[Akutsu (remote), Ushiba]

Abstract:
I reduced MICH and PRCL UGFs and added LP filters to reduce the control noise.
Then, MICH and PRCL feedforward was engaged.
Even though we just used the filters during O3, it's working and noise seems slightly reduced.
All the above changes are already implemented into LSC_LOCK guardian, so anyone can engage roll-off filters and feedforward by just requesting OBSERVATION state to LSC_LOCK guardian.

Detail:
Since current DARM signals have a large coherence with PRCL and MICH feedback signals, I optimized PRCL and MICH control filters.
Figure 1 and 2 show the OLTF of MICH and PRCL after optimization.
MICH UGF is about 16 Hz and phase marging is 35 degrees while PRCL UGF is about 22 Hz and phase margin is 42 degrees.

After optimizing control filters, I tried to engage feedforward of MICH and PRCL control.
This time, we just used FF filters during O3GK and any optimization has not been done yet.

Fighure 3 shows the spectra of DARM, MICH, and PRCL signals with several configurations.
Black, blue, and red represent without new filter, with new filter, and with new filter + FF.
As you can see, DARM error and feedback signals  around 100 Hz drastically improvede by filter optimization.
We can also see slight improve by FF at 50 - 80 Hz while slightly injecting noises around 150 Hz.
So, we probably need optimization of FF filters.

After the above changes, I implemented several corresponding engage state into LSC_LOCK guardian and confirmed it works well.
So, anyone can engage new filter by just requesting OBSERVATION state to LSC_LOCK guardian now.