Present:
Caltech: Szabi Marka,
Peter Shawhan, Stan Whitcomb, John Zweizig
Florida: Sergey Klimenko
Hobart: Steve Penn
LHO: Dick Gustafson,
Masahiro Ito, Mike Landry, Fred Raab, Robert Schofield
LLO: Brian O'Reilly
Michigan: Keith Riles
MIT: Laura Cadonati,
Mayur Desai
Oregon: Ray Frey, Rauha
Rahkola
Penn State: Mike Ashley,
Patrick Sutton
UWM: Xavier Siemens
S2 Status Update Reports
John continues to run a saturation-detection monitor on the H1/H2
data, looking for ADC/photodiode saturation as inferred from the AS_Q channel
and known filtering. The L1 search completed a while ago, but the sheer
volume of LHO data has slowed the H1/H2 jobs. At best, the monitor runs
at two times real-time. It's unlikely that the H1/H2 jobs will complete
before release of version 3 DQ segments. On the other hand, the DataQual
trends from all three IFO's are complete, and John expects to determine
in the next two weeks conservative thresholds on glitch rates and band-limited
RMS for throwing out data with a badly malfunctioning detector. That information
will be provided in version 3 segments.
KR gave a brief update on the status of the DQ segment repository. Gaby has reorganized the segment-indexed trend plot pages and is looking at large AS_Q outliers. Doug Lormand has determined that, despite what we thought at the time, we had no working magnetometers in the LLO end stations during nearly all of S2, which means that we have no vetoes of very local magnetic disturbances (other magnetometers on site do provide vetoes of truly intercontinental magnetic disturbances). Although a handful of scimons noted during the run that the EX magnetometer spectra didn't look quite right, the problem was not glaringly obvious because magnetometer channels and unconnected cables both display prominent 60 Hz combs. We should learn from this experience to ensure that PEM channels are checked out carefully before S3 and reliable reference spectra provided.
Planned for version 3 segments: cuts on DataQual glitch rates / band-limited RMS; cuts on calibration reliability, cut on known period of bad timing in LLO EY station, miscellaneous problems in H1/H2 already treated for L1 (e.g., end-of-lock noise degradation). Some discussion ensued on the LLO bad timing period, which was also characterized by nasty glitches. The consensus reached was to veto all of the LLO data between the time of a known power glitch, which is when it's believed the GPS antenna connection in EY was lost and the time when a backup antenna was connected. Brian offered to determine the GPS times for those events, and Szabi offered to look more closely at the start time to see if there is any confirming evidence of a timing problem starting then (e.g., an offset in the phase of a calibration line). Because the power glitch led to various locking troubles, the veto of the affected interval doesn't affect much science mode data. [Note added after meeting: Brian determined the approximate nominal start / stop times to be Mon April 7 17:00 UTC (GPS 733770013) and Wed April 9 7:06:32 UTC (GPS 733907204), corresponding to L1 segments 571-575, inclusive. The total science-mode data lost is 2821 seconds.]
Planned for version 4 segments: Cuts on inferred ADC/PD saturations
in H1/H2, large violin mode excitations (see below). Additional information
on artifacts found by the glitch investigation team may also be available
in time for version 4.
All point calibrations are complete for all three IFO's, and the
curves have been provided to Patrick Sutton who will generate revised alpha/beta
coefficients. Patrick estimates (fairly conservatively) that it will take
a few days to determine where to make transitions from one reference curve
to the next, and then another week or two to generate the coefficients
themselves. He expects to have a full set of coefficients for all three
IFO's within three weeks, to replace the curves and coefficients generated
in real-time during the run.
There was some discussion of running multiple jobs to speed up the coefficient-generation step and of the scarcity of DMT machines that can see the full S2 RDS. John agreed to try mounting the S2 data (temporarily) on the LHO sand and stone machines to allow not only Patrick's jobs to run, but also reruns of glitchmon(Masahiro) and LineMon(Sergey) to occur. KR noted that the current bottleneck in rerunning DMT monitors on S2 data would be greatly relieved by support of DMT machines at the various Tier 2 computing centers (UWM, PSU, MIT). A proposal has been made to Lab and Collaboration management to organize such support.
In comparing the new calibration with the online version from S2, the
calibration team finds a worst-case (H1) discrepancy of 20% and expects
that if another (final) recalibration is done later, that its difference
from the present calibration should be no worse than several percent. Because
it may be another three weeks before coefficients for the present calibration
are available, the team urges search groups wanting to get started on S2
analysis to use for now the online calibration, since it's already good
at the 20% level. Xavier, who is working on the pulsar analysis, readily
agreed to proceed using the online calibration and requested that ascii
files of the coefficients be provided from the existing SenseMon trend
frames. MikeL agreed to generated those files immediately. The calibration
team also urges the search groups to speak up very soon if their requirements
(e.g., binning, file formats) for S2 differ from those used in S1.
A.O.B.