Minutes of Detector Characterization Teleconference
(February 11, 2005)
Present:
Caltech: Duncan Brown, Laura Cadonati, Sergey Klimenko, Isabel Leonor, Vuk
Mandic, Peter Shawhan, Patrick Sutton, John Zweizig
CSUDH: Ken Ganezer
Columbia: Szabi Marka
Florida: Sergey Klimenko
LHO: Mike Landry, Greg Mendell, Vern Sandberg, Robert Schofield, Rick Savage
LLO: Gaby Gonzalez, Brian O'Reilly, Daniel Sigg
La Tech: Natalia Zotov
Loyola: Marc Cenac, John Whelan
Louisiana Tech: Natalia Zotov
Loyola: John Whelan
McNeese: Giovanni Santostasi
Michigan: Keith Riles
MIT: Peter Fritschel
Oregon: Ray Frey, Rauha Rahkola
Syracuse: Josh Dalrymple, Alessandra DiCredico
WSU: Sukanta Bose
S4 Plans & Investigations (including followups to Wednesday's E12 review):
- Scimon shift schedule (Keith Riles)
KR reminded everyone that the S4 scimon shift schedule is posted, and many
names remain to be filled.
- Calibration plans (Mike Landry, Gaby Gonzalez, Brian O'Reilly, Patrick
Sutton, Xavi Siemens)
Mike reported that measurements needed for the S4 calibration of H1 and H2
are nearly complete. A set of calibration files usable by DMT FOM monitors
and by online analysis code is expected to be ready by the end of next week.
The autocalibrator at LHO is now working again, and there's a good chance
that the photon calibrator will be workin on the ETMX mass, allowing a check
before the run of systematics on absolute H1 calibration scale. The photon
calibrator will not be operating during the run, however. Calibration injection
lines will be moved to match more closely those now used on L1. Sergey Klimenko
requested that the lines among the interferometers be closely matched so
that fewer frequency bands are affected in the WaveBurst analysis, prefererably
in the same 8 Hz band. PeterF cautioned that H1 and H2 lines should not be
as close as they were in S3 because of observed coherences in their vicinity.
Brian reported that the L1 situation is similar: nearly all measurements
done, with calibration files ready next week. In general, calibration
stability looks much better than in S3. There will not be a photon calibrator
available for L1. The strength of the 1151.5 Hz line will be increased to
ensure good SNR for short time intervals, but the strengths of the two lower-frequency
injections will be reduced. The online analyses plan to use the new GW channel
DARM_ERR for which time varying calibration should be much less pronounced
than for AS_Q, allowing the analyses to set their alpha/beta values to unity,
for simplicity. There will be no major mid-run calibration work, but time
will be used the day after the run ends to do some checks. There will be
occasional, brief open-loop-gain measurements during the run.
- Timing monitoring plans (Szabi Marka, Daniel Sigg)
Szabi has tried following up on some multi-microsecond timing jumps observed
in some LLO PEM channels during the early part of E12, but has not yet pinned
down the source of the problem. The size of the jumps is quite acceptable,
but he would like to understand the mechanism. Brian wondered if the timing
jumps had anything to do with recently observed channel hopping in the DAQ.
Szabi thought not, but will follow up on an elog entry Brian mentioned. Szabi
also mentioned that the atomic clocks at the sites agree well with the GPS
clocks.
- Reduced data set generation (Greg Mendell, Isabel Leonor, Igor Yakushin)
Greg reported that RDS generation went well during E12, despite some downstream
problems in moving data off site. One bug was found that had led to an 800-second
loss of level 3 data; it is now fixed. Isabel reported that she has collected
requests for RDS data channels and downsamplings and will circulate the latest
channel list next week. Some SPOB channels are still missing from the LHO
data stream, but are expected to be in place soon.
PeterF suggested creating a new level 4 RDS that would contain only DARM_ERR
downsampled to 2 kHz for the stochastic group. Gaby said that the inspiral
group could use such a data set but at 4 kHz. It was tentatively decided
to go with a 4 kHz level 4 for both groups, but to run it by Stuart Anderson
and the LDR people first. [Note added by KR: after much discussion among
many parties the following week, it was decided that the drawback of keeping
track of another RDS level outweighed the benefits of having the downsampled
data.] It was also tentatively decided to store not only the 16 Hz state
vector channel in level 3 RDS, but also the "activity" information (also
16 kHz) to avoid confusion over injections vs non-injections science mode.
PeterSh is working on a proposal to incorporate all needed information into
the state vector. He will circulate the proposal for comment next week.
- Hardware signal injection plans (Peter Shawhan, Vuk Mandic, Szabi Marka)
Vuk reported that the inspiral group has analyzed their E12 hardware injections
and recovered most of them, losing only some simulated to be quite distant
(>10 Mpc?). The bursts injected during E12 were the same as those used
in S3. There were also 100-minute stochastic injections on L1&A1 and
on L1&H2. Some new pulsar injection code was tested with 10 minutes of
injections on H1&L1 and a 24-hour injection on L1 only. The pulsar group
is asked to check these injections, given the injection code changes since
S3. PeterSh and Vuk will be sending out a formal request soon to the search
groups soliciting detailed injection requests. Sukanta mentioned that
the stochastic group is working on a plan that has some long injections near
the start of the run, about a 1-hour injection in the middle, with end-run
injections depending on analysis outcome of the earlier injections. Sergey
wondered whether playground data intervals would be used in S4, and if so,
whether injections should be restricted to those intervals. After some discussion,
it was concluded that such restriction would be impractical.
- PEM injection plans (Robert Schofield, Josh Dalrymple)
Robert reported that he will be doing two sets of PEM injections during S4,
the first at LHO during week one when he will be doing scimon shifts, the
second at LLO the week before the LSC meeting. Each set will be 4-5 hours
long. PeterF suggested that good use be made of LLO time during the LHO injections.
Robert said he would talk to Doug Lormand at LLO about getting some measurements
done during that time that were not completed during E12.
DMT Software Status:
- John Zweizig: Infrastructure status
(hardware & software)
John reported that there remain many DMT data dropouts at LLO. He has tried
reconfiguring shared memory on DMT machines to try to reduce page swapping.
Initial results show some reduction in loss, but not enough. He is now working
on spreading the load among the different DMT machines. He doesn't understand
why the cpus appear so loaded and suspects a newly introduced inefficiency
somewhere in the DMT infrastructure. One clue: Sergey found that BurstMon
was spending excessive time outside of the core process-data routine. KR
stated that if these measures are not successful, then we should start pruning
some monitors. He has sent a list of candidate monitors to John.
Sergey wondered if the occasional hangs of the DMT viewer could be fixed
before S4. John thought not; the mailbox communication between the DMT monitors
and the DMT viewer isn't as robust as we would like. KR suggested that operators
be given instructions on how to log into DMT machines and "nudge" non-responsive
monitors so that they resume serving DMT viewer objects. John has a nice
GUI for doing this, but it doesn't work at LLO at the moment because of some
firewall problems.
John reported that Patrick Sutton is working on a DMT viewer time stamp problem
in SenseMonitor that has become apparent only recently because of the many
data dropouts at LLO. Apparently SenseMonitor doesn't keep track of the dropouts
when filling the DMT viewer history object. John suspects the same problem
afflicts StochMon and PulsarMon, since those monitors were built on top of
SenseMonitor.
- John Whelan / Mark Cenac / Brian O'Reilly: StochMon
John reported that on their to-do list is fixing DMT trend channel names
to follow the new convention and changing the way the reference PSD's are
defined for the "other" interferometer. He plans to keep watch on StochMon
during S4 and update references, as appropriate. He has tried comparing the
Omega sensitivities calculated by StochMon during E12 with expected values
from an independently written Matlab program. Because of a technical problem,
the comparison isn't yet exact, but a quick n dirty comparison shows agreement
at the 5% level.
- Ken Ganezer: seis_blrms
Ken is trying to verify whether his latest seis_blrms program is finding
exactly the same data dropouts as seen by the older version. He sees differences
now, but the old and new versions are running on different DMT machines while
looking at the same data. KR suggested running the two versions on the same
DMT machine, e.g., fortress, and if they agree on dropouts, giving the latest
version to John for running online during S4.
- Sergey Klimenko: BurstMon
Sergey reported that the latest BurstMon is a week old and ran successfully
during E12. Following some discussions during E12, he has implemented a version
of BurstMon that does not compute hrss@50% and is thus much less cpu-intensive,
but does compute the other FOMs (e.g, pixel fraction, non-stationarity).
He proposed and it was agreed that the hrss version be run on each IFO up
to 1 kHz and that the non-hrss version be run separately for IFO up to 2
kHz. To avoid DMT trend name conflicts and DMT viewer object name conflicts,
he has appended a suffix to the non-hrss version channel names.
- Giovanni Santostasi: PulsarMon
Giovanni reported by e-mail that he doesn't yet understand the hanging of
PulsarMon and will be at LLO during the weekend to continue debugging.
- Rauha Rahkola: SuspensionMon
Rauha is stuying E12 triggers in order to refine his thresholds. During E12
there were thousands of triggers generated, many for the "high" threshold
setting. KR remarked that the high-threshold triggers should rarely be generated,
since those are supposed to tell us about optics having gone haywire. Rauha
agreed and will try to finalize the settings by next week. He mentioned that
he doesn't yet understand the apparent slow memory leak found by John in
the fall, but it appears benign enough that the program could probably run
throughout S4 without needing a restart.
- Duncan Brown: InspiralMon
Duncan ran InspiralMon successfully during the latter part of E12 and expects
it to be running at the start of S4. He will be at LHO the first week of
the run for scimon shifts and will finish ironing things out then, including
online documentation for scimons. John will be incorporating InspiralMon
into the DMT process manager lineup next week.
A.O.B.
- Next detchar telecon: Friday February 18 at 1:30 p.m. EST (S4 preparations
and DMT software)