This morning we sat in nominal low noise without going to observing from 19:21 to 19:51 UTC (Jan 9th) in a configuration that should be much better for the 1084Hz glitches. (WP6420)
On Friday we noticed that the 1084Hz feature is due to OMC length fluctuations, and that the glitch problem started on Oct 11th when the dither line amplitude was decreased (alog 30380 ). This morning I noticed that the digital gain change described in alog 30380 that was intended to compensate for the reduced dither amplitude didn't make it into any guardian, so that we have had a UGF that was a factor of 8 lower than what I used when projecting OMC length noise to DARM: 30510 The first attachment shows open loop gain measurements from the 3 configurations: before oct 11th (high dither amplitude), after october 11th (lower dither amplitude, uncompensated) and the test configuration (lower dither amplitude, compensated).
We ran with the servo gain set to 24 (to give us the nominal 6Hz ugf) and the lowered dither line amplitude from 19:21 UTC to 19:51 UTC Jan 9th. You can see the spectrum durring this stretch in the second attached screenshot, in the test configuration the peak around 1083Hz is gone, with just the pcal line visible, and the OMC length dither at 4100Hz is reduced by more than an order of magnitude. You can also compare the glitches from this lock stretch with one from yesterday to see that the glitches at 1084 Hz seem to be gone. This is probably the configuration we would like to run with for now, but we may try one more test with increased dither line amplitude.
Other notes because we don't have an operator today due to weather:
This morning all 4 test mass ISIs were tripped probably from the Earthquake last night that brought the EQ BLRMS to 10 um/second around midnight UTC. ITMY tripped again while it was re-isolating, no problem on the second try.
Richard topped added 400mL to the TCSY chiller around 10:15 or 10:30 local time, since we were getting low flow alarms. The flow alarms came back a few minutes before 11am local time.
I went through inital alingment witout problems and got to DC_readout transition. Then I measured the UGF of the OMC length loop in preparation for increasing the dither line height From that measurement and trends it became clear that when the OMC dither amplitude was reduced, the compensation of the OMC digital gain described in didn't make it into the guardian. This means we have been operating with a UGF in the OMC length loop that was a factor of 8 too low since mid october.
We arrived in low noise at 19:21 UTC with the OMC ugf increased to 6Hz. After about a half hour PI modes 27 and 28 rang up, and I wasn't fast enough to get them under control so we lost lock.
Here's a graphical version of what Sheila wrote, showing the time on Oct 11 when the 1083 Hz glitches started. The dither amplitude was reduced at 3:20 UTC, but the servo gain was increased to compensate. There are no 1083 Hz glitches at this time. Severe RF45 noise starts an hour later and lasts until the end of the lock. The 1083 Hz glitches are evident from the beginning of the next lock, and persist in every lock until the recent fix. The dither amplitude stayed low in the second lock, but the servo gain was reset back to its low value. Apparently, both need to be low to produce the glitches.
Keita tells me that people are concerned about making this change because of the increased noise below 25 Hz in the screenshot attached to the original post. We did not run the A2L decoupling durring this lock strech, and it was not well tuned. The shape of the HARD loop cut off at 25Hz is visible in the spectrum, which is one way of identifying bad ASC noise. The high coherence between CHARD P and DARM at this time is another way of seeing that this is angular noise (attachment).
So I think that this is unrelated to the OMC gain change and not really a problem.
1080Hz removal OMC gain/line conditions, does it make more low frequency noise?
Josh, Andy, TJ, Beverly
Conclusion: For two on/off times each for the two OMC gain tests (total 8 times) it looks like the high gain / low line configuration that takes away 1080 Hz (and also takes away some bumps around 6280Hz) coincides with a bit more noise below 25Hz.
Request: We hope this connection with noise below 25Hz is chance (It might have just been drift and we chose times unluckily) and we would like debunk/confirm it. We could do that with a couple cycles of on/off, (e.g. 5 minutes each, with the current configuration vs high gain / low dither configuration).
See the attached PDF. The pages are:
- 2,3: January 9th test configuration from Sheila's page: "We ran with the servo gain set to 24 (to give us the nominal 6Hz ugf) and the lowered dither line amplitude from 19:21 UTC to 19:51 UTC Jan 9th." Red/Orange are the test time with low line / high gain, and no 1080Hz
- 4,5: Similar experiment from October. Blue/green are the test time with low line / high gain, and no 1080Hz.
Also: There is no coherence above 10Hz between STRAIN and OMC LSC SERVO/I for any of these test times. So coupling must be non-linear.
Also: When the 1080Hz bumps disappear we also see a bump around 6280Hz disappear (attached image, sorry no x-axis label but its 6280Hz)
Our post crossed with Sheila's. If possible, we'd still like to see a quick on/off test with the A2L tuned. Could we have five minutes with the gain high and then ramp it down? Maybe with one repeat. Since this is a non-linear effect, we'd like to make sure there's no funny coupling with the CHARD noise. We're not too worried by excess noise below 25 Hz now, but it might be important when we're able to push lower.
While LLO was down I attempted to do a test by increasing the OMC length gain while in lock, which unlocked the IFO. So on/off tests aren't possible. Edit: I broke the lock by changing the gain after the integrator (which had been OK when not on DC readout), we can change the gain upstream instead without unlocking.
For now I put the new gain into the guardian so the next lock will be with the increased gain, and hopefully see that the low frequency noise is fine.
Now we have relocked, Patrick ran a2l, and Jeff, Evan, Krishna and I did an on off test by ramping H1:OMC-LSC_I_GAIN:
- high gain for about 15 minutes before 23:55 UTC Jan 12th
- low gain from 25:56:20 Jan 12th to 0:03:21 UTC Jan 13th
- high gain from 0:04 UTC to 0:13:22 UTC
- back to low gain
The attached screen shot using the same color scheme as in the presentation above shows that there is not a difference at low frequency between high gain and low gain.
We are back in observing in the low gain configuration, but the gain is set in an unusual way (and accepted in SDF so that we can go to obsevering). Keita would like us to hear confirmation from detchar before making this permanent.
Thank you Sheila, this looks really good. No change at low frequency. 1080Hz gone. The 6280Hz just varies on its own timescale. From our end we're happy with the configuration change since it only does good. Sorry for the red herring about low frequencies.
