This is a summary of the first 4 weeks of O3 on narrow spectral artifacts that degrade astrophysical search sensitivity to long duration narrowband signals (CW and/or stochastic GW searches). All search groups can be happy that the detectors are performing better than ever, so well done to commissioning teams!

Overall, there is a general reduction in spectral artifacts over O1 and O2, but there is still more work to do, especially when comparing H1 to L1. The 10 - 100 Hz band remains problematic, but there is good improvement in most of the rest of the frequency band. 

Data is taken from the Fscan automatically-generated SFTs of GDS-CALIB_STRAIN that are 1800 s long and use Tukey windowing. These SFTs are processed using the lalapps_spec_avg_long tool to generate the data files below.

O3 data can be found on the LHO cluster under:
~evan.goetz/CWDQ/O3/O3_[H|L]1_April2019.txt
where columns are given by
frequency, arithmetic mean PSD, arithmetic mean ASD, noise weighted PSD, noise weighted ASD

Attached are figures plotting this data showing the 10 Hz - 2 kHz band, and then zooms on specific regions (10-50 Hz, 50-100 Hz, 100 Hz bands from 100 Hz up to 1 kHz, 1 kHz - 1.25 kHz, 1.25 - 1.5 kHz, and finally 1.5 - 2 kHz). All plots show a comparison for H1 between O2 (June 2017 only) and O3 (first 4 weeks only). A final plot shows an O3 (first 4 weeks only) comparison between H1 and L1.

Some general points:
o While improved over O2, the 10 - 100 Hz band remains problematic for H1. Maybe there are some things done at L1 that can be applied also at H1 to improve this low frequency data
o There is good improvement in many bands; even as the broadband noise improved, more noise lines were not revealed. Good!
o Some new lines are present that were not there in O2 (dang, can't have everything...)
o Violin suspension resonances are spaced out more than in O2 (by design, I believe), but I can't tell if there is work to actively damp them during observing. The shoulders of the resonances are not too terrible, so maybe that implies some amount of damping work is in-place. When these ring up, lots of new lines can appear due to these loud peaks beating against other noise lines and introduce new noise into h(t). Keeping the violin line amplitudes to a minimum can help mitigate this effect.
o Some comb-like structures in the mid-600 Hz band seems to remain but with lower amplitude

Highest priority is to figure out coupling mechanisms of 10 - 100 Hz lines/combs and see what possible mitigation can be done. LLO may have some suggestions to assist with that.

Second priority is to understand some of the other line structures observed. Perhaps solving the low frequency line issues may also mitigate some of the higher frequency lines/combs

We will continue to monitor the data and provide further updates. Hopefully mitigations will help improve the spectrum! 

We struggled to reacquire after Georgia rephased the AS_A_RF36 WFS on Monday (alog 48857), even though her rephasing clearly made the ASC oscillations go away and improved some POP18 buildups.  Yesterday after maintenence I reverted the Monday rephasing, which enabled us to get locked.  At the beginning of the lock, we did not see the ASC oscillations, but after the IFO thermalized they were clearly there, and remained there throughout the rest of the lock stretch. 

I just gave a shot at rephasing the AS WFS that we use for BS, while we were locked with the arms still held off resonance.  BS ASC was using RF45 Q for its error signal (I had asked TJ to stop at ENGAGE_DRMI_ASC so that we were still on AS 45), and I drove BS in length (600 counts at 8 Hz in the BS M2 L2L filter bank).  I only changed AS_B_RF45 by +6 degrees, but I changed AS_A_RF36 by -20 degrees.  This is pretty different from Georgia's change of +25 degrees; our phasing results are 45 degrees different in DRMI-only versus full 35W IFO.  Interesting, and probably not so good.

Cause unknown so far. Bubba reported a dust devil heading toward the LSB OSB right around the time of the lock loss, interesting.

ETMY seemed to have moved a large amount after the lock loss. On the oplev, it went from +4 to -19 in pitch. I haven't checked the TMSY yet, but I had to move that a bit as well.

LLO informed us that they would be out of observing for ~20 min while they damped violin modes. Virgo was also down, so we got the okay to make a small change.

1639 Out of Observing

1645 Observing

TITLE: 05/01 Day Shift: 15:00-23:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Observing at 113Mpc
OUTGOING OPERATOR: Travis
CURRENT ENVIRONMENT:
    Wind: 9mph Gusts, 7mph 5min avg
    Primary useism: 0.01 μm/s
    Secondary useism: 0.10 μm/s
QUICK SUMMARY: Locked for 16.75 hrs, calm environment. The ASC Y oscillation is back, as expected after yesterday.

Andy, with input from Beverly, Laura, and Josh

The strong wandering lines in h(t) caused by the squeezer are becoming more frequent and more severe. This seems like it could be quite a problem for the searches, so I'm putting some information here to hopefully push this to be diagnosed and resolved.

The wandering line was noticed a week ago (alog 48772). It comes and goes, both in h(t) and in the squeezer channels. In the squeezer, it's an equally spaced comb of lines, with the first wandering from about 80 to 140 Hz. The line comes and goes quite suddenly. The first attachment shows what it looks like in one of the squeezer channels.

This seems similar to the 200 Hz wandering line from the LLO squeezer. The solution (LLO alog 43837 was a highpass before the TTFSS.

The attached PDF has spectra of several channels during the noise compared to a reference time. Nothing is visible on the fiber transmitted PD, but it is on the fiber PD, the IR laser LF, and a bunch of mixer channels.

Also of note is that the CLF RF6 and REFL RF80 RFMONs both get much noisier when the lines appear. There's a very small bit of extra noise also seen on FIBR_SERVO_DEMOD_RFMON.

I would speculate that this looks like a loop that's not quite stable, but everything is so coupled it's hard to tell where. Is it possible that the harmonics are generated by the noisy line going through the SHG and getting mixed with itself repeatedly?

   IFO has been locked and Observing for the past 4.25 hours, with a range of 111.3Mpc. The wind has dropped to under 10mph and primary microseism is down as well. Smooth first half of the shift. 

Shifter: Sumeet Kulkarni

Fellow/Mentor: Ling Sun

Summary:

• Average duty cycle 52.5%. No observing on Sunday. Longest lock period about ~18 hours on Friday.
• BNS Range averaged at 110 MPc. but showed lots of variation and dipped below 100 MPc. on a few instances.
• 48 Hz. spectral line / glitching continued to be visible, grew stronger as the week passed and peaked highest on Saturday. (Plot attached)
• Loud glitches caught throughout by hveto from H1:LSC-POP_A_LF_OUT_DQ with high significance.
• Whistles glitches were seen on all days of observing barring Tuesday and Thursday. They're correlated with motion in IMC-F and we observe a simultaneous increase in variance of the VCO frequency. alog48863
• Slowly wandering lines related to the Squeezer in h(t), first seen on April 21, were seen again on Thursday and Friday. alog48772
• Lost Lock towards the end of the day on Saturday at around 19:00 UTC due to high winds. Despite several attempts, the lock could not be regained on Sunday all the way until the end of the shift. alog48829 alog48830

Complete details can be found at https://wiki.ligo.org/DetChar/DataQuality/DQShiftLHO20190422.

Whistle glitches are being seen with an increasing frequency of occurrence starting from April 22.

The stretches of time when they were seen over the last week are as follows:

* April 22 (Monday): between 09:00-11:00 UTC; this coincided with an Earthquake.

* April 24 (Wednesday): between 01:45-02:30 UTC

* April 26 (Friday): between 13:00-16:30 UTC; this coincided with a period of high winds.

Today (April 30), they can be seen throughout the stretch of Lock between 02:00-15:00 UTC.

The whistles are accompanied by an increase in variance of the VCO frequency and a corresponding low-frequency (4-5 Hz.) glitching in the IMC-F motion. (plots attached from Wed. April 24).

While high winds and an Earthquake might have been the cause on a couple of instances listed above, whistles have also been seen when the wind and ground motion has not been high. It is not known what causes the increased VCO motion at other times.

Sharan Banagiri, Corey Austin, Philippe Nguyen, Anamaria Effler, Robert Schofield

Summary: For impulse injections from multiple locations, signals from accelerometers mounted on the HAM5/6 septum are the best match to DARM in arrival time, have the most consistent amplitude ratios with DARM, and have the most similar frequency structure to the effect of the impulse in DARM. Together with shaking data, this suggests that the septum is the dominant vibration coupling site in the LVEA at both sites. These vacuum enclosure impulse techniques promise to be a useful new tool for diagnosing scattering noise.

Shaking injections made throughout the corner station have shown, in recent PEM injections, as well as for some time before that, that above a few Hz, motion of the vacuum enclosure couples most strongly to DARM in the HAM5/6 area of both LHO and LLO. We have had trouble narrowing down the coupling site further using the shaking amplitude technique because the HAM5/6 region is relatively small and interconnected (if you shake one side of HAM5, the opposite side of HAM6 moves almost as much).

Impulse injection delays

To provide extra information, we investigated the use of impulse propagation delays to help identify coupling sites. In the past, we have narrowed down coupling sites by looking for the microphone that detects an acoustic impulse at about the same time that the signal appears in DARM. And, of course, we have tapped on the vacuum enclosure. But we hadn’t tried using propagation delays from impacts on the vacuum enclosure to accelerometers mounted on the enclosure. While the vacuum enclosure is made of steel, the propagation velocity of waves on the steel membrane and structure is much lower than the velocity for bulk steel, resulting in tens of millisecond delays for propagation between HAM5 and 6.

Figures 1 and 2 show, for LHO and LLO respectively, examples of impulse injection data. These and other injections indicate that vacuum enclosure impulses show up in DARM about the same time as it shows up on the septum accelerometer.

The down side to the impulse timing technique is that higher frequencies are needed in order to discriminate arrival times (here we have used a 70-200 Hz band). Thus, there is the danger that a coupling site that dominates at low frequencies is not the dominant site at high frequencies. However, we have checked bands at lower frequencies and not seen obvious differences and results from two other impulse-based techniques are consistent.

Impulse injection amplitudes

In addition to arrival timing, we also used the amplitude of the prompt impulse vs. the amplitude of the prompt signal in DARM to discriminate coupling locations. The examples in Figures 1 and 2 show that the amplitude in DARM is most consistent with the amplitude of the signal on the septum accelerometer. An advantage of the impulse technique over our usual shaking is that there appears to be a greater difference in amplitude between accelerometers on HAM5 and those on HAM6 for an impulse injection than for a steady state injection. This may be because, for the steady state, the many late reflections have built up the amplitude to nearly the same at all locations in the region (in equilibrium, the energy gets distributed more evenly).

Impulse injection frequency content

A third impulse technique that points towards the septum at both sites, is a comparison of the frequency structure of the impulse signal in DARM to the frequency structure of vacuum enclosure resonances at the various accelerometers. Figure 3 shows that, for both sites, the frequencies that appear strongest in DARM are also the frequencies with the greatest motion in the signals from the beam-axis accelerometers on each site’s septum. That is, the septum resonance pattern appears to best match the frequency pattern produced by the impulse in DARM. 

While results from these three techniques are not individually overwhelming evidence that the dominant coupling site is the septum, together they build a strong case.

We started out using soft hammers to impact the vacuum enclosure, but this evolved into using a long rod, with an inch-thick soft rubber pad mounted at the end, that could reach the tops of enclosures and keep the user further away from sensitive regions. In addition, the extended impact associated with the pad and the flexing of the long rod helped emphasize lower frequencies. A photograph is shown in Figure 4.  For this study, all impulse injections were analyzed by hand. It may be worth beginning to automate these techniques.

In addition to the HAM 5/6 area, we also made impulse injections at end stations and other LVEA locations during the PEM injection program, which will be discussed in a future log (preliminary: at LLO EY impulse injections are consistent with coupling at the manifold).

The rephasing of AS-A, since reverted, caused a big bump in h(t) from 20 to 30 Hz that I think was responsible for the drop in range by 10-15 Mpc.

The rephasing (alog 48857) improves the 9 MHz buildup and the slow yaw oscillation. But the range (plot 1) drops from 115 to 105 MPc (roughly) due to a feature from 20 to 30 Hz (plot 2). It looks a bit like a scattering shelf but has a featureless spectrogram - no arches.

I don't see any coherence of the new noise with the MICH_P or Y channels. I've set up a BRUCO run that might help figure out where the excess noise comes from.

I have taken fresh LSC feedforward measurements, after the IFO had been at 35W for about 2 hours, and the SR3 heater had been at 4W for at least 2.5 hours.  This used about 30 min of commissioning time. 

Note to self for the future: All of the FF_to_DARM templates should give identical results, since the feedforward signals go through the same path to DARM (via differential ETM output matrix elements), so make sure that the frequencies that have coherence are good enough in one template for all 3 LSCs, then we only have to take one version of this measurement, rather than the 3 I took today.

Since we haven't had a PRCL FF measurement before, I copied templates from MICH and SRCL FF, and then used the PRCL excitations from Sheila's noise buget folder.  I also set the PRCL FF elements of the LSC output matrix to match the MICH and SRCL values of ETMX=+1, ETMY=-1.  I'll try to find a moment to fix the screen, but the FF output matrix screen doesn't display the PRCL column of the matrix, so these are currently hidden.

Data is saved in the usual place: /opt/rtcds/userapps/release/lsc/h1/scripts/feedforward/ , and it's all checked into the svn.  I did update the references in all of the templates, since I can't remember how old they are, but they're pretty old.