Minutes of Detector Characterization Teleconference
(January 11, 2002)
Present:
Caltech: Drever, Shawhan
Carleton: Christensen
Dublin: Ottewill
Florida: Klimenko
LHO: Chin, Gustafson,
Ito, Landry, Raab, Schofield, Sigg
LLO: Coles, Daw, Katsavounidis,
Klimenko, Lazzarini, Marka, Saulson, Sutton, Zweizig
Michigan: Riles
Syracuse: Penn
Introduction & Announcements - KR
-
Today's meeting is devoted to software and scimon shift issues. Feedback
is solicited from scimon shift takers on how to improve the scimon experience,
including how to improve existing DMT monitors.
-
Also, interferometer commissioners have requested several interactive tools
for which volunteers are welcome.
-
Two mailing lists have recently been set up, ligo-dc@tenaya.physics.lsa.umich.edu
for general announcements / discussion related to detector characterization,
and dmt-discuss@tenaya.physics.lsa.umich.edu for DMT monitor authors. The
archives can be found at http://tenaya.physics.lsa.umich.edu/mailman/listinfo/ligo-dc
and http://tenaya.physics.lsa.umich.edu/mailman/listinfo/dmt-discuss
.
E7 Feedback - All
(The discussion was free flowing, bouncing back & forth
among several issues. Below is a summary organized by topic.)
-
Shift scheduling - There was a strong consensus that 12-hour night
shifts (8-8) were too long. Most speakers were in favor of simply returning
to three 8-hour shifts per day, but Mark Coles also reminded everyone of
the LLO operator schedule of three 10-hour shifts with six hours of overlap.
John Zweizig suggested that a 6+8+10 schedule be considered, with the 6-hour
shift in the night hours (0-6, 6-14, 14-24). No one argued for continuing
12-hour shifts any time of the day, partly to avoid drooping alertness
and partly to allow off-duty work on investigations and interaction of
travellers with site scientists.
-
Shift duties - Several persons suggested more automation of routine
items now on the checklist, leaving more time for investigations &
critical thought. Szabi Marka reminded everyone that the shift checklist
is open to modification by anyone on shift.
-
Detector / DAQ / LDAS monitoring
-
There was a general feeling that several DMT monitors were nearly "invisible"
in that their output (triggers, trends, log files, etc) were unavailable
or confusing to people on shift. Despite having exploited much of the I/O
infrastructure available to DMT monitors, particularly for recording info
for posterity, some authors clearly haven't devoted as much effort to giving
an informative summary to scientists in the control room. Each monitor
should provide a user-friendly html summary page that is self-explanatory.
-
Particular requests included a measure of current transient rates compared
to historical norms (several monitors would benefit from this). Also, a
global summary page of DMT monitor status and of interferometer performance,
as measured by standard-candle inspiral distance, would be desirable. Szabi
is working on dedicated DMT monitors to produce a status summary page and
to handle (audible) alarms. KR will talk with Sergey Klimenko and Adrian
Ottewill about useful summary pages for line monitors.
-
For part of E7, some monitors (that sampled the data stream less frequently
than 1/second) were hampered by a bug (now fixed) in the DMT viewer interface
code that prevented their plots from being displayed reliably (or at all).
-
KR urged that the Laboratory devote the necessary resources to make
it possible to see DMT-output trends with the playback mode of the data
viewer. That capability would automatically make the output of some
monitors (potentially) highly visible and useful in the control room. The
required change is non-trivial, however, involving modification of the
frame builder infrastructure.
-
A number of problems in the online LDAS system went undetected for some
periods during E7 because it wasn't clear to scimon shifters how to confirm
correct functioning, and because some LDAS self-monitoring infrastructure
doesn't exist yet. Albert Lazzarini confirmed that discussions are underway
on how to improve the system, including a web interface to be visible in
the control room.
-
Dave Chin requested that the DAQ fault channels now running at LHO be installed
at LLO so that frozen channels, e.g., those indicating lock state, can
be detected. (after the meeting Stan Whitcomb confirmed that the fault
channels will be available at LLO next month).
-
Dave also raised the issue of recording in the database when a DMT monitor
has aborted (by crash or deliberately). John agreed to think about how
best to keep track of aborted processes. The function might be carried
out, for example, by the trigger manager process. Another possibility is
increasing the number of conditions that define data segments in the database
and adding new segment types.
Requests for new detector characterization tools
KR relayed requests from IFO commissioners for several highly
interactive, GUI-controlled tools that would greatly aid in day-to-day
commissioning (and which would be useful on shift during engineering runs):
-
Package for automatic, periodic recording of reference spectra for
large numbers of channels, with easy means
for retrieving and comparing with real-time spectra. One possible implementation:
-
DMT monitor running in background that once per hour computes a PSD estimate
for hundreds of channels and stores them on a local site disk indexed by
day/hour/channel. Computation of PSD would follow usual convention of the
DTT. Spectrum for each channel would be stored as DTT-readable xml file.
-
Modified version of DTT with easy GUI selection of desired reference
spectrum to import, according to the day/hour/channel indexing.
John pointed out that the PSLMon monitor does something similar
to the first task, including the writing of spectra in DTT-readable XML
files. A relatively minor upgrade could store the files in a day/time/channel
directory structure. The main job would be the modification of DTT. Daniel
Sigg pointed out that the choice of optimal spectral parameters would vary
among channels and require some thought; KR suggested getting the infrastructure
in place with reasonable defaults before worrying too much about optimization.
This task still needs a volunteer!
-
Interactive GUI (launched from GDS diagnostics menu) allowing tracking
of band-limited RMS a la Ed Daw's blrms program with graphic display either
via the GUI itself or via the DMT viewer. Should be easy to launch a program,
select a channel, specify a band of interest and display a time series
of the power in that band. Multiple traces would be desirable too. Since
much of the underlying machinery already exists in the DTT, a modified
version of it may make the most sense.
Fred Raab suggested that a waterfall option would be desirable,
in which multiple spectra for the same channel taken at regular intervals
of time are displayed together with fixed vertical (and horizontal too?)
offsets to allow the eye to pick out trends and anomalies easily. Ed Daw
had exited the call before this point on the agenda, but John believed
Ed had in mind developing a tool along the lines described above. In a
discussion following the meeting, Ed confirmed that he does plan to provide
a tool similar to that described above. Thanks, Ed!
-
Interactive GUI (launched from GDS diagnostics menu) allowing a pixel-display
of band-limited RMS over many bands of a single channel to scroll across
the screen (time/frequency plotting). Again, channel selection should be
easy via GUI menus.What would distinguish this tool from what exists (e.g.,
tid) is a highly interactive interface with no command line or configuration
file.
Given his experience with the real-time bicoherence monitor display
and the recent availability of GDS library routines for GUI channel selection,
Steve Penn volunteered to take on this task. Thanks,
Steve!
Status reports on DMT infrastructure:
-
General - J. Zweizig
-
Things are running reasonably well. One unexpected glitch during E7 was
the inability to run more than 16 consumer slots simultaneously. Before
the frame table-of-contents reader was added to the DMT I/O interface,
more than 16 was precluded, anyway. With a recent fix, John can now run
32 monitors, but any monitors now in development need to be recompiled
in order to work properly.
-
As mentioned above, a bug was fixed that interfered with the passing of
data to the DMT viewer from processes with time strides longer than 1 second.
-
John has set up a somewhat ad hoc scheme that allows selected dmt monitor
objects from different sites to be displayed in the same dmt viewer. The
motivation was to display LHO and LLO lock histories from the LockLoss
monitor together, but the same method could be applied on a case-by-case
basis to other dmt objects. If enough demand arises, John will generalize
the new code to handle arbitrary objects.
-
The present average trigger rate from the glitchmon monitor is running
about 2 Hz, which is larger than the nominally allowed total trigger rate
from all DMT monitors. John has asked Masahiro Ito to retune glitchmon;
in particular, triggers tend to come in bursts over many channels; a simple
scheme to combine coincident triggers may solve the problem.
Status reports on DMT monitors:
-
Line noise (multi-taper method) & inter-channel correlation monitors
- B. Allen, A. Ottewill (Adrian reportingl)
John has fixed a recent problem with a root version incompatibility
for the version of the CorrMon monitor used to generate postscript plots.
Adrian wondered whether background monitors could generate the plots without
the root interface. Szabi said he is working on a library of routines for
producing plots in various formats, including postscript. Adrian volunteered
to help.
-
Line noise (quasi-monochromatic method) monitor - E. Daw, S. Klimenko
(Sergey reporting)
Two line monitors were run in E7 for each interferometer, one looking
at darm_ctrl, the other at carm_ctrl. Each monitor looked at 60 Hz &
harmonics and at known violin modes. A new feature is a histogram for the
dmt viewer that displays the number of significant lines found per time
interval vs frequency. KR suggested providing a summary html file with
one row per line with current best estimate of amplitude and phase, modelled
on Daniel's MultiVolt summary page.
-
Band-limited / seismic noise monitors - E. Daw (no report)
-
Bilinear cross-couplings & broad-band non-Gaussianity monitor - S.
Penn
A new version of the bicoherence monitor is available to produce eps
files, which will make the output more useful for the E6/E7 correlations
investigation.
-
Lock transition and servo instability monitors - D. Chin, K. Riles
(Dave reporting)
Dave has made many improvements to the LockLoss monitor before and
during E7, including a revamped html summary page with multiple updated
livetime estimates and multiple lock history plots for the DMT viewer.
The ServoMon monitor has received less attention, but improvements include
monitoring of more violin modes and rescaling of config file threshold
units for band-limited rms to agree with DTT conventions. With John's fix
of the dmt viewer object serving bug (see above), ServoMon plots are now
reliably visible.
-
Rayleigh behavior monitor (non-Gaussianity vs frequency) - S. Finn,
G. Gonzalez, P. Sutton (no report)
-
Power spectral transient monitor - S. Mohanty (no report)
-
Transient identification & event catalog monitors - J. Sylvestre,
R. Weiss (no report)
-
Glitch monitor - M. Ito
Masahiro is trying to retune glitchmon to address the trigger rate
problem John described above. One source of bursts of triggers are the
turning on of whitening/dewhitening filters about five seconds after lock
is acquired. John suggested that Masahiro exploit the latency option of
operational state conditions that define lock so that he doesn't start
looking for glitches until at least 10 seconds into a locked stretch. KR
and Dave will look at defining a standard delayed lock acquisition condition
for general use. This issue provoked some discussion on automated flagging
of control parameter changes when nominally in detection mode. It
was suggested that a button be added to the main lsc medm panel that operators
should turn on when in physics mode and turn off during any tweaking. Peter
Shawhan volunteered to modify conlog to set an audible alarm if any changes
are made to any control parameter when in physics mode (but asked that
someone at the sites set up the panel button and associated variable).
Daniel suggested that "clean" lock segments stored in the database should
require that no control parameter changes have occurred during that segment.
-
Magnetic field and seismic transient monitors - R. Frey, R. Rahkola
(no report)
A.O.B.
-
Next regular teleconference: Friday February 1 at 11:00 a.m. EST
(8:00 PST).