As part of gearing up for aLIGO CW and Stochastic analysis detector characterization,
a number of us are developing and starting to use new or enhanced spectral line finding
tools. We plan to apply these various tools to this summer's OAT data in the hope of
1) making the tools better via exposure to real data; 2) identifying spectral lines early
on that can be mitigated; and 3) cataloging lines that we have to live with in aLIGO.

A wiki clearinghouse for OAT spectral line studies can be found here. For now it has
mostly documentation on tools and channel selections, but over time it will be populated
with study results.

One important tool is an enhanced version of the familiar Fscan program. Greg Mendell
has started creating daily Fscan plots for locked-arm data (defined by the mean value of
H2:ISC-ALS_EY_REFL_PWR_MON_OUT16 over 1 minute lying in the range 4300-8000 -- see wiki
above for explanation). The Fscans generated so far cover only the days of July 16 and 20,
chosen because each had more than five hours of locked-arm data.

I have used the half-hour SFTs generated during the Fscanning to look at average spectra
for the calibrated OAT feedback channel H2:SUS-ETMY_M0_LOCK_L_IN1_DQ for July 16 (5 hours)
and July 20 (9 hours) with 0.56 mHz binning. Two different averages were computed, one
being a simple arithmetic average, the other a noise weighted average that is almost, but
not quite the same as a harmonic mean. (The same method was used to compute unweighted
and weighted averages for h(t) for all of S5 and S6, as discussed in more detail here.)
Generally speaking, one expects the weighted average to be lower than the unweighted
average and to have a smaller variance. Large differences between unweighted and
weighted averages indicate non-stationary bands.

The figures below show the results:

Full spectrum (0-1000 Hz) for July 16 (unweighted and weighted averages)
Full spectrum (0-1000 Hz) for July 20 (unweighted and weighted averages)
Full spectrum (0-1000 Hz) for July 16 and 20 together (weighted only)
Zoomed spectrum (0-2 Hz) for July 16 and 20 together (weighted only)
Zoomed spectrum (2-10 Hz) for July 16 and 20 together (weighted only)
Zoomed spectrum (10-100 Hz) for July 16 and 20 together (weighted only)
Zoomed spectrum (100-200 Hz) for July 16 and 20 together (weighted only)
Zoomed spectrum (200-800 Hz) for July 16 and 20 together (weighted only)

Some quick observations:

The full spectrum is "rich" with structure (how much of it is relevant to
eventual aLIGO operations is not clear, however)
The July 20 noise is below that of July 16 over most of the full band (good to see!)
Here is a small sampling of line artifacts above 10 Hz (band of direct interest to the CW/Stochastic groups):

120 Hz is easily seen, but 60 Hz is much weaker in July 16 data (stronger in July 20 data -- was there a 60-Hz notch on this channel on July 16?) There is a peak near 180 Hz (179.9 Hz), but I don't think it's a 60-Hz harmonic. 
There is a persistent comb at 2.74 Hz seen out to at least the 25th harmonic in the July 16 data, but quite mitigated in the July 20 data
There is a very strong line at 29.375 Hz with rapidly weakening harmonics visible at 58.75 Hz, 88.125 Hz and 117.5 Hz 
Other strong lines below 200 Hz include 77.30 Hz (and harmonic at 154.60 Hz), 78.15 Hz, 89.54 Hz (and harmonic at 179.08 Hz), 90.40 Hz (and harmonic at 180.80 Hz), and 105.92 Hz
 


Chores ahead of us include dealing with some auxiliary channel issues in Fscans, getting Virgo's NOEMI line-finding tool to work with only locked-arm data, resurrecting automated coherence measurements between the OAT feedback channel and ~100 auxiliary channels -- and digesting all of the results once they appear. 

Future reports will be posted both here in the alog and on the wiki page.