Minutes of Detector Characterization Teleconference
(November 12, 2004)
Present:
Caltech: Peter Shawhan, Patrick Sutton, John Zweizig
Carleton: Nelson Christensen
CSUDH: Ken Ganezer
Florida: Sergey Klimenko
Hobart: Steve Penn
LHO: Dick Gustafson, Mike Landry, Greg Mendell, Fred Raab, Vern Sandberg,
Robert Schofield
LLO: Natalia Zotov
Loyola: Marc Cenac, John Whelan
LSU: Gaby Gonzalez
Michigan: Keith Riles
Oregon: Ray Frey, David Strom
E11 Preparations:
- Scimon shifts
(Keith Riles)
KR summarized the E11 scimon shift schedule. All 20 expert shifts are filled,
along with several trainee shifts. Seven visitors will be on site for part
or all of the run.
- Investigation plans:
- Calibrations (Mike Landry, Gaby Gonzalez, Brian O'Reilly)
Mike summarized calibration plans for E11. H1 and H2 calibrations will
be scheduled for Monday and Tuesday, preferably including DC actuation calibration
and leading to a handoff of calibration files to Patrick Sutton on Tuesday
night for incorporation into EasyCalibrate. Patrick confirmed he will be
available Tuesday night. John Zweizig mentioned that he has a somewhat automated
system (stand-alone C++ program) now for assembling the input xml file for
EasyCalibrate from a variety of sources in both ascii and frame format. He
will work with Mike and Patrick to get everything ready for the start of
the run Wednesday morning.
- Glitches (Laura Cadonati, Erik Katsavounidis)
Erik reported that the glitch group will squeeze E11 investigations into
the S3 analysis timeline. He wondered if WaveBurst triggers could be produced
quickly during or right after the run to guide investigations into the most
troublesome noise sources. Sergey confirmed that he and Igor can run WaveBurst
offline as soon as data and calibration information is available. After
some discussion, it was realized that the frame-format calibration alpha
and beta measurements needed for the quasi-online analysis by the inspiral
group, which will be produced daily, would serve WaveBurst needs. Sergey
will plan to start generating WB triggers during E11 running. Erik wondered
whether the new BurstMon DMT monitor could be used for even faster generation
of triggers. After more discussion, it was decided the effort to add triggers
to BurstMon would be better spent on WaveBurst, in part because BurstMon
is a single-IFO burst finder with a high false alarm rate.
- Steady-state correlations (Nelson Christensen)
Nelson and his undergraduates will plan to make and post E11 tables of
sharp lines in H1 and H2 AS_Q that correlated with environmental channels,
as has been done for S2 and S3. In addition, they will study glitches that
correlate with strong inspiral triggers from the quasi-online (once per
day) analysis run. After some discussion, Nelson was urged to give the glitch
studies higher priority than the line finding. He plans to start work the
first week of December.
- Environmental disturbances (Robert Schofield)
Robert provided a list of investigations he will pursue after the run that
he welcomes assistance on from scimons during the run. KR will post the
list on the E11 web page (when he finds the time to make the E11 web page!).
JohnZ mentioned that the RMSBands DMT monitor config file for environmental
channels may be out of date. He will send Robert the present file and ask
for corrections/updates.
- Data Quality (John Zweizig, Peter Shawhan, Keith Riles)
JohnZ will be running a new DMT monitor called Station to quantify non-stationarity
of the data and will be studying the output during the run to see if it
adds useful information beyond what other non-stationarity monitors (e.g.,
DataQual, Burstmon) now provide. He has received a new monitor called
SpectrumFold from Vladimir Dergachev which is hoped will measure the strength
of the 0.25 Hz comb now plaguing the pulsar group's S2 and S3 analyses.
On a side note, Gaby asked whether there were any photodiode saturation
DQ flags for S3 like those John produced for S2. John reported that the change
in photodiode configuration from S2 to S3 made it much harder to infer saturation
from the raw data stream, as he had done before. At the same time, though,
the increase in photodiodes made saturation rarer (presumably) and for S3
a diagnostic was put in place to measure ADC saturation. John has found
some instances of ADC saturation, and those are flagged in the current release
of S3 DQ segments.
KR wondered if both John's and Peter Shawhan's segment statistics and generation
programs would be running during E11. Both confirmed they will be. Peter's
segment table will have one significant change in that a line will be included
for any segment in progress, along with a "current'" comment. This last
feature is needed for the quasi-online analysis for which Greg Mendell is
writing scripts to query Peter's tables. KR will post the zeroth-order E11
DQ segments in the usual web directory once the run ends.
- Critical DMT issues:
- John Zweizig: Infrastructure status
JohnZ reminded everyone that the nominal deadline for final code revisions
is today, but offered to accept some revisions as late as Monday. Fred Raab
wondered if anyone will be checking for trend channel name integrity / consistency
before the run, to avoid any frame builder crashes, given the problems seen
in the M5 mini-run. John will work with Dave Barker to fix residual problems
and verify no new ones have arisen. John believes that only SuspensionMon
still has a problem with channel names (blanks in the string). Ray Frey
said he would make sure Rauha Rahkola gets the names fixed over the weekend.
John mentioned he is working with Duncan Brown on getting results of quasi-online
analysis into the control room via display by the DMT viewer. He is writing
a DMT monitor that will read trigger files (xml) produced by analysis jobs
and create graphical displays in the usual format (e.g., trigger rates vs
time). The monitor will be config file driven, allowing use with different
offline search programs. He and Duncan hope to have a test run of the system
during E11 for the inspiral search. Gaby pointed out that the latency of
inspiral job results will be 12 hours or more. KR suggested including a mechanism
to alert a user to periods (most recent past) for which job results
are not available, to distinguish zero trigger rate from absence of information.
- Patrick Sutton: Calibration infrastructure
Patrick reported that he hopes to add a method over the weekend to EasyCalibrate
to allow retrieval of calibration line amplitude for use by SenseMonitor
during the run. Otherwise it will be unable to report the amplitude to the
DMT viewer or in trend files, although the amplitude information will have
been used internally by EasyCalibrate to produce a correctly calibrated strain
spectral density estimate
.
- John Whelan / Mark Cenac / Brian O'Reilly: StochMon
JohnW reported that the inclusion of the overlap reduction function is
all but done and will be in the cvs archive by the end of the weekend. KR
wondered if John had been able to confirm quantitatively that the Omega
sensitivities quoted by StochMon during M5 were correct. JohnW said he hadn't
been able to retrieve the necessary spectra files produced by StochMon during
the run and wondered where they had been stored. JohnZ offered to help him
find the files.
- Ken Ganezer: seis_blrms
Ken reported that he believes his code is working, but he hasn't had time
to test it thoroughly on fortress. KR urged him to get it shaken down over
the weekend.
- Sergey Klimenko: BurstMon
Sergey reported that the problem in M5 with an invisible summary html file
has been fixed. So has an overall factor of 4 in reported sensitivity due
to an oversight in defining seed hrss50 values. The false alarm rate the
monitor will report sensitivities for will be about 0.5 Hz in E11. He asked
for guidance in arranging online documentation of the monitor. KR suggested
(for now) putting in a bare-bones BurstMon.html file with a brief description
of the algorithm and a link to the extensive documentation page on a U. Florida
web page.
Sergey uses a 1-minute time stride in BurstMon and would like to run on
a machine with a 1-minute memory buffer to avoid data losses. JohnZ said that
the gypsum machine already has that size buffer, and travertine could be
set up for it too. KR suggested that if CPU usage became critical, one could
stop running WaveMon or run it offline for trigger generation. (Note added
later by KR: JohnZ found that the data losses in BurstMon seem to be caused
by excessive trapping of floating point underflows, a problem to which LineMon
is also susceptible.) Sergey plans to do waveform injections for five different
waveform types when measuring hrss50.
- Giovanni Santostasi: PulsarMon Sample plots:
1
2
Giovanni reported by e-mail that he should have final E11 code to John
by Monday, with new figure of merit plots showing discrete values for known
pulsars. Sample plots from the DMT viewer were provided, showing target strain
sensitivites for known pulsars in order to reach spin-down limits and ellipticity
sensitivities for those pulsars for a 1-year integration at current noise
levels. The incorporation of EasyCalibrate isn't quite working yet. He will
give John a version based on the old SenseMonitor code for E11 running, if
he can't track down the EasyCalibrate problem over the weekend.
- Rauha Rahkola: SuspensionMon
No report. Ray will talk to Rauha about the residual channel name troubles.
S3 Calibration Status
(Mike Landry, Gaby Gonzalez)
Mike summarized the status of the S3 calibration for all three IFO's, giving
a talk that had been prepared for the LSC meeting last weekend but was not
given, for lack of time. He mentioned that the review committee has been supplemented
by John Zweizig, given the committee's workload. Briefly, most of the modelling
has now been reviewed successfully, with a small number of tasks to be completed.
Validation is nearly complete for Xavi's demodulation technique for determining
alpha/beta, and validation of his h(t) frames has begun. An example of the
agreement between modelling, measurements, and fits was shown for one dewhitening
filter. Cross checks include computing imaginary parts of coefficients that
should be real but may not be if modelling is wrong, time evolution of imaginary
components, and consistency between calibration determined by a ~150 Hz injected
line and the usual ~900 Hz line. Cross checks work very well for H1 and H2,
but there is a tiny discrepancy (~1 degree phase) in L1 for the ~900 Hz line.
Final results are expected in 2 weeks (again!). Although old S3-era calibrations
were used for M5, updated calibrations will be performed for E11 (see above).
For S4, the same DC calibration methods will be used, but others may also
be explored to beat down the dominant systemic uncertainty. The plan is to
have good reference measurements in place <i>before</i> the run
starts to allow reliable online calibration and data analysis. There will
also be real-time comparisons for different methods to obtain coefficients.
By then the photon calibrator should be ready for testing as a cross check.
A.O.B.
- Next detchar telecon: Friday November 19 at 1:30 p.m. EDT