Today Jenne asked the very reasonable question: "why do we need to update the LSC feedforward so much?"
Here is an accounting of the number of times we retuned the LSC feedforward since the engineering run and why.
- April, at 76W: the feedforward had not been previously tuned at this power and LSC noise was significantly limiting the low frequency noise in DARM 68709; We noticed that the noise coupling changed with thermalization, so the FF was updated to be tuned for the thermalized state 68934
- May, at 76W: We chose to retune because the DARM offset and SRCL offset changed, which changed the LSC coupling, 69646, 69785. The first alog here was a failure because it was my first time tuning the LSC feedforward and I made many mistakes. The second alog was also a failure, for reasons that we believed were due to thermalization problems
- June, at 60W: We went down in power, meaning that the feedforward was "updated" to the previous setting at 60W, 70648. Note, that 60W setting was a very out-of-date setting because we had previously relied on the NonSENS cleaning to clean the LSC noise out of darm. Gabriele and I retuned the feedforward to be better for 60W on June 28, 70925. We found at this time that the effectiveness of the MICH feedforward depended on what SRCL feedforward was on.
- July 5, 60W: We saw that there was probably some iterative tuning we could do to the June 28 feedforward, so we retuned MICH specifically, 71084. Note: this injection and fitting was done when OM2 was hot.
- July 12, 60W: OM2 TSAMS was turned off and the range suffered. One thing contributing to the drop in range was the MICH coherence increased. I believed this was because the appropriate MICH FF with a hot OM2 is different from a cold OM2. I based this conclusion on the fact that we have significant alignment changes when OM2 TSAMS is hot vs cold. 71285 Then, the OM2 TSAMS was turned back on and I believe we returned to the hot OM2 tuned ff, 71510. At the time Jenne still noted some residual MICH coherence
- July 20, 60W: Gabriele discovered that the 1.34 Hz peak in DARM is due to the SRCL FF, so he implemented a new high pass filter to reduce this peak. A new high pass filter means the feedforward needed to be adjusted slightly. 71548, 71565, 71585
- Aug 4, 60W: Gabriele made a huge leap forward in reducing our low frequency noise with adjustments to the CHARD and DHARD Y controllers and M0 damping gain. After this update, a bruco indicated some coherence with MICH and SRCL that prompted another iterative tuning of the feedforward. 71935, 71961
Overall, yes, we have had to update the LSC feedforward quite a bit. I think some of this is expected. When we make a major configuration change such as IFO power, DARM offset, or TSAMS setting, that can significantly change the coupling of the LSC to DARM. As a reminder, the LSC feedforward had not been updated for the 60W configuration before we increased the power further, and both at 60W and 76W the LSC loops were a significant contributor to the low frequency noise, see 68869 and 68382 for 60W and 76W DARM noise budgets.
If we remove the significant IFO changes from the count, our reasons for updates are varied: better tuned for the significant thermalization at 76W, improvements to the sub-10Hz contribution with better high pass filters, minor updates with iterative tuning that are only possible after we have a baseline feedforward and an improvement in other competing noise sources, and finally because I made some mistakes on my first try. Approximately half of the times we have updated have been due to significant IFO changes; the other half have been the other reasons I just listed.
We might need to make further changes to the SRCL feedforward based on Gabriele's latest alog regarding the DARM RMS, 71994. We will try other avenues first, such as reducing the noise in SRCL, but if we update the high pass, we will have to update the whole SRCL feedforward.
Jenne's comment was related to LLO's approach, which is that they rarely need to update their feedforward. I think we have had more work to do to improve our low frequency noise (even to approach LLO's sensitivity), and much of that work will continue to turn up additional LSC coherence that can be reduced. However, if we stop changing the IFO configuration and resolve some of these RMS issues, I believe we will no longer need to make adjustments of the feedforward.
