Minutes of Detector Characterization Teleconference (February 14, 2000)
Present:
-
Dublin: Ottewill
-
Florida: Klimenko, Whiting
-
Hanford: Mauceli, Raab, Savage, Schofield, Sigg, Zweizig
-
Livingston: Kovalik, Marka, Saulson
-
Louisiana State: Johnson
-
Louisian Tech: Greenwood, Simicevic
-
Michigan: Riles
-
MIT: Daw, Fritschel, Zucker
-
Oregon: Brau, Rahkola, Strom
-
Penn State: Finn, Mohanty
Introduction & Overview (KR):
-
This meeting is meant to focus on software development for the Data Monitor
Tool (DMT).
-
After John Zweizig's DMT status report, persons who have signed up to write
priority-1 DMT algorithms will be asked to report briefly and to provide
realistic milestone dates for installation of code in the DMT at Hanford.
These milestones will be reported to the collaboration at the March LSC
meeting at LLO.
-
In addition, persons working on priority 2 and 3 DMT tasks are welcome
to report afterward. Sergei Klimenko will speak on wavelet analysis and
has provided a pdf file of his transparencies.
Data Monitor Tool Status (John Zweizig):
-
John's outline
-
Version 1.1 was released in early January with various
updates promised in the December teleconference.
-
The new version is compatible with the old egcs compiler;
it does not require the gcc version which has given a number of
persons trouble on non-Solaris platforms.
-
A slightly revised version 1.1a with various
bug fixes was released recently.
-
A new Web
page has been added to the DMT Web site that gives more explicit detail
on what is required to install the DMT at other sites.
-
Ed Daw has written a step-by-step
guide for installation (from scratch) on Solaris.
-
Ed and Adrian Ottewill have been providing feedback
on compiler incompatibilities.
-
The DMT/root combination seems presently incompatible
with Digital's unix operating system.
-
A recent change to the build files for the FrameCPP
library required a hacked up kludge in the DMT build files. A better fix
is imminent.
-
The data errors in the Frame broadcaster at LHO (reported
at December teleconference) have been fixed, and no errors (i.e., lost
frames) have been seen since January 14.
-
The Fortress multi-CPU workstation is now set up
to write designer data sets.
-
The machine has two 70 GB RAID disk arrays and an
attached AIT-2 tape drive robot. For now mounting is done manually, since
the robot software is not yet working on Solaris 2.7.
-
Fortress is now used by LHO scientists to look at
data that would otherwise be lost within a day. The designer data sets
can be viewed by the CDS xmgr-based data viewerer and the GDS root-based
viewer.
-
Some minor configuration changes on the LHO DMT machines
have made them more reliable.
-
A data conditioning API is under development by S.
Finn, J. Romano and P. Charlton to provide signal processing for LDAS.
A wrapper for incorporation into the DMT is under development.
-
Sam Finn requested direct support for the valarray
data type, including a constructor and assignment operator. John said that
can be provided.
-
Daniel Sigg wondered how fast the valarray operations
would be. Sam stated that they are indeed fast, and the code is being written
with that in mind.
-
John and Daniel are working out a scheme for communication
with background DMT processes, based on sockets. The scheme would allow
individual DMT processes to serve external information requests, using
standard library calls.
-
The graphical display will be based on the GDS root-based
display manager written by Daniel for viewing results of online diagnostics.
Display requests can be provided via parameter files or scripts.
-
Much of the communication software is written, but
is not yet ready for release.
-
Other recent DMT changes include min/max routines,
standardized test macros and incorporation of Alex Ivanov's DAQFrame code
which allows rapid reading of partial frames. Reading partial frames allows
more DMT monitor processes to be running in parallel on the same machine.
Status reports on performance characterization priority 1 tasks:
-
Line noise:
-
B. Allen, A. Ottewill (Adrian reporting):
Line removal code is being ported from GRASP to the DMT. Adrian had
trouble getting the DMT to run under Digital unix and has moved to a linux
platform, where he now has it running. He has written a DMT version of
a seismic monitor and now understands how to port the line noise and inter-channel
correlation code software (another task below) to the DMT. He expects to
have both algorithms running by the March LSC meeting and will try installing
them on the LLO machines before the meeting. John confirmed that he will
be at LLO before the meeting to get the DMT infrastructure running there.
Adrian mentioned that the correlation code uses routines from the LAPACK
library and asked whether linking to that library would be permitted within
the DMT. John said he had no objection and didn't forsee a problem.
-
S. Finn, S. Mukherjee (Sam reporting):
They are now working on the data conditioning API discussed above.
Once that is complete and the DMT/API interface is in place, they should
be able to install a Kalman filter for monitoring violin modes relatively
quickly. They expect to have this ready not long after the LSC meeting.
The group has also signed up for inter-channel correlation software (another
task below), but can no longer identify a person to work on it. No milestone
date can be estimated yet.
-
R. Gustafson (KR reporting)
Dick has just taken up full-time residence at LHO and will use any
line identification software available to help track down and fix 60 Hz/harmonic
contamination. This is part of a broader role of serving as a site-based
liaison to DMT software writers, giving feedback on algorithm effectiveness
and ease of use.
-
W. Johnson, R. Svoboda (Warren reporting)
Bob Svoboda has joined the KamLand experiment (neutrino oscillation
search) and will be unavailable to help convert Warren's Matlab-based analysis
to the DMT. Warren asked that his and Bob's names be withdrawn from this
task.
-
A. Sintes (not present)
John stated that he believes Alicia succeeded in getting a DMT-compatible
compiler installed at AEI recently, but then ran into a problem with the
FrameCPP interface (see above).
-
S. Scott, B. Whiting (Bernard reporting)
The person (Phil Charlton) working on this software will be leaving
ANU soon to join the Caltech group and has several tasks to complete before
leaving. Currently he is working on tests of different line-noise removal
algorithms, including a Kalman filter. Bernard requested that Sam provide
some simulated data for additional testing, to which Sam agreed. A best
estimate for a milestone date for a DMT-installed algorithm is early April.
-
Characterizing seismic noise:
-
G. Gonzalez (not present)
-
D. Greenwood
Dick reported that he and Neven have been analyzing data taken with
high-sensitivity seismometers last summer and fall. They have focussed
on the micro-seismic peak and have not yet started looking at gravitational
gradients. They have also sent their data to Gaby. They do not forsee writing
DMT monitors in the near future. KR urged them to communicate with Mauceli
(Oregon) who has written a seismic monitor for the DMT. Dick mentioned
that they are currently using a sophisticated package written at the Livermore
Lab that may not be suitable for adapting to an online monitor.
-
W. Johnson
Warren provided a web page
describing work to date on seismic data analysis, using a Matlab-based
package. The page outlines the time-frequency analysis used, provides sample
plots and provides links to download the matlab code. In brief, the analysis
looks at the time dependence of rms motion in frequency bands (octaves).
Data from LLO (taken by Shourov Chatterji, Tom Evans, and Joe Giaime)
allow identification of transients caused by persons walking in the vicinity
of the Y end station. As mentioned above, Warren does not have anyone on
hand to convert the Matlab code into a DMT monitor, but is quite
willing to work with others.
-
Comments by KR: Several groups have real expertise in seismic analysis
but little experience in the software coding needed for the DMT. This seems
to be an area where direct collaboration among the groups is called for.
Although not explicitly signed up for this task, the Oregon group has taken
the initiative in writing a first-order DMT seismic monitor (to be reported
on at the March LSC meeting). The other groups above may find it most profitable
to work with Oregon to ensure the best possible DMT seismic monitor emerges.
Note that Adrian Ottewill of Dublin has also written a seismic monitor
(see above).
-
Stack vibrations:
-
J. Giaime (Warren reporting)
A former MIT student, Andrea Stulte, carried out transfer function
measurements at LLO and put together a draft report before leaving MIT.
Joe is working to complete that report by the March LSC meeting. Warren
did not know Joe's plans in regard to a DMT algorithm. It was mentioned
that a report from Mark Barton describes transfer functions measured at
LHO. (Note from KR: in a later conversation with Joe, it became clear
that this monitoring task needs better definition before milestone dates
can be estimated. An important issue is the number of accelerometers available
for direct monitoring.)
-
S. Penn (KR reporting)
(Steve was en route to LHO during the teleconference.) Recently he
has mainly worked on providing calibrated transformation matrices to convert
raw signals from test mass shadow sensors and from the optical lever sensors
into orthogonal pitch, yaw and position determinations. He will be carrying
out more tests of his code during the LHO visit. His work on the DMT was
hampered until recently by the DMT's need for the gcc compiler. Although
he installed the DMT code on a linux machine, he then discovered that root
does not have a gcc-compatible version for his powerpc linux machine. Now
that the DMT can once again use the egcs compiler (see above), he will
try again at the first opportunity.
-
Inter-channel correlations:
-
B. Allen, A. Ottewill (see above)
-
S. Finn (see above)
-
Operational state:
-
R. Gustafson, K. Riles (KR reporting)
KR is working on DMT code to allow specifying logical conditions on
one or more channels as a prelude to analysis. Specifiable conditions include
lock configuration, ambient rms in a given channel. laser intensity, etc..
The code allows user-friendly Boolean sets of conditions to be specified
and for those conditions to be built upon. Code development is being carried
out at Michigan with testing done remotely at LHO, since the dmt/root package
is not yet running on Michigan HP workstations. In the longer term, a linux
pc will be installed at Michigan for code development. Logging in remotely
at LHO works, but graphics display is quite painful. The first version
of the operational state condition code should be ready by mid-March.
Status reports on transient analysis priority 1 tasks:
-
Frequency band transients:
-
S. Mohanty
Soumya is now porting his Matlab code for generic band-limited transient
detection to the DMT and expects to have it ready by the end of March.
-
Servo instability:
-
K. Riles
This code looks for characteristic increases in rms motion in specified
frequency bands that indicate potential onset of servo instability. To
be useful, the code needs a well defined operational state to be defined
(see above). An initial version of the code, using very simple band-limited
rms measures, is expected to be ready by the end of March. As other
generic techniques for detecting band-limited transients become available
(e.g., Mohanty algorithm above), they may be adopted instead.
-
Event catalog:
-
R. Adhikari, R. Weiss (Fred reporting)
Fred believes that Rana will be starting work on this very soon, now
that he has passed his MIT qualifying exam (hurray!). Fred promised to
monitor this work closely.
Status reports on lower priority DMT tasks:
-
Wavelet analysis of transients:
-
S. Klimenko
Sergey provided a pdf file
describing a wavelet analysis approach to detecting transients. The analysis
package (WAT = Wavelet Analysis Tool) being developed at Florida is meant
to help in three areas: data compression/reduction, transient signal characterization,
and identification of unmodelled GW sources. Other persons involved include
G. Mitselmakher, A. Sazonov, and B. Whiting. Wavelets provide an alternative
to Fourier transform analysis and are widely used in pattern recognition.
Well tuned wavelet forms allow small number of parameters to be used, in
principle. Examples are shown of wavelet representation of signal and noise.
The wavelet version of a low-pass filter allows compact representation
of a time series. Presently WAT exists as a C++ library with matlab, frame
and root interfaces. Eventually it will conform to the LAL code specification.
It has not yet been decided whether to port into the DMT.
-
Frame writing from Matlab:
-
E. Daw, S. Finn
As part of the data simulation effort, Ed has
ported the DMT frame-writing code into a Matlab-compatible routine for
easy direct generation of simulated frame data from Matlab.
A.O.B.
-
Prospective speakers for the detector characterization
parallel sessions at the March LSC meeting are urged to contact KR or Daniel.
-
An important topic of discussion at the March meeting
is the mechanism for distribution of special data sets to LSC institutes.
In particular, this group would like a straightforward way to obtain customized
/ designer data sets for studying particular characterization issues. A
collective proposal from the working group to the collaboration may be
the appropriate mechanism.
-
Be sure to periodically check the working
group bulletin board for important updates and miscellaneous links
to useful documentation