Minutes of Detector Characterization Teleconference
(September 7, 2001)
Present:
Caltech: Kells, Matone,
Whitcomb
Carleton: Christensen
Dublin: Ottewill
Florida: Klimenko
Louisiana Tech: Greenwood
LHO: Butler, Gustafson,
Landry, Raab, Rahkola, Savage, Schofield
Michigan: Riles
MIT: Fritschel, Shawhan,
Sigg
Oregon: Leonor, Strom
Syracuse: Penn
Preliminary E5 Reports:
-
Lock losses - KR
Details and figures can be found here.
In summary, there were three main categories of lock loss seen in E5 (plus
one smoking-gun earthquake-induced loss from the south pacific):
-
13 Hz runaway oscillation in Michelson servo control signal that was improved
with better DC alignment of the recycling mirror and is attributed to loss
of optical gain from tidal drift in alignment.
-
Saturation of error signal for common mode servo after tidal drift.
-
"Deliberate" software railing of servo control signals after tidal drift
due to necessary limits imposed during lock acquisition, which in the future
will be widened after acquisition.
These three types of loss should be much less important once the 2K IFO is
fully commissioned with 1) feed-forward tidal compensation (common &
differential mode), 2) wave-front-sensing control of the recycling cavity
alignment, 3) higher unity gain frequencies in the Michelson and Recycling
servo loops, and 4) higher DC gain in the arm common mode loop.
-
Tides - D. Strom
Details and figures can be found here.
In summary, the common and differential mode control signals from E5 were
fitted to a set of offsets and nominal tidal parameters, including
temperature parameters for the oven used in common-mode tidal compensation.
Significant deviations from expectation are observed, especially in the
common mode. Had lock stretches lasted longer, it is strongly suspected
that residuals would have been still larger. Significant phase lags are
observed. The numbers suggest the understanding of common mode arm control
in this model is off by as much as a factor of two.
-
More on tides - F. Raab
Because the common-mode tidal compensation was turned on (and varying)
for part of the run and because the thermal modelling of that compensation
via reference cavity temperature is thought to be problematic (see above),
it was decided to fit only the differential mode arm control signal to
the differential mode tidal model, with one global fudge factor as a free
parameter. A plot
was shown of the fit's chi**2/dof vs the fudge factor which should
ideally be unity from "first principles". The fit prefers a value of 0.93
with 1 sigma limits of 0.8-1.1. This seems compatible with David's corresponding
coefficient of 0.85 above. Fred emphasized that the temperature control
of the reference cavity is now much improved since E5, after recent work
by Rick Savage and Hugh Radkins, with better coupling between heater and
cavity and with better stability.
Transients reference web page - R. Rahkola, M. Ito (Rauha reporting)
-
A web
site has been set up for storing and displaying sample transients seen
in various interferometer and environmental channels. Some examples have
been posted, and scientists are highly encouraged to submit others to the
site, using the web interface buttons. Eventually, there will be automated
posting of information, but for now Rauha and Masahiro will take care of
posting what is submitted. There was some discussion of what numerical
parameters, such as detection efficiency, should be displayed and several
suggestions made, including linking in documents that describe algorithms
used to detect transients (see above too) and providing both distilled
and raw information on transients.
Status reports on DMT infrastructure:
-
General - J. Zweizig (not present)
KR reported that John has recently written infrastructure for creating
data segments in the meta-database that allows specification of "good"
data via operational state conditions. For the moment the definition is
simply that both arms are in lock. KR also mentioned that he and John are
discussing how best to improve documentation of DMT monitors and infrastructure
packages, and how to improve access to that documentation. An e-mail with
guidelines to DMT authors will be issued soon. One guideline will include
placing a definitive (but perhaps evolving) document in the DCC for each
monitor or package.
-
Histogramming - M. Ito
Masahiro has implemented xml support in the DMT viewer histogram GUI
control. This will allow saving of histogram data and the parameters used
in displaying them, much like what is provided in the Diagnostic Test Tool.
Status reports on DMT monitors:
-
Line noise (multi-taper method) & inter-channel correlation monitors
- B. Allen, A. Ottewill (Adrian reporting)
There is not much to report since delivery last month of the multitaper
line tracking code. The revised inter-channel correlation monitor code
has been waiting for histogram support in the DMT viewer. Adrian hadn't
been aware that Masahiro has completed the first version of that infrastructure
and was eager to learn more about it. He has been using a home-grown histogramming
program in root for the time being. Histograms will be made available in
the next releases of both monitors.
-
Line noise (quasi-monochromatic method) monitor - E. Daw, S. Klimenko
(Sergey reporting)
Not much has changed since E5 when the latest line monitor was exercised
in a private directory. The next version will be runnable by the central
DMT process manager, but Sergey and John have to coordinate on a revised
CVS system for DMT code management. Sergey plans to add trigger
generation to the monitor when glitches in amplitude or phase are seen.
He remarked that the code can look at only a finite number of lines and
channels and solicited advice on which lines were most important to measure.
KR suggested injected calibration lines, 60 Hz & harmonics, and violin
modes, including harmonics, as deserving high priority.
-
Band-limited / seismic noise monitors - E. Daw (not present)
-
Bilinear cross-couplings & broad-band non-Gaussianity monitor - S.
Penn
Before the recent LSC meeting, Steve had sped up his code with
a change in the way memory is allocated. Unfortunately, that change seems
to cause the program to crash ungracefully when it ends. Steve has tried
several possible fixes without success and is now following up a suggestion
by Kent Blackburn that a buggy gcc compiler is at fault.
-
Lock transition and servo instability monitors - D. Chin, K. Riles
(KR reporting)
Documentation for both monitors is being revised, and a problem with
buffer overflows in data transfer to the DMT viewer, seen in E5, is being
investigated.
-
Rayleigh behavior monitor (non-Gaussianity vs frequency) - S. Finn,
G. Gonzalez, P. Sutton (not present)
-
Power spectral transient monitor - S. Mohanty (not present, reported
by e-mail)
A DatEnv version of the existing transient detector class is about 70%
done. A complete monitor should be ready for first testing within a week.
-
Transient identification & event catalog monitors - J. Sylvestre,
R. Weiss (not present)
-
Glitch monitor - M. Ito
To make the monitor more flexible, the thresholds applied to channels
can now be set individually via a config file at run-time.
-
Magnetic field and seismic transient monitors - R. Frey, R. Rahkola
(Rauha reporting)
The earthquake monitor was commissioned shortly before E5, and although
it failed to detect the south pacific quake on Sunday night of the run,
it has detected (correctly) a number of other quakes since then. A revised
version of the monitor with lower thresholds and a longer time scale for
computing ambient noise level does now detect the E5 quake. KR suggested
in order to make the monitor more robust against distant quakes with slow
rise times that the monitor have absolute thresholds for triggering in
addition to those determined on-the-fly from ambient conditions. Rauha
mentioned that he also plans to add band-limiting methods to make the monitor
more robust.
A.O.B.
-
Next teleconference: Friday October 5 at 11:00 a.m. EDT (8:00 PDT).