Minutes of Detector Characterization Teleconference
(June 14, 2002)
Present:
Caltech: Aso, DeSalvo,
Brown, Drever, Kells, Lazzarini, Marka, Sanders, Shawhan, Zweizig
Carleton: Christensen
Florida: Klimenko
LHO: Ito, Landry, Savage
LLO: Zotov
Michigan: Riles
MIT: Ballmer, Katsavounidis,
Saulson
Oregon: Leonor
Penn. State: Schlaufman,
Sutton
Syracuse: Penn
S1 Preparation:
-
Meetings - Another teleconference will be held in one week (Friday
June 21 @ 11:00 a.m. EDT) to review further progress on S1 preparations,
including DMT software improvements. Additional meetings will be scheduled
at that time, when more site scientists are on the line.
-
DMT code deadline - John Zweizig wants all new DMT monitor code
to be run in S1 provided to him no later than Friday June 21.
-
Scimon shifts - It was decided not to synchronize scimon shifts
across the time zones. They will begin at 8 a.m., 4 p.m. and midnight local
time at both sites. KR had a good response to an e-mail requesting more
scimon volunteers, but still needs more experts at both sites, especially
at LLO. Here is the current shift schedule.
-
Scientific investigations - KR proposed a strawman
list of S1 investigations, many of which are very similar to what have
been done in previous engineering runs, but with a few new ones suggested
by Peter Saulson. In particular, Peter suggested that the charge of some
of the groups should include explicit comparison and evaluation of DMT
tools that perform similar functions to understand better their strengths
(and weaknesses). Volunteers are solicited to work on these investigations.
Albert mentioned a joint LIGO/VIRGO investigation of photodiode sensitivity
to the environment and promised to contact Robert Schofield who will lead
the investigation on correlated inter-site environmental transients.
What has been learned about DMT vetoes from E7 upper limits analysis?
-
Bursts - E. Katsavounidis
-
The Burst group has a well defined analysis pipeline that includes triggers
from both DSO search engines and DMT monitors. The primary DSO engine has
so far been Julien Sylvestre's TFClusters program.
-
Trigger rates have been found to be time dependent with up to 50% variation.
-
Typically the first 60-70 seconds of a lock stretch are vetoed as noise
level decays after lock acquisition.
-
The efficiency of DMT triggers to veto DSO candidates has been studied
together with loss of livetime in applying the vetoes. It was decided in
DMT trigger comparisons to fix livetime at 95% and evaluate efficiency,
which were typically 30-40% for the Hanford 2K, even in combination, and
typically 60-70% for the LIvingston 4K.
-
The DMT monitors / channels found to be most effective to date include
AbsGlitch and ZGlitch, with ZGlitch doing somewhat better on the whole.
Running inspiral templates on auxilary channels has also been tried, along
with a dedicated waveform analysis of a particular PSL system glitch. The
inspiral templates are found to be fairly orthogonal to the DMT glitch
monitors.
-
The different interferometers had different glitch pathologies. The Livingston
4K AS_Q channel, for example, had many glitches correlated with glitches
on the light level of the PSL reference cavity. The Hanford 4K and 2K had
a more diverse set of problems.
-
The combining of glitch triggers from different channels did not seem to
pay off, suggesting that a glitch felt in one auxiliary channel was very
likely to be felt in others too.
-
KR urged Erik and collaborators to ensure that the DMT authors of the monitors
found to be useful in E7 be aware of thresholds and urged to update them
for S1 (absolute thresholds will presumably have to be adjusted, for example).
He also mentioned that Duncan will be giving a report next week on a control
room web tool for monitoring DSO trigger rates with a few minutes latency.
-
Inspiral - N. Christensen
-
The inspiral group is working closely with Erik and Peter on DMT vetoes.
-
They have found AbsGlitch especially useful, but have also focussed strongly
on inspiral template analysis of auxiliary channels like POB_Q, MICH-CTRL
at LHO and REFL_Q and MICH-CTRL at LLO. No attempt has been made so far
to account for transfer functions between those channels and AS_Q.
-
The raw trigger rate in LLO was found to be about 200 times higher in the
LLO 4K than in the LHO 2K.
-
As reported by Erik, one can DMT-veto 30-40% of DSO candidates while maintaining
high livetime.
E8 DMT / RDS recap:
-
Overview - KR
-
KR thanked the LHO staff for supporting the E8 run and the DMT monitors
/ testers who visited LHO for the run. There had been tremendous progress
in the utility of many individual monitors since E7 and in the DMT infrastructure
(web interface, alarms, trend viewing in the data viewer, many new dedicated
blades).
-
But much remains to be done before S1. Minutes
from Monday's software status review meeting in the control room indicated
considerable work in progress. Authors need to finish up all that they
can and get revised code to John no later than next Friday.
-
DMT infrastructure - J. Zweizig
-
A pre-release of DMT version 2.0.2 was used in E8. A final tagged version
will be released this afternoon. Version 2.1 will be used in S1 and will
include the monitors submitted to John by next Friday.
-
There is much work to do at LLO to bring it to the same state as LHO in
E8, including the commissioning of the new DMT blades.
-
There is a lingering message-passing problem among the blades that John
doesn't yet understand and which caused some start-up troubles for some
monitors in E8.
-
It's very likely that at least one new blade will be moved into the control
room at both sites and dedicated to interactive DMT monitors, primarily
RayleighMonitor and BicoMon. Steve Penn reported an odd observation, namely
that BicoMon seemed to run faster when displaying to (some) remote machines
than when displaying to a hardwired monitor. No one understood why that
would be.
-
DMT summary page - S. Marka
-
Szabi's new summary page is in progress. He is waiting for various graphical
displays, alarm descriptions and abbreviated html pages from various authors.
-
After some discussion, it was decided that audible alarms would probably
be premature for S1, the worry being that operators/scimons would simply
disable persistent false alarms. Authors are encouraged to set alarm severity
levels with discretion.
-
Reduced data set generation in E8 - I. Leonor
-
Reduced-data frames were generated for 45 hours of E8 using 152 channels
at Hanford, with mostly two times decimation for 16 kHz channels. The total
data set for S1 using the same channel selection / decimation would be
about 1.5 TB. Isabel urged everyone to look over the posted
list of channels. KR has asked the upper limits group chairs to review
the list.
-
A new version of LDAS with faster decimation has just been released. Isabel
will use it to regenerate the E8 RDS as a verification test. There was
some trouble in E8 when running RDS generation simultaneously with Greg
Mendel's power spectrum generator (for the CW source search) in that RDS
generation was 4 times slower than real time. The latest LDAS version should
help, but probably won't be enough, which means S1 RDS generation will
lag behind real time.
Updates from DMT monitor authors (progress since Monday meeting and
prospects for S1):
-
Dave Chin: LockLoss and ServoMon
The automated generation of Szabi's requested 2-week summary of livetime
is done, and Mode Cleaner lock OSC conditions have been defined for LHO
IFO's. The bit needed for LLO does not yet exist, but Mike Landry will
make it available soon. The DMT viewer history storage for both LockLoss
and ServoMon is nearly ready. So are various new lock OSC conditions based
on the IFO state vectors. The present 1-second update of DMT viewer lock
history has an 8ksec time span. That will soon be changed to 2 hours to
conform to the standard defined at Monday's meeting. Other tasks to be
ready for Friday's deadline: displaying ServoMon trigger rates in
the DMT viewer and trending them; setting alarms in LockLoss, and standardizing
trend channel names. It's very unlikely that the elog-requested feature
of computing integrated livetime from an arbitrary, user-specified date/time
will be ready by S1. Szabi may be able to provide this function for up
to two weeks in the past on the global summary page, using a cgi script
and Dave's 2-week livetime data points. It's also unlikely that a requested
1-year history capability will be ready either.
.
-
Ed Daw: blrms and PeakMon (no update)
-
Masahiro Ito: glitchMon
The latest version of the glitchMon code is now in the CVS archive
and seems to run fine. An instability problem was tracked down and fixed.
Masahiro is now working on tuning the config file.
-
Sergey Klimenko: LineMonitor and WaveMon
All of the standard DMT features have been installed in LineMonitor
except the setting of alarms. Sergey was concerned about setting audible
alarms this early, but is agreeable to quite alarms displayed on Szabi's
alarm summary page. He will try to have alarms ready by Friday on violin
modes. WaveMon (new correlated glitch finder) is not as mature as LineMonitor
but is rapidly getting there. Sergey's test version now writes trends,
a summary html file and trigger files. It will soon write triggers to the
database. He should have it all ready by Friday.
-
Szabi Marka: IRIG-B and TimeMon
Both monitors now support trend writing, DMT viewer plots and alarms.
Szabi is working on documentation.
-
Adrian Ottewill: MTLineMon and CorrMon (update by e-mail)
Test versions of the monitors write trends and summary html files.
It's not clear what else will be ready by Friday.
-
Steve Penn: BicoMon
A rare bug in root when running with threaded applications was causing
crashes roughly every 400 frames. This was fixed by simplifying the communication
between the calculating part of the program and the graphics display.
The absence of an expected upconversion signal reported Monday by Robert
Schofield appears to be an artifact of insufficient frequency resolution.
Steve is changing the range of parameters allowed to permit mHz resolution
if desired and is changing the way the user defines the parameters to be
more intuitive. The documentation is being upgraded to include some sample
Matlab-generated artifacts as demos and will also include artifacts found
in LIGO data. Graphics display speed is still a problem, where X appears
to be the bottleneck when trying to keep up in realtime with a 16 kHz channel.
Patrick Sutton mentioned that he has a very fast graphics painter, but
that it's unstable at the moment.
-
Daniel Sigg: MultiVolt (no update)
-
Patrick Sutton: SenseMonitor and RayleighMonitor
RayleighMonitor is in good shape. Patrick is working on various schemes
for providing fast display and graphics output files. SenseMonitor still
needs some work. Sergey's line tracking code will be used to generate corrections
to the transfer function used in computing NSB range. A trended output
will soon be available. Patrick will talk with Mike and Szabi to
see what can be done to automate the use of the latest calibration file.
-
Natalia Zotov: PTMon
There remains a problem with output which may not be solved by next
Friday. Natalia thinks it unlikely the monitor will be ready for collaboration
use by S1, but plans to run private debugging versions on the sand/stone/delaronde
development machines.
-
John Zweizig: BitTest, HistCompr, PSLMon, and SegGener
The documentation for all four monitors has been updated. SegGener
used only physical locks in E8, but will use the new "run mode" OSC condition
for S1 which uses the operator/conlog bits in the state vector. PSLMon
now supports windowing of band limited RMS. ZGlitch (part of PSLMon)
now writes trends of trigger rates.
A.O.B
-
Next teleconference: Friday June 21 at 11:00 a.m. EDT (8:00 PDT).
This telecon will be devoted to S1 preparations.