as a function of time. Technical problems created
a number of gaps in the injected line record, however.
-
Violin Modes - S. Klimenko
Sergey has looked so far at the excitation
of fundamental violin modes (~345 Hz) in the Livingston 4K IFO in the
AS_Q and REFL_I channels. After excitation at the start of a locked stretch,
the lines typically ring down with a time constant of 2-4 minutes with
Q's in the range of several hundred thousand. Detailed quasi-spectra (histograms
in frequency of exceeding a SNR of 3) around each "line" are consistent
with the expected families of four lines per test mass. Sergey does not
yet know which lines come from which masses, but hopes to deduce that information
from data in single arm and Michelson configurations. Albert Lazzarini
wondered what the expected amplitude should be from pure thermal noise
for the modes; Fred Raab believed it to be an rms of about 50 mFm, well
below what can be seen now, given the IFO noise level.
-
Signal Injection and Remote Data Access - P. Shawhan
Peter reported that three different types of signals were injected
at selected times during E7: inspirals, bursts, and stochastic. The stochastic
noise was injected in the two 4K IFO's in two 512-second stretches. Varying
amplitudes of signals were used during the injections, some of which were
large enough to be seen by eye in AS_Q. These injections were carried
out by Peter and Sukanta Bose. John Zweizig reported finding some of the
injected bursts with DMT glitch monitors, but one class of bursts (nicknamed
"blips"), with much high-frequency content, eluded detection.
Peter also reported on the status of accessing data remotely (as opposed
to direct access by LDAS or DMT programs running on site). The getframes
utility does not currently work on the 16-second frames stored in the Caltech
HPSS archive, but does work (with fewer options than nominal) on data stored
at the sites. At the moment the 16-second frames also cannot be read by
the online Diagnostic Test Tool or Data Viewer programs. One temporary
way to get around this last problem is to use getframes to read in data
and output it in 1-second frame format. It is not expected that the 16-second
frame reading problem at Caltech will be fixed soon.
-
Inter-channel correlations - N. Christensen
Nelson and Dennis Ugolini have been looking at correlations between
AS_Q, MC_F and various accelerometers. They have seen strong correlations
in the 0-20 Hz and 50-70 Hz bands. They are also starting to look at bicoherence,
using Steve Penn's xbic program. They will be looking soon at correlations
with microphones. Results will be posted on the E7
Correlations Results page .
Inter-site Power Mains Correlations - A. Lazzarini
Albert, Andrea Vicere, and Robert Schofield have been working to understand
an effect first reported by Sergey Klimenko at the May
2001 detchar telecon from analysis of E3 data. Sergey had found
that while strong correlations in phase between the power mains at the
sites decayed rapidly, as expected for Washington and Louisiana, there
was a small but statistically significant correlation that persisted with
Fourier peaks in the phase difference near 0.5, 1.0, and 2.0 mHz. He believed
that the correlation existed because of a GPS-enforced long-term synchronization
between the Eastern and Western power grids and assumed the peaks at low
frequencies arose from the periodic manual adjustment of frequency within
each grid.
Albert et al. have pursued this idea and written a detailed
document summarizing their findings. From interviews with power grid
officials, they have constructed both analytical and matlab models of the
phase adjustment that gives good qualitative agreement with what Sergey
observed and what they find in a 1-year data set of phase errors every
half hour, provided by the western power grid. (Within each grid, phase
drift is allowed to accumulate until it reaches a maximum value, at which
point the frequency is adjusted, usually by +/- 20 mHz. The maximum accumulation
in the east is 8 seconds and 2 seconds in the west. In addition, the instantaneous
frequency is not allowed to deviate more that 0.02 Hz from nominal. Typically,
adjustments are made every half hour to an hour.) When a data set for the
eastern power grid becomes available, it will be analyzed similarly and
compared to the model. In summary, the long-term correlation between the
power mains observed at the sites seems to be understood. Sergey mentioned
that he will be looking soon with his DMT monitor at the full 17-day E7
data set and will try to verify with fine timing resolution more of the
model predictions.
Status reports on DMT infrastructure:
-
General - J. Zweizig
-
A bug was recently fixed that interfered with DMT decimation of data in
16-second frames (affected Steve Penn's interactive bicoherence monitor
- see below).
-
John is working on an offline emulator of the length sensing & control
(lsc) software running in the lsc microprocessor. The goal is to check
calculations independently and raise an alarm if calculated quantities
in the data stream are not correct because, e.g., of a roundoff error.
The DMT monitor can sample a much longer time stride than the online 16
kHz real-time processor, allowing more efficient but equivalent code to
be written. The emulator project is motivated by strange glitches seen
in the running of the online code. John has run into problems with hard-coded,
difficult-to-find filter constants used in the real-time system and with
non-standard nomenclature for channels. But he now has a monitor that works
well for the Hanford 4K system and is tracking down missing constants /
channels for the other two interferometers.
-
KR wondered if progress has been made by anyone on defining an online "physics
mode" button to be pressed by operators when lock has been established
and no further adjustments are expected. Peter Shawhan confirmed his plan
to create an alarm for any changes to control parameters when in physics
mode. Stan Whitcomb confirmed that he will ask someone at one of the sites
to create the button (define/load variable into EPICS database, create
a dedicated MEDM screen). Stan offered to help define the channels Peter
should monitor in setting an alarm. (This should be very similar to the
set being used in E7 to define "clean" locks.) Once this new infrastructure
is in place, then John's locked-segment generator for the database can
do in real-time and automatically what Gaby and others have had in E7 to
do by hand or with ad hoc scripts.
Status reports on DMT monitors:
-
Line noise (multi-taper method) & inter-channel correlation monitors
- B. Allen, A. Ottewill (no report)
-
Line noise (quasi-monochromatic method) monitor - E. Daw, S. Klimenko
(Sergey reporting)
Sergey has added a summary html output to the line monitor, displaying
a list of lines being monitored, with information for each on central frequency,
width, and amplitude. The new version will be sent to John in a day or
so. Sergey proposes to run six monitors at LHO and three at LLO, with three
monitors per IFO looking at DARM_CTRL, CARM_CTRL, and MC_F. Initially
the lines monitored will be the power mains, violin modes and harmonics.
Others may be added, as needed. Someone wishing a particular line to be
monitored should contact Sergey.
-
Band-limited / seismic noise monitors - E. Daw (no report)
-
Bilinear cross-couplings & broad-band non-Gaussianity monitor - S.
Penn
Steve has written up instructions (now
linked to detchar bulletin board) on how to use the interactive bicoherence
program xbic. Steve is working on a new version with an interactive GUI
control, with channel and bandwidth selection similar to that of the Diagnostic
Test Tool (using Daniel Sigg's GDS library routines). The current version
is not entirely robust in that it can fall behind the data in real-time
running, in which case it deliberately skips frames.
-
Lock transition and servo instability monitors - D. Chin, K. Riles
(KR reporting)
Dave has fixed some small bugs in ServoMon and revised the summary
html page.
-
Rayleigh behavior monitor (non-Gaussianity vs frequency) - S. Finn,
G. Gonzalez, P. Sutton (Patrick reporting)
Because of technical problems, the present version Rayleigh monitor
was not run during E7. A new version is under development that will incude
a GUI interface (instead of the existing configuration file) and include
a summary html page.
Patrick also confirmed that another monitor is under development
at Penn State by Kevin Schlaufman to provide a measure of overall IFO sensitivity
to inspiral sources in the form of a maximum distance to which a standard
candle source can be seen. A status report will be given at the next detchar
telecon (March 8).
-
Power spectral transient monitor - S. Mohanty (report by e-mail)
The last iteration of e-mail exchanges with John didn't resolve the
data access troubles reported previously, and Soumya has been tied up with
GEO run and workshop preparations. He expects to return to this problem
in late February.
-
Transient identification & event catalog monitors - J. Sylvestre,
R. Weiss (no report)
-
Glitch monitor - M. Ito (on vacation)
-
Magnetic field and seismic transient monitors - R. Frey, R. Rahkola
(no report)
A.O.B
-
Following up on discussions preceding and during E7, Fred Raab mentioned
that a desire he had had for simple, interactive histogramming of a time
series turned out to be easy to satisfy the xmgr display of the Data
Viewer. He encouraged others to reflect on simple tools they use that might
be good to publicize.
-
Next regular teleconference: Friday March 8 at 11:00 a.m. EST (8:00
PST). Also, there will be a collaboration-wide teleconference on Wednesday
February 20 at the same time. KR has been asked to raise detector characterization
issues on February 20 for discussion at the March LSC meeting (March 20-23)
.