Today we have changed the plant inversion for the DHARD pit loop (which is really differential ETM).  

The first two attachments show the data collected, the data used for fitting, and the fit both at a CARM offset of 50 pm and on resonance.  The third attachment shows the fit after some poles and zeros were removed by hand.  Our strategy was to invert the plant for the on resonance case, and aim for a stable but very low bandwidth loop to engage durring the CARM offset reduction.  The resulting controller is plotted in the 4th attachment, this includes FM3,4,6,7,9,10 in the DHARD filter bank.  We are using a gain of 16 and the signal is AS A 45 Q.  

On resonance, we now have a ugf of about 150 mHz, a gain margin of 10dB and plenty of phase (last attachment).  The RMS of the error signal is still dominated by fluctuations between 0.5-0.1 Hz.  This was measured with the oplev damping on at full bandwidth. 

We later turned the OpLve damping down and saw an instability at 0.55 Hz, which could be due to multiple UGF crossings in this loop.