Locked Segments in LIGO E7
Locked, Clean Segments and Playground Data
News about Playground data (April 15): After
the playground lists below were posted, some bugs were discovered in
the code that generated the clean segments for the
interferometers. After a laborious and thorough process, Dave Chin has
now generated new lists of "really clean" segments, found in his web
page. (We should all thank his efforts in this!)
I have checked the new lists against our original triple-coincidence
playground data (PlayL1H1H2.txt). We had
16 segments in our list. The file CorrectedPlayL1H1H2.txt has the
original list, the new clean segments that have some intersection with
the original ones, and the intersection between our original list and
the new list. Here is a qualitative description of the changes:
- For most segments (1,2,3,6,9,10,12,13,14,15,16) the only
difference is that the segments should have ended 2 seconds
earlier. (One of the interferometers, not always the same, was
unlocked in those two last seconds, so those seconds should
look pretty bad!).
- Segment 4 (starting at 693793929) is now broken into two segments,
due to some H1 filters being switched on and off in the middle. Notice
that if only the L1-H2 coincidence is taken into account, this segment
is OK as in the original (H1 is the interferometer that makes the last
segment end two seconds earlier).
- Segment 5 (starting at 693836545) and Segment 8 (starting at
693961593) should end two seconds earlier AND start 4 seconds
later. This is also true if only L1-H2 coincidence is considered.
- The times in segment 7 (starting at 693943838) include times for
which there were missing frames in LLO due to the Y2K+2 problem. (See
more details about missing frames in the RDS web
page.). We could tell the interferometer was locked and the
interval was clean, but we have no data! There were some efforts to
recover this data, but I'm not sure what the status is.
- L1 in Segment 11 (starting at 694147877) has a particularly
noisier spectrum due to dewhitening filters not being switched, so
although it is a valid coincident segment, it is not "clean" in the
sense that the detector doesn't have the right parameters.
Since the consensus seems to be that L1-H2 coincidence is probably the
only one that matters, I include here a
list of L1-H2 coincident clean segments that have a non-zero
intersection with the original playground list. Notice that these are
in general (but not always) longer than the original playground
segments.
Original playground lists
Using the conlog tool and database information at the sites, we can
tell when the interferometer was being tweaked or filters were being
switched on. We use this information to find a list of "clean"
segments, when the interferometer was locked AND it wasn't
"touched". Most of the "touches" that may happen are certain to
trigger burst and inspiral searches, while the rest probably makes the
data unusable (such as alignment or calibration procedures). Also, the
interferometer does not have its typical noise spectrum until all the
right filters have been switche don after the instrument locks. For
all these reasons, data analysis should not be done on the "unclean"
segments. The file L1comments.txt
illustrates the concept, with a comment field for each segment, either
"CLEAN" or something descriptive of what was happening. Some segments
were broken into unclean portions leaving the clean portions as a
separate segment.
Dave Chin has followed this idea, and produced nice data, presented in
a web page:
http://tenaya.physics.lsa.umich.edu/~dwchin/LIGO/E7/LockSegments/.
Using this information, I generated some "playground sets" that the
bursts adn inspiral upper limits can use without biasing the future
results of astrophysical searches on the whole E7 set. These are the
steps I followed: I first found clean segments when L1 was locked, and
both H1 and H2 were not locked; clean segments when H1 was locked, and
both H2 and L1 were not locked, and clean segments when H2 was locked
but L1 was not (without a condition on H1). I then took away from
these segments the ones that were taken with L1 in "robust" mode, and
the times near or coincident with Gamma Ray Bursts, as reported by Szabi Marka to us. At some of
chosen times for L1, signals were injected. I did not exclude these
times, but rather added a comment for the segment. So, here is the
playground set.
The format used in each file is a list of initial GPS time, followed
by duration in seconds. We have in these sets a total of 13.8 hrs in
H1, 29.6hrs in H2, 27.6hrs in L1. A graphical summary of the chosen
data, day by day, can be found in a PDF file: E7PlayTime.pdf.
[[Note! On Feb 1, we
discovered D. Chin's lists of clean segments in each interferometer
were missing some short locked segments (shorter than
~5min). Therefore, in some of the segments in the lists above, there
is not a strict anti-coincidence: for example, in PlayH1noH2L1.txt, H1
or H2 are locked for few minutes at a time. We decided to stick with
the same lists, instead of shortening them to strict anti-coincident
times, since short segments will probably not be useful for
coincidence analysis anyway.]]
We add to the "playground" a list of 16 clean segments in which all three
interferometers were locked. The segments are between 15 and 20
minutes long, and they add up to 4.8 hours, or 10% of the 46 hours we
have for triple-coincidence analysis.
[[New (Feb 06 2002): I
discovered one of the segments in the list above was a time when
calibrations were being done. (I apologize for that, I thought the
clean segments excluded those times). The segment in question starts
at GPS 694147877 (Jan 4 2:51 UTC).]]
Finally, we have chosen a few segments of each list giving them
"priority" for any analysis to be done. We suggest looking at these
segments first, so we can collect information and compare results on
the same segments even if people cannot analyze the whole playground.
Enjoy the playground!!
Clean segments with injected inspirals:
Peter Shawhan injected inspiral signals in the interferometer hardware, as mentioned in elogs and in his email. I looked for the clean segments in each interferometer around the time when the signals were injected:
Summary table: how many hours of data?
E7 run for 18 days, or 432 hours. The numbers below follow from the
lists of locked segments presented here, and the lists of clean
segments generated by D. Chin, in here. They were updated after changes in Feb 1.
all segments segments >15min
L1 locked 284hrs (71%) 249hrs (62%)
L1 clean 265hrs (61%) 231hrs (53%)
L1 longest clean segment: 3:58
H1 locked 294hrs (72%) 231hrs (57%)
H1 clean 267hrs (62%) 206hrs (48%)
H1 longest clean segment: 4:04
H2 loked 214hrs (53%) 157hrs (39%)
H2 clean 162hrs (38%) 125hrs (28%)
H2 longest clean segment: 7:24
all L1, H1, H2
locked 140hrs (35%) 72hrs (18%)
clean 93hrs (21%) 46hrs (11%)
longest clean segment: 1:18
H2, L1
locked 160hrs (39%) 99hrs (24%)
clean 113hrs (26%) 70hrs (16%)
H2,L1 longest clean segment: 1:50
Plain Locked Segments
The files L1.txt, H1.txt and
H2.txt have a list of locked segments longer than
5 minutes for each interferometer. These were generated by stitching
together information from the metadata database on Segments, using
Guild. The source information so obtained is in H1IfoLocked,H2IfoLocked, and L1IfoLocked. The code that generated the list
of locked segments is a Matlab program, GuildSingleIfoStats.m. The Matlab
program also produces summary statistical information, and statistical
plots, found in FinalH1Figs.pdf,
FinalH2Figs.pdf,
FinalL1Figs.pdf.
Similar information for segments when two or three interferometer were
locked simultaneoulsy was generated with GuildTwoIfoStats.m and GuildThreeIfoStats.m. We also
generated information on times when one interferometer was locked and
the other two were not, or when one was locked and one other was not,
using the scripts GuildThreeMinusTwoIfoStats.m
and GuildTwoMinusOneIfoStats.m
This is the complete list of segments:
The Matlab commands used can be found in the script RunAll.m, and the statistical information obtained
(and posted in LLO elog) is found in the file FinalSummary.txt.
WARNING!!!! The criterion used for defining locked
segments is only a threshold crossing in the power stored in both
arms. There is NO guarantee that settings were not changed during the
locked segment. In fact, settings WERE changed in most, if not all,
the segments at the beginning and/or end of the segment. The change in
settings are almost guaranteed to trigger any burst or inpiral search.