Minutes of Detector Characterization Teleconference
(October 31, 2003)
Present:
Caltech: Philip Charlton, Ron Drever, Albert
Lazzarini, Syd Meshkov, Kaice Reilly, Peter Shawhan,
Patrick Sutton, Stan Whitcomb
Cal. State Dominguez Hills: Ken Ganezer
Cardiff: Joe Romano, Tania Regimbau
Florida: Sergey Klimenko
Hobart-Smith: Steve Penn
LHO: Stefan Ballmer,
Mike Landry, Luca Matone, Vern Sandberg
LLO: Valera Frolov,
Andri Gretarrson, Brian O'Reilly, Natalia Zotov
Michigan: Keith Riles
MIT: Laura Cadonati,
Alessandra Di Credico, Erik Katsavounidis
Oregon: Ray Frey, Isabel
Leonor, Rauha Rahkola, Brian Stubbs
S3 Planning - General
Operations planning - Stan Whitcomb
Earlier in the week Stan circulated instructions
for joining weekly S3 telecons at 5:00 EST (2:00 PST)., where general run
status will be reviewed and proposed configuration changes discussed. Because
L1 's performance is rapidly changing (good news!), there will likely be
some continued commissioning during the first week of S3, but unlike in
E10, there will be significant time set aside for science mode running.
Mike Zucker will make the call each night.
Interferometer status - (Mike Landry - LHO, Valera Frolov - LLO)
H1 and H2 are both locking readily, with decent duty cycles.
H1 has inspiral ranges spanning 1.7 to 2.5 Mpc, where the variation is
dominated by fluctuations in somewhat broadband noise centered on 100 Hz
(not the infamous 100-150 Hz glitches, however, which remain mysterious).
H2 is running with typical inspiral ranges of 800-900 kpc. DC calibrations
on both IFO's were completed yesterday. Swept sines will be carried out
tonight, with calibration line injection amplitudes to be tuned today.
The calibration lines are injected from the operator 4k and 2k workstations
via awg scripts (unlike signal injections). Dave Barker has redefined the
LHO state vector bit to require the excitation test points for calibration
lines and pulsars to be defined for science mode to apply. This safeguard
is to ensure that any injections during science mode are properly recorded
in the data.
L1 inspiral ranges are now getting above 1 Mpc after fixes within the
last several days. Lock stretches of half to one hour have been observed,
but have been interrupted by commissioning work. It is thought that several
hours can be achieved. The IFO seems relatively stable (not well quantified
yet), with low glitch rates. Calibrations have not been seriously addressed
yet, but work will begin tonight. Baseline measurements and determination
of injection line amplitudes are the main goals in the first night. The
"up" script for reaching science-mode-ready state is not quite final. Brian
relayed a statement from Mike Zucker that science mode running will be
emphasized over commissioning in the coming days, if all goes well.
Scimon shifts - Keith Riles
As was true last week, one shift (Sunday Nov 9 day shift before
LSC meeting) remains unassigned to any institution. A volunteer is most
welcome.
Scimon procedures - Keith Riles
There have been a number of updates of instruction sheets in
the last week, but work still remains. KR will be doing some updating at
LHO over the weekend. Stefan raised several issues that had come up in
going through the checklist for today's LHO day shift:
-
The script for checking that SpectrumArchiver is functioning properly seems
to be missing. KR will send Tim Olson a note on it.
-
The pages for checking the status of inspiral and burst DSO's running in
LDAS don't seem to be working. Albert said he expected things to be up
and running in the next day or so.
-
There was some discussion about the burst group's plans. In S2 the group
produced nice summary web plots in near real-time of DSO results. Erik
reported that, except for WaveBurst, the burst group's DSO's are not quite
ready to run on real-time streaming data, but will be soon. Once those
are running, then the old web infrastructure can be used again. WaveBurst
requires coincident streams from both sites. TFClusters and the PowerExcess
DSO's will look at single-IFO data. All will rely on the level 3 reduced
data set (AS_Q only). Sergey mentioned that he planned to create some infrastructure
for displaying WaveBurst results on the web.
E10 Investigation Reports & S3 Plans:
-
Glitches
/ Burst/Inspiral search group vetoes (Laura Cadonati, Natalia Zotov)
Laura summarized a study of the E10 glitches reported by glitchmon
on the three IFO's. Using Peter Shawhan's segment list, she looked at 117
hours of H1 data (358 AS_Q glitches), 62 hours of H2 data (917 AS_Q glitches),
and 2 hours of L1 data (76 AS_Q glitches). In each case triggers from every
non-PEM auxiliary channel monitored by glitchMon were evaluated for coincidence
with AS_Q glitches. Dead-times, veto success (fraction of auxiliary glitches
matched to AS_Q glitch), and veto efficiency (fraction of AS_Q glitches
matched to auxiliary channel glitch) were evaluated. Time-lag histogram
plots were also produced and displayed on the results page if a statistically
significant peak at zero time lag was observed. Summary:
-
H1: No single, dramatic correlation, but the following channels
showed significant correlations: AS_I, POB_Q, POB_I, REFL_Q, SPOB_MON
-
H2: The POB_I channel is quite important here, with 55% veto success
and 25% efficiency and only 0.06% deadtime. AS_I is quite similar, with
lesser but still significant help coming from AS_DC, AS_AC, POB_Q, REFL_(I/Q/DC),
and QP(DX/DY)_DC.
-
L1: Four channels are found to be very important: POB_I, AS_I, MICH_CTRL,
PRC_CTRL, with MICH_CTRL giving veto success of 100%(!) and efficiency
of 59%, PRC_CTRl giving veto succcess of 95% and efficiency of 84% (!).
Caveats: only 76 AS_Q glitches were available to be studied, and the IFO
is a very different beast now from what it was in E10.
Laura also mentioned that Julien Sylvestre has generated TFClusters
triggers for E10, but she did not believe they had been analyzed yet.
Natalia reported that a third of the glitch triggers found by PTMon
in H1 data are in the worrisome range of 80-120 Hz. She finds correlated
glitches in several other primary IFO channels in most cases.
-
Calibration
(Mike Landry, Brian O'Reilly)
See above remarks under IFO status.
-
Timing (Szabi Marka, Daniel Sigg)
No update from last week's report.
(Szabi in Japan, Daniel on shift tonight).
-
Hardware
signal injections (Peter Shawhan, Szabi Marka)
The mechanics of hardware injections are working well now after the
work in E10. A problem was discovered this week, however, in which numerical
precision errors in LAL (code now fixed by Teviet Creighton) were causing
the pulsar injection code to inject line signals at other than at the 10
chosen pulsar values. Peter discovered the problem when looking at FFT's
of the excitation channel. The code seems to be working fine now. Sukanta
Bose is working on revisions to the stochastic injection code and on the
setting of amplitudes. When that code is ready, a request will be made
at the weekly S3 telecon for dedicated injection time. It had been hoped
to carry out a series of burst/inspiral injections over the last couple
of nights, but the IFO's did not cooperate. Setting aside time for those
injections will be discussed at this afternoon's S3 telecon.
It is no longer necessary (or desirable!) for operators/scimons to
restart stopped injections. A checking script should either verify that
injections are running or reveal why they are not (e.g., other injections
going on temporarily). Any other outcome should be reported by e-mail to
Peter. Two new workstations have been installed in the LHO mass storage
room, dedicated to injections into the IFO's and freeing up the control
room workstation previously reserved for these.
-
Reduced data set generation - channel
list (Philip Charlton, Isabel Leonor)
Philip reported RDS generation is going well at Hanford, but that the
level 2 generation will need to be restarted at LLO because of a channel
list update. During the telecon data started flowing from LHO to Caltech
at 10 times real-time so that we should be reaching steady-state soon.
There was no opportunity this week to measure total latency between data
taking and transfer to Tier 2 centers, but that measurement will be available
soon and is expected to be an hour or so. (Note added next day: latency
from LHO/LLO to CIT stabilized at 1.5 hours. Additional latency from CIT
to other sites not yet available.) Raw data tapes will be shipped from
the sites to Caltech roughly once per week where level 1 and 2 RDS will
be regenerated for storage there and distribution over the network to Tier
2 centers. Level 1 RDS tapes will also be made for each site and shipped
to the other, giving a total delay of up to two weeks between data recording
at one site and RDS appearance at the other.
Isabel reported that there were a few last-minute changes to the level
1 RDS channel list because some channels had been renamed or had not been
made ready in time. The channel list should now be considered final. Daniel
has fixed a problem in the DTT that prevented reading of merged, multi-IFO
RDS frames.
-
Data quality (John Zweizig, Keith Riles)
John could not make the telecon, but KR reported that John is testing
out 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. KR has promised some scripts for viewing and annotating segment
information, but has not gotten far yet on the coding.
-
Stochastic search group coherence
studies (Tania Regimbau)
Tania showed a coherence plot of H1/H2 for the entire E10 run, averaged
over 30-second intervals (bin size 0.03 Hz). There is strong structure
at the 60 Hz line harmonics and at very low frequencies (below ~30 Hz),
as expected qualitatively. The flat noise floor is about where it should
be for this length of time average. Aside from a handful of other small
peaks, the coherence looks very nice, a dramatic improvement over what
was seen in the S2 run. Peter Shawhan pointed out that the injected pulsars
do not appear to be showing up in the coherence, a reassuring indication
that those injections are truly non-intrusive. He suggested that someone
from the pulsar group compute the expected coherence from these injections
as a diagnostic.
-
Violin modes (S. Klimenko)
Sergey is monitoring all of the major violin mode harmonics on each
IFO. A problem in LineMon that was causing loading of its DMT machine during
E10 has been fixed with a refined heterodyne algorithm. Patrick noted,
however, that even with the fixes, there is only 10% idle CPU on the LineMon
machine (obsidian) at LHO when both IFO's are in science mode. He (or John)
may have to move monitors around to ensure sufficient headroom on all machines.
Patrick wondered how one would know if LineMon were losing data. Sergey
replied that the DMT viewer plot should show a gap, the html summary page
would show inconsistent total times, and its machine's spi page would show
lost frames. (BrianO mentioned that a problem was seen with seis_blrms
in E10 when it could not keep up with the data, in which there would be
apparent seismic spikes on the figure of merit screen, leading to scimon
confusion.)
Sergey has added a new heterodyne option to LineMon with an anti-aliasing
filter to try to reduce the large fluctuations seen in that line tracking
option in E10. He wondered when he could try it out at the sites. KR suggested
that he bring it up at this afternoon's S3 telecon and imagined it would
be a good thing to try in a few days, after things settle down a bit. Rauha
wondered whether LineMon should provide audible alarms on missing calibration
lines. Sergey said that his standard measure of line strength would do
so, but not the heterodyne option recently commissioned. Since both are
running, we should be okay. KR remarked that if it is established within
the first week of running that the heterodyne determination adds no value
to the standard technique, we may want to disable that tracking option,
relieving the load on obsidian.
-
Bilinear couplings (S. Penn, V. Chickarmane)
Steve has a new BicoMon background monitor nearly debugged (hope to
finish today) that allows monitoring of multiple bilinear couplings measures
in one program. KR was leery about starting up significantly new code after
the start of S3, but suggested Steve talk to John and raise the issue at
this afternoon's S3 telecon.BicoMon does not (yet) provide a figure of
merit display; so changes are not critical to operations, but bugs are
a worry.
-
Correlated inter-site environmental transients (R. Schofield)
No update (Robert on owl shift tonight)
-
Local environmental disturbances (R. Schofield)
Ditto.
-
Data Access (Peter Shawhan)
Peter is working on an interface problem that prevents the DTT from
reading new RDS frames via LARS. He mentioned that the fix that allows
the DTT to read old merged RDS frames may not appear in LIGOtools for a
while.
DMT Software Status Review:
-
Status of DMT infrastructure
-
Patrick Sutton (filling in for John Zweizig): Version
updates, deadlines, etc.
All DMT monitors were restarted Thursday night
at both sites. All DMT authors are urged to verify that their monitors
are working properly! As mentioned above, there is serious loading
of two LHO machines (obsidian & gypsum) that may prove troublesome
and require moving monitors around. All of the new monitoring code received
by last Monday night's deadline has been installed and is running. There
seems still to be a problem with the web emulator of the DMT viewer going
into loops that drain the stone and delaronde cpu's. This will have to
be watched carefully and perhaps that interface turned off until it is
fixed.
-
Szabi Marka (alarms and web interface)
(Szabi in Japan). KR has sent Szabi an e-mail concerning the alarm-silencing
version of the DMT alarms page. The one at LHO seems to work fine and presumably
should become the default alarm page. KR does not know the corresponding
URL for the LLO version of the alarm silencer.
-
Status of DMT monitors
-
Dave Chin / Tim Bodiya / Keith Riles: LockLoss
and ServoMon/SpecMon
No recent changes in LockLoss/ServoMon. KR will
exercise a new version of the interactive SpecMon program this weekend.
-
Ken Ganezer: seis_blrms
Ken had some questions about accessing the running versions of seis_blrms
at the sites. He has started looking at the code on the saiph machine at
Caltech.
-
Masahiro Ito: glitchMon
No report.
-
Sergey Klimenko: LineMonitor and WaveMon
LineMon reported on above under violin modes
entry. No recent changes to WaveMon
-
Szabi Marka: IRIG-B, TimeMon, and ShapeMon
No report
-
Tim Olson: SpectrumArchiver (report by e-mail)
Tim is waiting for a response from John on how
to use a new synchronization class he would like to use for running SpectrumArchiver
under the process manager.
-
Steve Penn / Vijay Chickarmane: BicoMon / BiLinMon
See Steve's report above under bilinear couplings..
-
Brian Stubbs: SuspensionMon
SuspensionMon is running stably at both sites.
Some earlier trigger problems at LHO seem to be fixed. KR urged running
guild to verify that the triggers do actually make it into the database.
Since alarms are not yet supported by the program, it's important that
scimons periodically check either the log file or the triggers to make
sure no mirror side motion has been kicked into unstable oscillations.
KR asked that Brian evaluate a reasonable checking scheme (once/day) and
write up a short scimon instruction sheet describing it. KR wondered if
the thresholds need to be raised on the monitors, since we generally want
triggers only when the side servos go haywire. One (presumably) should
not have received the triggers Brian verified had been generated this week.
-
Patrick Sutton / Mike Ashley: SenseMonitor / LineAmp
and
RayleighMonitor
There were a few minor fixes to SenseMonitor
since E10. The calibrated power spectra and inspiral range integrand are
now displayed properly in the DMT viewer. Patrick asked that someone define
and archive reference curves for these soon to serve as benchmarks. Patrick
has also revised RayleighMonitor to provide better defaults for launching
in the mode usually used in the control room. There is also new functionality
in that an option allows whitening the power spectra before looking for
deviations, where the whitening is based on the first N (parameter) spectra
observed during initialization. The whitening option is turned off by default.
-
Natalia Zotov: PTMon
Natalia has fixed the retriggering problem seen
before and now sees reasonable glitch rates. She can see via guild the
triggers being written to the LHO database, but cannot see them in the
LLO database (this was resolved after the telecon by Peter Shawhan who
discovered a stale config file was being used for PTMon at LLO and by Patrick
who fixed the problem).
-
John Zweizig: BitTest, DataQual,
HistCompr, PSLMon, SatMon , and SegGener
No update..
A.O.B.
-
Rauha mentioned that he is working on a scheme to notify the operators
via a pop-up window on a control room workstation when a gamma ray burst
(GRB) has been detected by the worldwide satellite network (typically 0-3
seen per day). He would like operators to hold off an hour after such an
event before deliberately taking down an interferometer. (Ray reminded
us that during the biggest GRB of the S2 run, L1 was down, and H1 was up
only because the operator had decided to wait until the turn of the hour
before leaving science mode.) Rauha has not yet worked out a mechanism
but wondered if he could one of the DMT machines. KR suggested consulting
John before carrying out any experiments on any DMT machines running online
monitors. Another method may be to go through the red and river gateway
machines at LHO and LLO, respectively. Stan should also be consulted about
implementing the scheme once it's ready..
-
Next detector characterization meeting: Wednesday
November 12 at LHO (LSC meeting)