Minutes of Detector Characterization Teleconference
(October 24, 2003)
Present:
Caltech: Erika D'Ambrosio, Philip Charlton,
Ron Drever, Peter Shawhan, Patrick Sutton, John Zweizig
Cal. State Dominguez Hills: Ken Ganezer
Cardiff: Joe Romano
Florida: Sergey Klimenko
Hobart-Smith: Steve Penn
Italy: Alessandra Di Credico
LHO: Doug Cook, Dick
Gustafson, Mike Landry, Luca Matone, Richard McCarthy, Fred Raab,
Hugh Radkins, Rick Savage
LSU: Gaby Gonzalez
Michigan: Keith Riles
MIT: Peter Fritschel,
Albert Lazzarini, Kaice Reilly, Bernard Whiting
Oregon: Ray Frey, Isabel
Leonor, Robert Schofield
Penn State: Mike Ashley
Syracuse: Peter Saulson
Washington State: Sukanta
Bose
S3 Planning - General
Operations planning - Stan Whitcomb
Each observatory site will have local run coordinators with
authorization to make decisions on non-standard configurations and to be
on call to deal with miscellaneous problems. At Hanford, these duties will
be shared by Fred Raab and Daniel Sigg. At Livingston, they will be shared
by Mike Zucker, Brian O'Reilly and Valera Frolov. The rotation scheme will
be locally decided. Stan will announce soon the S3 teleconference schedule.
Interferometer status - (Fred Raab - LHO, Gaby Gonzalez - LLO)
During E10 H1 had a 75% duty cycle. The inspiral range was
disturbingly variable, ranging as high as 1.8 Mpc, but it was discovered
that lowering the laser power stabilized the range at its high end. It
also lowered glitch rates that were presumably due to a hidden saturation.
H2 had a 45% duty cycle during E10, having started pretty ragged. The IFO
stabilized as the week wore on, but it's not clear why. A typical inspiral
range was 600 kpc. The large, every-few-minutes glitches on H1 have persisted
and are not yet understood. The best monitor of them so far has been watching
the raw AS_Q spectrum in "movie mode'. The present DataQual glitch rates
do not pick them up well, although the DataQual band-limited RMS for that
band (around 100 Hz) does show excess at glitch times. RayleighMon picks
up about 20% of the glitches. Fred suggested adding narrower bands (40
Hz) up to about 230 Hz to DataQual's lineup. KR suggested adding a 12-sigma
glitch rate monitor to DataQual to catch what are anomalously large excursions.
John agreed to try modifiying the config file for weekend running. Fred
remarked that having the inspiral range available now in common-mode (but
pre-science-mode) running was proving convenient and requested that DataQual
report in that mode too. John confirmed he will be changing the config
file later in the day.
L1 has struggled all week, with most time given over to commissioning.
There were few good locks and almost no science mode running. Although
the source of the noise that has degraded inspiral ranges by ~10 w.r.t.
S2 is not yet known, a number of suspected culprits have been ruled out.
A laser power supply problem that added to troubles was replaced. Fred
suggested trying to lower laser power in case L1 is subject to the same
saturation trouble evident in H1.
Scimon shifts - Keith Riles
Thanks to Caltech and Hanford, six of the seven unassigned
expert shifts at LHO have now been filled. The one remaining shift unassigned
to any institution is the Sunday day shift (Nov 9) preceding the LSC meeting.
A volunteer is most welcome.
Scimon procedures - Keith Riles
Most of the scimon procedure instruction sheets leftover from
S2 were updated just before or during the first weekend of E10, with a
few new sheets added during the week. Nonetheless, a handful could use
some more attention
E10 Investigation Reports & S3 Plans:
-
Calibration
(Gaby Gonzalez, Mike Landry)
A full-up calibration was done for H1, including verification of the
DC global scale that had been propagated from S2. The LSC model was also
updated. Similar work has been done for H2, except that the DC scale hasn't
yet been checked (but will be soon). Calibration functions will be posted
later in the day for H1 and H2. It has been verified that the autocal and
SenseMon inspiral ranges agree to within a few percent, confirming that
SenseMon is correctly tracking the relevant dynamic input matrix element.
The inspiral range for H2 started the week at 800 kpc, but the addition
of a dewhitening filter to improve stability lowered the range to 600 kpc
for reasons not understood. Fred remarked that there is some skepticism
about the DC scale for H2, since at 1-2 kHz, the sensitivity comes within
2-3 of design sensitivity despite low power. Peter Fritschel wondered if
violin mode notches were doing any good to reduce excitations. Fred replied
that the notches seemed to do nothing and that large excitations are seen
even on masses that are not driven. The source of the excitations will
be investigated further before S3.
Gaby reported that no serious full calibration has been attempted on
L1 yet, but a quick n dirty attempt last weekend gave a range of about
200 kpc and provided Patrick Sutton with functions to use in SenseMon.
-
Timing (Szabi Marka, Daniel Sigg)
Szabi reported by e-mail from Japan.Briefly,
a new atomic clock system has been commissioned at both sites and seems
to be working.
-
Hardware
signal injections (Peter Shawhan, Szabi Marka)
Peter and crew have been working mostly on the pulsar and stochastic
signal injections during the last week. A number of technical obstacles
were overcome with code optimization, monitoring scripts and cron jobs.
There remains one problem with simultaneous injection of stochastic signals
since the random number seed depends on the GPS start time, requiring (at
the moment) that the jobs at the different sites start within the same
second. A scheme to make the simultaneity requirement more tolerant will
be tried out soon. Workstations in the control rooms (control7 at LHO,
control14 at LLO) are dedicated to the injections and should not be used
for other work, to avoid overloading. There remains one failure mode in
which a test point is cleared, the injections continue, but there is no
DAQ recording of the excitation. Mike will check with Dave Barker to see
if the state vector can be set up to turn off science mode if the test
point disappears. In general, Stan is concerned about an injection system
that could prove to be an operational headache and suggested considering
an increase in signal amplitude with injections carried out during only
10% of the run. KR wants to try out the full-run injections arrangement
for a while, but revisit the issue soon into the run, if failures occur
frequently. Peter has put together an information
page on the injections. Some of that info will be condensed into a
scimon instruction sheet.
-
Reduced data set generation - channel
list (Philip Charlton, Isabel Leonor)
Philip reported that the bugs have been worked out of the RDS generation
at the sites; some earlier troubles with knowing promptly when data has
been recorded have been fixed. There is now a latency of about half an
hour for level 1 RDS generation and an hour to an hour and a half for level
3 (AS_Q only) generation. The transfer of level 3 data from the sites to
CIT and then to the Tier 2 centers (UWM, PSU, MIT, AEI) has proven more
difficult because of LDR (LIGO Data Replicator) troubles and network firewalls
between LLO and CIT. There is also a bottleneck in which RDS transfers
out of CIT must await data from both sites. So the slower link holds up
the process. Because of startup troubles, level 3 RDS transfers for E10
data are about four days behind, but it is believed that the latency will
be a few hours once the transfers reach steady-state. Philip will test
this as soon as the E10 data transfers finish.
Isabel reported that the extra weather channels Robert requested have
been added to the RDS level 1 list. There are still a few channels slated
for the RDS that are not yet present in the LHO & LLO data streams.
KR asked her to send him the list of missing
channels ; he will try to find out the status/plans for those channels.
-
Data quality (John Zweizig, Keith Riles)
John has been looking at the 3-day DataQual summaries from LHO. He
has been seeing some glitch rates from H1 that appear alarming, both the
ambient rates for 4-sigma excursions which should be about 1 Hz, but which
are running at several Hz in good times, and some much larger excursions.
KR wondered how much of this was due to the apparent laser power saturation
problems (see above) seen most of the week and suggested looking at the
last 24 hours of data from after the power was tweaked down. For S3 John
is also working on a new scheme to keep track of data quality segments
in near real-time by running cron jobs to look back over trends produced
by a variety of DMT monitors. The "database" will be directories that are
filled with statistics derived from those trends. He expects to have the
new infrastructure by the start (or soon after the start) of S3. Peter
Shawhan mentioned that he will be generating a segment list for E10 shortly.
-
PEM injections (Robert Schofield)
Robert will be doing the first round of LLO PEM injections on one or
more nights of Nov 4-6. He will work with Josh Dalrymple of Syracuse who
will later do the second round of LLO injections near the end of S3. Robert
will do the first round of LHO PEM injections Nov 9-11 and the second LHO
round near the end of S3. In each round of injections at each site, approximately
four hours of nominally science-mode data is needed.
-
Glitches / Burst/Inspiral search group vetoes (Laura Cadonati, Erik Katsavounidis,
Peter Saulson, Julien Sylvestre)
Erik and Laura were unable to attend but reported by e-mail that problems
with the burst DSO's had prevented looking closely at E10 data yet, but
they expect to have some results by next Friday's detchar telecon. Their
plan for S3 is to make glitch investigation as close to real time as possible.
Julien reported by e-mail that at least qualitatively the E10 data looked
like the S2 data from the TFClusters perspective, but with larger trigger
rate variations. Peter Saulson reported on a look at DataQual glitch rates
for all three interferometers. By and large, E10 rates are running significantly
higher than in S2, consistent with John's findings above. Fred mentioned
observing the odd behavior that DataQual glitch rates actually go down
sometimes when large glitches occur. John believes this is because the
very large glitches are distorting the noise spectrum enough to hide the
rate of smaller glitches. Fred also mentioned an observation by Stan that
there are variations of an order of magnitude in the 60 Hz line in AS_Q.
Generally, stability of the line correlates well with stable inspiral range.
He wondered whether that variation affected DataQual glitch rates. John
believed not because of low-pass filtering and notches used in DataQual.
-
Stochastic search group coherence studies (Joe Romano, Fred Raab)
Joe reported that Tania Regenbau has been working on a long-term E10
H1/H2 coherence measurement, but had had some LDAS troubles and wasn't
quite ready to report. Fred reported on a several-hour H1/H2 measurement
he had done with the DTT which had not turned up any significant coherence
in the astrophysics band other than in the 60 Hz harmonics. John suggested
constructing a figure of merit for the LHO control room based on the H1-H2
coherence. Some discussion ensued on whether that was a worthwhile check,
given the tininess of the expected correlation and the little one could
likely do about an excess. KR suggested that an alarm on the coherence
rising above a certain high threshold would be useful to catch gross screwups
in the LVEA leading to coherent acoustic contamination (Fred's example:
a phone left on that rings).
-
Violin modes (S. Klimenko, J. Castiglione)
Sergey reported that the violin modes he has monitored in E10 data
have amplitudes that are consistent with thermal most of the time, with
most excursions occurring at the start of lock stretches, as expected.
Dick wondered if the beam splitter test mass vibration modes are being
monitored by LineMon, since an excess has been observed in H2 during E10.
Sergey thought they were being monitored, but would double-check. Sergey
urged anyone with line-tracking requests to contact him soon.
-
Bilinear couplings (S. Penn, V. Chickarmane)
Steve's new BicoMon background monitor was running since the second
day of E10. Although the results summaries were being written sensibly
to the html directory, he ran into the problem that the monitor occasionally
falls behind the data (clear from the time stamps in the summary files),
at which point the information sent to the DMT viewer gets zeroed out because
of an inconsistency check. John suggested using his new DMT synchronization
class to avoid the DMT viewer dropouts. Steve has implemented a config
file with cross-bicoherence channel choices suggested by Vijay Chickarmane
and Nelson Christensen.
-
Correlated inter-site environmental transients (R. Schofield)
Robert reported that the merged PEM RDS for S3 would be made at LHO
after level 1 RDS tapes are shipped from LLO. Isabel will take care of
the merging jobs. She mentioned, however, that there is presently an incompatibility
between the LDAS-merged frames and the DTT, which Robert will use in his
studies. KR will see if this can get sorted out by the start of S3.
-
Local environmental disturbances (R. Schofield)
From E10 elog entries, Robert reports that liquid nitrogen dewar creaks
are clearly causing some lock losses. Other seismic disturbances have been
reported too. Since not all elog reports of seismic disturbance state which
seismometers have been looked at, it's hard to draw conclusions about sources.
Robert urged everyone to look at seismometers from all locations when investigating
lock losses. Winds were a problem at the start of E10 at LHO. (See Robert's
PEM injections report above for S3 plans.)
-
Data Access (Peter Shawhan)
Peter summarized the data creation / transfer plans for S3. Raw data
will be written at each site to tapes which are copied and shipped to Caltech
(~once per week) for creation of level 1 (hundreds of channels) and level
2 (handful of channels) reduced data sets for distribution to Tier 2 centers.
Levels 1 and 2 will also be generated and kept locally at the sites. Level
3 (AS_Q only) will be generated at the sites and sent over the network
to Caltech and then out to the Tier 2 centers (see Philip's report above).
A data access page based on the S2 version has been linked to the S3 web
page. Volunteers are welcome to fill in some of the missing information
on accessing certain data sets.
DMT Software Status Review:
(Since the telecon had nearly ended by the time the agenda
items below came up, KR asked for only urgent reports concerning recent/planned
changes to DMT code. Some e-mail reports were also summarized.)
-
Status of DMT infrastructure
-
John Zweizig: Version updates, deadlines, etc.
There are some serious loading issues on certain
DMT machines due to a problem in the web viewer interface (web emulator
of DMT viewer). Szabi has been looking at the problem this week, but hasn't
been able to track it down yet. If it persists through next week, John
will probably shut down the web serving to avoid interference with monitoring
programs. John expects a handful of code and config file changes
in the next few days. He has set a deadline of Monday night for any code
changes to be run during S3. He will be unreachable Wednesday morning through
Friday afternoon, but Patrick Sutton has agreed to fill in for him during
that time, in particular, to restart all monitors at both sites Thursday
night in conjunction with the installation of S3 LDAS databases at the
sites.
-
Status of DMT monitors
-
Dave Chin / Tim Bodiya / Keith Riles: LockLoss
and ServoMon/SpecMon
No recent changes in LockLoss/ServoMon other
than a config file fix to allow new LHO LockLoss trends to be seen by the
data viewer. KR exercised SpecMon last weekend at LLO and sent more bug
reports and change requests to Tim Bodiya.
-
Ken Ganezer: seis_blrms
Ken asks that bug reports on seis_blrms be sent to him.
-
Masahiro Ito: glitchMon
No report.
-
Sergey Klimenko: LineMonitor and WaveMon
(reported by e-mail after the telecon)
Heterodyning is now implemented in the LineMon
version that looks at excitation channels. So far the heterodyning method
sees larger fluctuations than Sergey's standard algorithm. Sergey
attributes the increased variance to the absence of background noise estimation
in the heterodyning.
-
Szabi Marka: IRIG-B, TimeMon, and ShapeMon
No report
-
Tim Olson: SpectrumArchiver (report by e-mail)
Tim is working on a version to run under the
process manager and expects to have it ready for S3.
-
Steve Penn / Vijay Chickarmane: BicoMon / BiLinMon
See Steve's report above.
-
Brian Stubbs: SuspensionMon (report by e-mail)
The latest Hanford version of SuspensionMon ran
stably, but the Livingston config file had problems. Brian saw some odd
behavior in the H1 end mass channels he is trying to understand. Coding
work for DMT trends is not quite done. Trigger statistics don't seem to
indicate that SuspensionMon's triggers are being received.
-
Patrick Sutton / Mike Ashley: SenseMonitor / LineAmp
and RayleighMonitor
See remarks above on SenseMon. The new code seems
to be working beautifully.
-
Natalia Zotov: PTMon
Natalia was unable to connect to the telecon
(20 ports weren't enough?!). So she reported by e-mail
. Briefly, she has retuned downsampling and threshold parameters, which
seems to have produced more robust results at LLO. She will be looking
at LHO data this weekend.
-
John Zweizig: BitTest, DataQual,
HistCompr, PSLMon, SatMon , and SegGener
No time to report.
A.O.B.
-
Next detector characterization telecon: Friday October
31 at 1:30 EDT