[Sheila, Gabriele]

The REFL DC centering loops (DC1 and DC2) were not working very well. We measured the loop transfer functions for all pitch and yaw degrees of freedom, and tried to tune the gains. We were confused by the different shapes of the loops, the different signs and gains.

Therefore:

We measured again the output matrix to RM1 and RM2, by adding DC offsets at the suspension M1 stages and measuring the motion on REFL_A and REFL_B. We did this with single bounce from the PRM. The old matrix was completely wrong. The plot below shows the effect of adding offsets at the output of the DC1/2 pitch/yaw loops, before and after inverting the matrix. Before the loops were completely coupled, after they are quite well decoupled.

After this, we fixed a couple of more things in the control filters: a different sign in DC2_P, and reduced the Q of all ~1 Hz zeros to 3 (from 7). With this configuration, all centering loops can be closed with the same control filter, with a bandwidth of about 2 Hz, same sign in all cases, gain of -200 in pitch and -300 in yaw. The plots below shows the measured open loop transfer functions.

The configuration of the DC1 and DC2 and of the output matrixes is shown in the attached plots.