What has been done so far :

- Filters were installed in the lockin demodulator and were tuned for a 10Hz signal excitation.

        *9:11Hz bandpass for the optical lever signal

        *1Hz ellip low pass and 20Hz notch for the demodulated I and Q signals

- Lock2Osem matrix filled to excite the pringle mode of the second mass from the pitch oscillator (UL, LR = 1 LL, UR=-1)

- Sent a 10Hz, 50000cts signal from the oscillator, to excite the pringle mode of M2, but I couldn't see the demod Q and I changing when modifying the coil gains, so I quit after few iterations

- M2 length to M3 Pitch transfer functions, without coil balancing. It matches the model below 2Hz, cf attachement. The signal/noise ratio of the tf is too low above that. I had to divide the model supposed to be in rad/N.m by 100 to match with the op lev TF, and I'm not sure where this factor comes from. The red curve is a fit of the modeled tf, and the cyan is the inverted tf.

What I intented to do this evening is to install this inverted filter in the L2P M2 drivealign matrix, take a new L2P TF and compare it with the previous one, to see the improvement., but didn't have time to finish.

If anyone wants to take over : 

Script live in /SusSVN/trunk/BSFM/H1/BS/Common/Scripts/L2Pdecouplingfiltersdesign.m

TF data is in /ligo/svncommon/SusSVN/sus/trunk/BSFM/H1/BS/SAGM3/Data