Stefan, Sheila, Hang

We realized that previously we kept losing locks at the CARM_TO_ANALOG stage could be explained by the following mechanism:

When we ramp up the CARM gain, it created glitches in the CARM loop. While the CARM loop could handle those glitches just fine, due to loop cross-coupling, the glitches would also show up in SRCL, and consequently saturate the SRM M1 stage.

To solve this issue, we modified the SRM M1/M3 L2L cross-over config.

The first approach was to reduce the SRM_M1_LOCK_L gain from -0.02 to -0.01. After this modification we successfully transitioned through the CARM_TO_ANALOG stage fine.

The second approach was to move the LP filter in the M1 stage from 20 Hz down to 12 Hz (in SRM_M1_LOCK_L, FM4), while maintaining the M1_LOCK_L gain to -0.02. This also reduced the high-freq component in the SRCL glitches. We were able to use guardian to transition through the CARM_TO_ANALOG stage twice with this approach without problems.

In the attached plots we also show the new SRM M1/M3 L2L TF in [m/ct] (with M1_LOCK_L gain -0.02, LP at 12 Hz), and the original M1/M3 L2L TF (M1_LOCK_L gain -0.02, LP at 20 Hz).