Yesterday, after getting some suggestions from Akutsu-san, I improved the wind shield of the SR2 OPLEV at both TX and RX side. After that, the measured residual velocity immediately reduced by at least a factor of 10, meaning that the noise level is indeed suppressed by the wind shield. However, there are some peaks observable in all signals at around 8-10 Hz (see figure 1), which make the longitudinal residual rms velocity to be at around 1-2 µm/s which is still above the type B requirement (0.5 µm/s as stated in Sekiguchi-san's thesis. So, I decided to measure the residual motion from 0 - 5 Hz and make this the current standard of measure residual motion. 

After that, I measured residual motion again with TM damped and the result was not satisfying so I attempted to increase the gain of the TM damping filters but then it has gotten worst. I guess this was because I have modified the OL setup by a little bit while I implement the wind shield and this invalidates the diagonalization matrix that I have implemented previously.

Today, I attempted another run of diagonalization to decouple L, P and Y. Then, I measured residual motion again and the result was still not satisfying. I then noticed that the coils are very imbalanced so I attemped to balance the coils by setting up a pringle mode and rougly balanced the coils (still need improvement). See figure 3 for the current gain adjustments. After that, I measured new TFs and redesigned the damping filters.  Although there is significant improvement in pitch residual velocity , there is still no significant reduction in L residual velocity and it stays at around 0.6~0.9 µm/s. See figrue 4. 

One thing to note is that there is a peak at ~ 0.2 Hz which coincide with the typical seismic peak. This could mean that IP damping is introducing low frequency ground motion to the optic because the LVDT is coupled to the ground. Note that this is not the ground motion measured by the OL (as it is also coupled to ground) because the peak is observable in P and Y. Since the inertial damping implemented in 8905 has a much higher gain than the usual damping, it could introduce even more seismic noise to the optic at low frequency so I backed off and copied the normal damping filter to inertial damping filter block and make it a temporary inertial damping filter while disabling the high gain filters. The 0.2 Hz peak contributes to 0.2~0.3 µm/s of the total residual velocity so it would be nice if we could get rid of it. 

The longitudal rms residual velocity increased gradually as opposed to increasing in sharp steps at resonances. This means that the noise level is still high and the measured rms is actually accmulated noise that spans 0 - 5 Hz. Therefore, other efforts should be made to further reduce the overall noise of the OL or else we have to declare this as the lowest residual velocity we can measure. I will keep trying to get the residual motion to where we want for the type B paper because we are so close