Evan, Sheila, Elenna, Jenne
Last night as we tried to power up, we faced ring ups in our ASC loops that prevented us from reaching and staying at 50W. One of those ring ups caused a lockloss. Evan and I spent some time trying to track down what caused this lockloss. We looked at SRC2, as it seemed that those loops rang first. However, we looked at the oplevs, and clearly saw they were ringing in common soft pitch. See oplev ndscope. Evan and I next measured the CSOFT pitch loop at 2W, 10W, and 25W. I will add our CSOFT OLTF template to userapps asc/h1/templates/CSOFT after I post this alog.
You can see a plot of our CSOFT p measurement, where I have plotted only the regions where the measured had high coherence. As we power up to 25 W, a feature appears near 0.4 Hz where the gain drops, and the phase crosses zero. Evan and I have slightly different ideas about what this could be, a mechanical resonance, or possibly radiation pressure effects? Either way, this might be what is preventing us from powering up. Sheila and I think that a best short term move might be to increase the CSOFT P gain.
Also worth noting is that in the past we have also had 3.3 Hz ringing in CSOFT pitch, however the ring up is much slower. It does not prevent us from power up, but it does prevent us from sitting at high power without moving to our lownoise state. I have attached a plot of a previous higher frequency CSOFT P measurement, compared with a CSOFT pitch model that I have made (model in orange, measurement in blue). There is another feature around 3.3 Hz that could contribute to this instability. Raising the gain might make this feature worse, so we would need to think about how to approach updating CSOFT pitch to avoid both the 0.5 Hz problem and the 3.3 Hz problem.
Sheila and I noticed a 0.47 Hz resG in the CSOFT P filter bank in FM2, and guessed that it was designed to help us avoid this 0.5 Hz oscillation. I took the IFO to 25W, and turned on the resG. I remeasured the CSOFT P OLG at 25W with the resG, see attached plot. Since the resG did not appear to make things bad, I kept it on and took us to power up. And it worked! We have made it back to NLN and 50W power.
This is likely a stopgap measure, as we are not sure if this resG is the best solution for the low frequency instability. However, it works well enough for now without aggravating the 3.3 Hz instability, so I saved the change to SDF. I will update the guardian code to turn on this filter at the 25W state.
Marie Kasprzack emailed us to remind us that the 3.3 Hz issue might be related to QUAD damping issues that LLO saw and fixed in the past with additional damping infrastructure. We do also have that infrastructure, so Sheila and I will look into implementing it in the near term.
