Minutes of E3 Teleconference
(February 19, 2001 at 1:00 p.m. EST, 10:00 a.m. PST)
Present:
ANU: A. Searle
CIT: R. Drever, P. Lindquist, B. Mours, P. Shawhan,
A. Vicere, S. Whitcomb,
H.
Yamamoto, J. Zweizig
Fermilab: T. Nash
Florida: S. Klimenko
LHO: W. Butler, G. Gonzalez, M. Landry, F. Raab,
T. Summerscales
LLO: M. Coles, E. Daw, W. Johnson, J. Kern, J.
Kovalik, R. Riesen, A. Rizzi, G. Stapfer
Louisiana Tech: D. Greenwood
Michigan: K. Riles
MIT: R. Adhikari, P. Fritschel, M. Zucker
Nashville: S. Marka
Oregon: R. Schofield, D. Strom
PSU: G. Gonzalez
Syracuse: S. Penn
Wisconsin: D. Brown
Introduction (KR)
-
The E3 run will begin at noon CST Friday March 9 (10 a.m. at Hanford) and
end at noon CST Monday March 12.
-
The Hanford 2-km interferometer will be in a full, recycled configuration.
The Livingston 4-km IFO configuration is not yet certain, but may be in
recombination (both arms, but no recycling - see below).
-
Running will be in 24-hour mode, with one IFO operator on shift at Hanford
and two at Livingston, in addition to scientific monitoring shifts manned
by one expert and one trainee (see below).
-
Physicists have signed up for 13 suggested investigations similar to the
ones carried for E2 (see below).
-
One purpose of this meeting is to field requests from investigators for
special running conditions or for data channels to be included in the reduced
data set (RDS). Comments for improving E2 procedures that didn't work well
are also welcome.
-
The followup teleconference will be the regularly scheduled detector characterization
meeting on Friday March 2 at 11:00 a.m. EST (8:00 PST). At that meeting,
each investigation leader should present a (brief) plan of investigation
and make any final special requests. Requests that are difficult to satisfy
should be made sooner than that!
-
DMT software writers are reminded that Monday February 26 is the deadline
for giving final DMT monitor code to John Zweizig for official E3 running.
Interferometer running configuration (Stan Whitcomb)
-
The Hanford 2K now runs stably in fully recycled mode, with lock stretches
of half an hour to an hour. The incident power on the recycling mirror
is 100 mW and gives about 100 W in the arms. The best sensitivity is at
about 1 kHz: 2 * 10-16 m/sqrt(Hz).
-
Each of the Livingston 4k arms has been locked individually for up to 40
minutes, but recombining the beams from the arms has not yet worked. There
is a high probability of achieving stable recombination by E3, but single-arm
locks will be used for the run if recombination locked stretches don't
consistently last 20 minutes or longer.
Shift Planning (KR)
-
The shift schedule is in good shape, with nearly every scientific monitoring
slot filled (present shift list). One more
person is needed for the last Livingston owl trainee shift.
-
The shifts will be synchronized at the sites, with day shifts starting
at 6 a.m. at Hanford and 8 a.m. at Livingston. Although one expert and
two trainees were present for scientific monitoring shifts during E2, it
was felt that only one trainee was necessary/appropriate for E3.
-
Shift duties for E3 will be quite similar to those for E2. There was some
discussion of how software bugs found on shift should be reported. In E2
the shift takers included bug reports in an end-of-shift online questionairre.
John suggested adopting the GNATS bug reporting system used by LDAS in
which bug reports in different software categories can be automatically
directed to the right person.
E3 Investigations (KR)
-
All tasks have at least one volunteer, but
some could probably use more help. In particular, the following tasks seem
a little short-handed:
-
Fraction of noise in 5-50 Hz due to seismic motion. Phil Charlton is leading
this, with help from Dick Greenwood and Neven Simicevic. Phil plans to
attack this problem using the data conditioning API of LDAS. KR would like
to see someone with DMT experience lend a hand too. It would be natural
for this team to work with the correlations investigation team led by Nelson
Christensen.
-
Data merging. Only Peter Shawhan was signed up for this at the start of
the meeting, and there is much to be done. Gaby Gonzalez volunteered to
help. Others are welcome. See below for discussion of existing merging
options.
-
Frequency noise. Only Rana Adhikari and Andrea Vicere are signed
up for this comprehensive and important investigation. Others are strongly
urged to join in.
-
The largest investigation team is for environmental disturbances (led by
Robert Schofield). This represents a consolidation of several E2 investigations
that seem in hindsight to have been split artifically. There are many disturbances
and many existing tools with which to study them. A more coordinated approach
seems appropriate here.
-
There was some discussion of how line monitoring for E3 should differ from
what was done for E2. Stan Whitcomb stressed the importance of looking
at correlations between the sites in the power mains. In addition to looking
in the GW channel for 60 Hz contamination, one can look at dedicated power
line monitors available at Hanford and soon to be available at Livingston.
Looking for correlations in glitches, drifts, etc. may require producing
trend or similar outputs at each site separately for later comparison (see
below).
-
Sergey Klimenko estimated he would need to look at about 200 GB of data
for line monitoring and wondered how best to go about it, since he doesn't
have that much free disk space at Florida. John said that running programs
on fortress at LHO and on decatur at LLO would probably be best. Sergey
wondered what other E2 teams had done to look at large data samples. KR
agreed to carry out a poll of E2 team leaders. Tom Nash expressed interest
in seeing the poll results, in order to help revise the part of the LSC
data analysis white paper dealing with computers and architecture.
-
John suggested that the various data serving programs should log
data requests to give us a better measure of what data samples people find
most useful. Peter reported that CACR already logs all such requests, and
any data passed to the Hanford general computing machines is also logged.
But internal CDS transfers are not presenlty logged. John volunteered to
speak with Dave Barker and/or Alex Ivanov to see if logging can be implemented
for the network data server.
-
The only investigation team to make a special request was the calibration
stability group (led by Mike Landry), which would like four hours of dedicated
time at the beginning and end of the run. KR suggested starting that effort
somewhat before the official Friday noon starting time for the run.
Mike wondered whether each investigation team was expected to station one
or more persons at each site during the run. KR suggested that small groups
focus on one site or the other for analysis to be done in real-time. The
calibration group does need to carry out work at both sites, however. Szabi
offered to help out at Livingston and said he believed Luca Matone would
be available to help too. Mike and Benoit will be at the Hanford site during
E3.
-
More on data merging and access:
-
Peter reported that there are two methods of data merging from Hanford
and Livingston available now, both somewhat cumbersome. One can read data
online but separately into two matlab programs for side by side comparision
(e.g. of seismic transients). Or one can transfer RDS frame files from
one site to the other and merge the files together (or transfer both to
a third location). The same data can also be obtained from the CACR archive
in a slightly different form.
-
Ed Daw pointed out that currently only the decatur workstation at LLO has
access to the Livingston RDS and that given the troubles experienced by
fortress during E2, we should not plan for physicists to be logging in
to decatur to get at the data. It's possible that NFS sharing will be available
at LLO by the time of E3 to allow other computers to see the data. At Hanford,
it's now possible to see the RDS data from both fortress and stone.
-
Analysis of data merged at CACR must wait for about a week after the run.
Stan pointed out that doing a merged-data analysis correctly is more important
than doing it in real-time.
-
Peter reported that the data viewer and diagnostic test tool have near-real-time
access to Hanford data from general computing machines outside the control
room.
Reduced Data Set Channels (KR)
-
A list
of RDS channels recorded during E2 was posted on the agenda page. E3 investigators
are asked to look over the list and request any additions needed for E3
no later than the March 2 telecon. The question was raised whether the
corresponding channels are present at Livingston. Szabi agreed to look
into this; he knows of at least a few differences in channel names.
Run Planning (Szabi Marka)
-
Szabi urged all DMT software writers to test their code soon on Livingston
computers. Ed pointed out that only one dedicated dmt computer (delaronde)
is available at LLO, in contrast to the two dmt computers (sand & stone)
at LHO. A dual-processor linux machine on order is unlikely to be installed
and commissioned in time for E3 to spread the load.
-
John mentioned that with a fix to the framecpp api he's working on, the
cpu demand on dmt machines might diminish considerably. He hopes to have
the fix ready by E3. If not, it may be necessary to perform triage on E3
monitors at LLO to avoid cpu saturation. The heaviest dmt loads come from
the data set summary generator (dss - Benoit Mours) and the transient identifier
(tid - Julien Sylvestre) which each requires roughly 1 cpu of the four
available on the sun-450 dmt machines. Benoit mentioned that he is speeding
up the dss program, but it will still require a large fraction of one cpu.
-
Szabi and John plan to run a test of all dmt monitors in the week leading
up to E3 to verify that the monitors run and keep up with the real-time
data.
-
Szabi mentioned that Livingston may be swamped with curious, pre-LSC-meeting
visitors. He was urged to be ruthless in ejecting idle hands from the control
room.
-
Szabi also mentioned that all disks at LLO will be backed up the day before
the run begins.