Present:

• AEI-Potsdam: Chassande-Mottin
• Caltech: Blackburn, Lazzarini, Shawhan, Yamamoto, Zweizig
• Dublin: Ottewill
• Florida: Klimenko
• Hanford: Raab, Savage, Schofield, Sigg, Whitcomb
• Livingston: Kovalik, Marka, Saulson
• Louisiana State: Johnson
• Louisian Tech: Greenwood, Simicevic
• Michigan: Gustafson, Riles
• MIT: Shoemaker, Sylvestre
• Oregon: Mauceli
• Penn State: Finn, Mohanty, Mukherjee, Siddiqui
• Syracuse: Penn

• This meeting will focus mainly 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 and 2 DMT algorithms will be asked to report briefly and to provide realistic milestone dates for "pre-delivery" of code to John for incorporation into the DMT.
• At the end we'll have a discussion on inserting information into the meta-database and our group's participation in the first LDAS mock data challenge this summer.
• KR proposed holding DMT teleconferences the first Friday of each month at 11:00 EDT. It was pointed out that the Milwaukee group holds weekly group meetings at this time. KR will contact Bruce Allen about this and send out an announcement of the next meeting's starting time.

• John's presentation
• John has been working on client/server infrastructure for trigger generation and background monitor communication. He has also looked at the engineering run data and found a number of odd glitches.
• A new base class called TrigClient has been defined, allowing monitor processes to create trigger records in the XML format needed for meta-database ingestion (P. Shawhan script).
• An example of header and implementation code was shown. It will be straightforward for existing monitor code to be adapted to the new base class.
• In order to allow control room operators to view status information from DMT monitor processes, client/server software has been written to allow monitors to serve up objects that can be requested by a monitor display manager (not yet written). Presently these objects can include integers, floats, strings, and DMT TSeries, FSeries and FSpectra.
• The steps for converting an existing monitor to an object-serving monitor were outlined, and an overview given of the client server system. Daniel Sigg has volunteered to work on the display manager.
• It was suggested that the object-serving monitors be flexible enough to serve objects only when requested to do so and that one be able to pass parameters to the monitors. For example, if a problem is detected, one might request power spectra or other plots for a particular set of channels, but would not want those calculations done routinely.  John will look into this.
• He has run into some namespace problems in getting the new code to run under root's foreground interpreter environment, but the code does now work in background when linked to root libraries.
• John has looked at the data from the April 3/4 engineering data run as a test of trigger generation and to check the data's integrity.
• The DMT played an important role in that run, since its frame writing class was used in recording a subset of DAQ channels to disk from the fortress machine. A number of glitches occurred during the run. Fortress itself hung up for about 1.5 hours during the night before the problem was noticed (an automated diagnostic based on pinging was immediately created to detect such problems in the future!). In addition, 18 1-second frames were lost in the run for reasons that aren't yet fully understood, but appear to be related to a faulty initialization in the data server.
• Trigger generation didn't work during the run because of another problem, but John has recreated data taking conditions since then using shared memory on fortress in order to test the generation after the fact.
• Two monitors, DaqSlice and ResTrend, were tested. The first looked for 1/16-second data segment discontinuities, prompted by an observed anomaly (that is not yet understood but being investigated) in which a single channel appears to get a "stale" 1/16 second segment of data instead of the right segment. DaqSlice also looked for channel saturation and only operated when the single arm was in lock (determined from intensity of light transmitted through end mass).  The ResTrend monitor kept track of power in certain test mass resonance bands.
• The trigger generation test involved sending triggers from fortress to the sand machine for logging by the LDAS system into the LHO meta-database.
• All of the 42 AcquiredLock and LostLock triggers were successfully recorded, but 14 of the 2808 channel saturation triggers and 6 of the 4191 1/16-second discontinuity triggers were not recorded, for reasons not yet understood. John and Peter will get together to discuss improving the tracking of trigger logs to allow systematic investigation of database recording failures.
• John is preparing a memo on the problems uncovered from the engineering data run.
• He hopes to release a new version of the DMT  package in a week or two, containing the new trigger and monitor client/server code, along with line tracking code recently provided by Adrian Ottewill (see below).
• He will also soon produce guidelines on submitting monitoring code for inclusion in the DMT.

• Line noise:

 
• B. Allen, A. Ottewill (Adrian reporting):

Adrian has given a first release of line tracking code to John. For the moment, it includes clapack code in the form of a shared-object library. John plans to incorporate clapack into the DMT eventually. So this temporary solution is appropriate. The line tracking code, translated from code in the GRASP package (see Adrian's presentation at the March Livingston meeting), used DMT routines for data access and the root library for graphics. The selection of channels and frequency ranges to monitor is specified via a configuration file. Adrian wondered whether he should provide a GUI for setting such parameters in the future. He was encouraged to hold off on that for now and concentrate on correlation code (see below). The monitor display manager may provide a natural mechanism for parameter setting. Adrian plans to write a letter requesting access to the recent engineering run data.
 
• S. Klimenko (presentation)

The Florida group has installed the DMT on its Sun system and has converted existing code to track lines and their harmonics over to the DMT environment. The defined classes were outlined and usage examples shown. The code can be used both to track and remove lines. It will be provided to John for incorporation into the standard DMT release soon. The Florida group is working on a statistical evaluation of the line removal technique.
 
• Characterizing seismic noise:

 
• E. Daw (e-mail report:)
  1) A preliminary version of the seismic band limited RMS monitor has

been running at Livingston on Delaronde for about 2 weeks. The
program measures the band limited rms in two bands 0-1Hz, and 1-5Hz
every 16 seconds, and logs the data in text files on delaronde. A
new text file is started once a day. Szabi Marka has installed a
daemon that restarts my monitor if Delaronde is rebooted, and sends
me email to inform me of the trouble. A better developed bank of
filters with more bands and extending to higher frequencies will be
installed shortly.

2) I have written a simple data histogramming utility. The utility
acquires an arbitrary length sample from a single channel, and plots
the timeseries, and the projection of the timeseries into a histogram,
side by side in a root canvas.

3) A second monitor is under construction to monitor the optical lever
readouts from the south endstation at Livingston. Joe Kovalik is currently
using a PC and labview system to look for long term drifts in the mirror
alignment. I have connected the optical lever outputs to the DAQ system,
and intend to start the monitor later this week.
 

• Inter-channel correlations:

• B. Allen, A. Ottewill (Adrian reporting)

This code was promised for May 16, and Adrian still considers that realistic. The code will be in a comparable state as his line tracking code (see above), using a configuration file for setting parameters, DMT routines for accessing data and the root library for graphics.
 
• Bilinear cross-couplings:

 
S. Penn
Steve has looked over the presentation by Albert Lazzarini at the Livingston meeting and believes he can implement the techniques described there in the DMT within two months.
 
• Operational State:

 
R. Gustafson, K. Riles (KR reporting)
The code described at the Livingston meeting has been converted from the DMTBase base class to the DatEnv class and some debugging done, but the package is not ready. The pre-delivery to John should occur before the next teleconference in four weeks.

• Bandlimited RMS:

 
E. Daw (see above)
 
• Time / Frequency plots:

 
(See Mohanty & Sylvestre talks below)
 
• Non-Gaussian noise:

 
Although no one has yet signed up to take a lead role in this task, Sam Finn mentioned that a graduate student, Sharif Siddiqui, will be working with him this summer in this area and that code will be written for the DMT. He'll say more at the next teleconference.
Steve Penn mentioned that his work on bilinear cross couplings will also touch naturally on non-Gaussian noise quantification.

• Power spectral transients:

 
• S. Mohanty

This code for detecting generic transients in frequency bands will be provided in two packages. The first, which should be ready by the end of the month. supports time-frequency plots indicating via a pixel display sudden increases in power in frequency bands.Soumya will look into making this infrastructure generally available as a diagnostic. The second package, which should be ready by July 1, will contain automated pattern recognition to generate triggers based on the time / frequency plots.
 
• Servo instability:

 
• R. Gustafson, K. Riles (KR reporting)

Nothing new to report, other than a delayed milestone. This code should be ready roughly a month after the operational state condition (see above), i.e., early July.
• Event catalog:

 
• J. Sylvestre, R. Weiss (Julien reporting)

Two techniques are being used to identify and classify transients. One is power monitoring in time / frequency space, which may provide another tool for generic time / frequency plots. The second technique is based on matched filters, of which a handful have been implemented. The DMT code is now running in real-time at Hanford and has also been tested on the engineering run data. Julien is now writing a GUI for controls and display of results.
The package has as inputs a channel list, filters list (time domain), channel/filter patterns with automated decimation selection,  channel list for power monitoring, and parameters for defining time series length and transient thresholds. Its output is a list of candidate transient events and can include associated frames of data.
He is working on automatic classification of transients detected in the power monitor and expects to have code to give to John (including the GUI) by mid-June. A later release would allow for pre-whitening and other user-specified filters.
Soumya suggested Julien look into the windowed Fourier transform technique he is using to identify power spectral transients and that he and Julien exchange notes.
Fred Raab mentioned that an REU summer student, James Keef, will be arriving at LHO on June 5 to spend a couple of months running DMT code to shake it down. He can try evaluating / comparing the different time / frequency plotting routines described above.

• Flickering optical modes:

 
No volunteer!!!

• Transient detection using adaptive denoising methods:

 
• E. Chassande-Mottin

This task was originally titled "impulse recognition" (see below), but Eric emphasizes that although his adaptive denoising technique will detect a variety of transients, including impulses, it will not (at least initially) "recognize" them, i.e., classify them. He still plans to have the detection code ready by October 1.

• Impulse recognition:

 
• Although no one has volunteered for this task, it has much in common with the matched filtering technique used in the event catalog (see Sylvestre talk above). Determining the appropriate filters to use to recognize impulses needs work. A volunteer is welcome!

 
• Magnetic field transients:

 
• E. Mauceli

Much of the code is written. Evan is waiting for new magnetometer hardware with sensitivity good enough to be useful. Such a magnetometer was used during the engineering run, and Evan was encouraged to use that data for determining first-guess thresholds to use. He will be at LHO next week and will pursue this further. He hopes to give code to John a week or so later.

• KR encouraged everyone to look over the meta-database tables documentation provided by Peter Shawhan. Although much of detector characterization involves diagnostics and alarms to be used immediately on site, there is also much information that needs to be transmitted downstream through the meta-database or through frames for use in astrophysical analysis.
• The detector characterization group has been encouraged to participate in the first LDAS mock data challenge (MDC) to be held in August 2000. That MDC (led by Sam Finn) will test the data conditioning API. (Sam had to sign off before this item on the agenda came up and did not participate in this discussion, but in a later e-mail emphasized the need for careful coordination.)
• KR proposed the following areas as natural to test in this MDC:
• Triggers for vetoes (seismic transients would be a good candidate for testing)
• Line tracking & removal
• Regression coefficients & corrections
The first should be straightforward, but the other two require some thought / discussion,
since whatever data conditioning API code uses dmt-derived constants must use those constants exactly as intended. In addition, Kent Blackburn has raised the issue that deriving useful regression coefficients may require long integration times that may not be well suited to the DMT.
• The main purpose of bringing detector characterization into this MDC is to test the infrastructure of the DMT/LDAS chain and to confront nitty-gritty technical issues, such as those uncovered in the engineering run (see Zweizig talk above).
• A teleconference in late May will be organized to discuss what is needed, and volunteers to participate were solicited. Persons who volunteered at this meeting or who have already expressed an interest include: K. Blackburn , S. Finn, A. Lazzarini, E. Mauceli, A. Ottewill, K. Riles, P. Shawhan, D. Sigg, R. Weiss, and J. Zweizig.