Gabriele, Camilla, Elenna
We have not updated the in-loop LSC feedforward since the increase in the DARM offset and the new SRCL offset (also small increase in MICH gain). These changes had a large effect on the feedforward, and Jenne has been able to subtract some but not all of the LSC noise using nonsens cleaning. I made three injections last night that would allow Gabriele to fit the feedforward and update the filters. I ran a MICH and SRCL injection, as well as a filter bank injection (input off, filters off, gain 1) to get the other preshaping effects in the feedforward. I did not run a PRCL injection, as we will need to update MICH and SRCL first, and then measure PRCL. The PRCL fit will depend on the MICH and SRCL feedforward.
The injection measurements are saved in "/ligo/home/elenna.capote/LSCFF".
Both new filters are labeled "5-16-23" and are placed in FM9 of MICHFF and SRCL1FF respectively. These filters should be used with a gain of 1. I updated the ISC_LOCK guardian to use these filters in lownoise_length_control.
Given that the PRCL feedforward is so dependent on MICH and SRCL FF, in the event that we are unable to implement a new PRCL FF in a timely manner, it is worth checking if the current PRCL feedforward is still improving noise or injecting noise with the MICH and SRCL update.
Unfortunatelt this made the LSC noise worse, which is the exact opposite direction we want to go. I am reverting the change.
The error was likely a sign flip. These filters should be tried one more time.
Overall, this implementation was a failure. The new SRCL filter did not work and caused a lockloss (there was a large high frequency portion of the filter we forgot to correct for). The MICH filter increases the noise in DARM. Both Gabriele and I are confused why the new feedforward has failed to improve the noise.
However, I think we need to try again. I am attaching a quick coherence I took from this lock, showing the high coherence from all three LSC loops. This is costing us significant range. This likely explains most of our range drop after we implemented the new DARM and SRCL offsets.