We lowered the HARD YAW loop gains at the LOWNOISE_ASC state. Please see the attached plots for the comparison between new (red traces) /old (blue traces) OLTFs.

For CHARD YAW (first plot), it seemed that we lost some phase in it and our original loop had only 10 deg of phase margin... I lowered the gain from 0.75 to 0.3 so that we have a UGF of 3 Hz and 30 deg phase margin. As we now have digital SS torque compensation to keep the plant power-independent and ISIFF to reduce the microseismic motion, it seemed that 3 Hz (or even lower) UGF should be sufficient. On the other hand, the  phase at > 3 Hz needs some further investigation. With such a small phase margin we cannot adding extra low-pass to reduce its noise further.

For DHARD YAW the phase behavior is much better, but still for noise consideration we still lowered its gain from -60 to -40 to get a roughly 3 Hz UGF. From the measurement it suggested that we could engage FM2 in DHARD_Y to further reduce its noise, but before we can test it a fast lockloss happened.

I put the new gain settings into the guardian for now (without engaging the extra LP in DHARD_Y FM2). Before the fast lockloss happened I didn't see ASC oscillations due to the reduction in the ctrl BW, so it should be fine. After I relocked the IFO, I saw the violin modes got rung up again... The ETMY mode 1 and 6 were high (RMSLP_LOG10_OUTMON reaching ~ 5.5 or so), and the ITMX mode 1 reached RMSLP_LOG10_OUTMON value of almost 7...

I took the damping params from lscparams.py which seemed to be doing their work. However the damping was extremely slow, especially for the ITMX mode 1. For now I parked the IFO at the ENGAGE_DC_VIOLINS state (I also ran once the REMOVE_WHITENING in the OMC guardian to avoid DCPD saturation) and decided to go home. The new LSC MICH setup (LHO:44703) seemed fine for both grabbing lock and keeping IFO locked. I have not yet tested to go to NLN with the new, low-bandwidth C/D H yaw loops due to the high violin mode rms now.