Minutes of Detector Characterization Teleconference
(October 4, 2002)
Present:
Caltech: Marka, Shawhan,
Zweizig
Hobart-Smith: Penn
LHO: Landry, Matone, Raab,
Savage, Schofield, Sigg
LLO: Coles, Gonzalez, Zotov
LSU: Daw, Johnson
Michigan: Riles
Oregon: Frey, Rahkola
Penn State: Sutton
Wisconsin: Brown
S1 Reports: (S1
Web Page
)
- Calibration Status (M. Landry)
Results from a number of calibrations performed before, during and after
S1 have been posted on the
calibration web page
. The page also has links to a document on calibration drift correction by
Gaby Gonzalez, and sample plots by Patrick Sutton and Sergey Klimenko
from tracking the injected calibration lines. Mike is working with Rana to
finalize L1 computed sensitivities and is working with Patrick Brady and
others on providing time-dependent parameters to allow correction in astrophysical
search engines for calibration drift. Presently the inspiral group uses raw
frequency series information, and the burst group uses the results of pole/zero
fits. Patrick Sutton is rerunning his new version of SenseMonitor over the
entire data for the three interferometes and will post the results and archive
them in the DCC. Mike will give an update on calibrations at the October
18 detector characterization teleconference.
- Data Quality (J. Zweizig, E.Daw)
John has compiled a
large number of plots
of band-limited RMS and of glitch rates from the three interferometers
during S1. His plan is to define objective criteria for vetoing "bad data".
Auxiliary channels will be sought with which pathologies in AS_Q can
be correlated without biasing, for example, searches for astrophysical bursts.
John's plots (1-minute binning) show clearly some periods of increased glitch
rates in the H1 AS_Q channel he is investigating. John is rerunning the glitch
monitoring with power line filtering included to reduce overall rms. Preliminary
results show the expected average ~1 Hz rate, given Gaussian noise
and a low threshold setting. John mentioned that the saiph DMT workstation
at Caltech has completed its task of S1 tape copying and is now available
for running DMT jobs. saiph can see all of the data from both sites.
Anyone with a Caltech general computing account should be able to access
saiph. In order to help the upper limits analysis groups pre-filter
data for the pre-Nov 1 analysis, John will finalize in the next two weeks
the criteria to use in defining new segment tables in the database (more
refined version of what his online SegGener DMT monitor creates). KR
offered to query the detector characterization
group
for any favorite data stretch veto criteria people have found.
Ed has been reworking his PeakMon DMT monitor plots into the form of
2-dimensional histograms
(jpeg and png formats) with time as one axis. He finds that with a
high-pass filter, he sees periods of quite clean tails in the histograms,
with occasional pathological periods. He finds the 2-d histogram bin contents
to give better resolution than his former contour plots. The histograms are
produced by a matlab script that reads in summary data produced by PeakMon
and have 16-second binning of filtered (140Hz high-pass) AS_Q. Duncan
has looked at some of the glitchy periods Ed sees and finds them correlated
with periods of increased inspiral candidate rates. For example, an average
of 20-30 candidates / 256 seconds jumps occasionally to about 400 / 256 seconds
for L1. These candidates are characterized by high signal-to-noise but (fortunately)
also high chi-square for looking like true inspirals (binning by frequency
band. John will give an update on the data quality work at the October 18
detector characterization teleconference.
Looking ahead to S2:
- The S2 run is nominally slated to begin February 14, 2003 and last
for about eight weeks. KR brought up a number of issues to work on between
now and then:
- Fixing and refining DMT monitors - Although great progress
was made between E7 and S1, there were still many start-up troubles with
DMT monitors in S1. A number of new and useful features had been added
to monitors in the weeks leading up to S1, but the features hadn't been shaken
down thoroughly. Config files were out of data, and so on. We need to start
S2 in better shape.
- Figures of merit to watch in the control rooms - A
number of new DMT viewer plots were used in S1 to keep track of interferometer
performance, some of which worked well and others of which were flaky or at
times misleading. The known problems need to be fixed, and new figures of
merit (e.g., time histories of many AS_Q bands) commissioned.
- Alarms - We chose not to enable audible DMT alarms
in S1 for fear of swamping operators and scimons with false alarms that would
soon be ignored. But we should be aiming for reliable alarm generation (occurring
rarely but correctly). We need a mechanism for deciding which DMT alarms
we trust and choose to enable.
- Graphical real estate in control rooms - KR felt strongly
that more wall space could have been profitably used during S1 for displaying
DMT monitor outputs and asked whether more could be made available for S2.
Fred and Mark said that projection display area could be increased by about
a factor of two. Peter Shawhan stressed the need for standardized plots where
one knows what to expect at a glance and can quickly spot deviations. The
standardization would include pre-set axis limits, placement, etc..
- Shakedown engineering run E9 - After some lengthy discussion
a consensus emerged that before the S2 run there should be a set of "mini-runs"
scheduled during convenient weekends in the late fall at the sites. These
runs would not necessarily be held simultaneously at LHO and LLO, being somewhat
opportunistic w.r.t. ongoing commissioning work, and they would be short,
perhaps no longer than several hours, just long enough to collect representative
IFO data and to test DMT monitor improvements/fixes.A more formal E9 engineering
run would then be held sometime in the month preceding S2, in order to shake
down the DMT monitors for primetime science running. This would be different
from E8 during which DMT authors were strongly encouraged to add new features
to monitors. Only bug fixes and config file tuning are envisioned for E9.
Fred Raab pointed out that despite a 2 1/2 month interval between E8 and
S1, new, improved monitors that were nominally declared ready for primetime
weren't and argued for holding E9 sooner rather than later, with a formal
DMT code freeze well before the start of S2. KR volunteered to
query the detchar group
for suggestions and come up with a specific run schedule proposal.
DMT (or other) Software Status Review:
Since time was running short after the above discussions, KR asked DMT
authors to report only significant code changes since S1:
- Patrick Sutton reported having fixed up and tested SenseMonitor to
use the 973 Hz injected calibration lines to correct the raw computed inspiral
ranges (see above). Calibration line tracking seems to be working well.
- Duncan Brown has written a new tool based on xmgr graphics for displaying
the results of inspiral search engine results in real-time. This tool will
soon become part of LigoTools and be available in the control room.
A.O.B
- KR inquired on behalf of Albert Lazzarini as to who from outside
LIGO Lab would be attending the GWDAW meeting in December in Kyoto. Of those
on the line, only Patrick Sutton planned to attend.
- Next meeting: October 18 telecon at 1:30 p.m. EDT. In order
to support the ongoing upper limits analysis of S1 data in a timely way,
KR proposed to hold detchar telecons more frequently than once per month
during the fall and to schedule reports from the S1 investigation teams,
with more urgent investigations reporting first. Proposed telecon dates
were Oct 18, Nov 8, Nov 22, and Dec 13. KR will propose a report schedule
to S1 team leaders.