Since we spent so much time on the fast locklosses this afternoon and early evening I decided to work on a different problem.  We need to turn up the bandwidth of CSOFT P in order to be able to move our soft offsets around after we power up, but it seems that a cross coupling with CHARD P is making the loop unstable with higher gain. 

The guardian now engages a version of CSOFT P which is the same as the rest of the soft loops (FM1, 3+4,5,9,10) we can change to the filters we used for the high bandwidth version in O2 by turning off FM5 +FM10 and engaging FM5, ctrl2g, who's gain Stefan lowered by 20dB today.  This can be done with a gain of 3, 6 is too high and oscillates.  

I took a few long measurements to try to understand why this loop is now unstable.  The attachment shows a measurement, with the plant TF shown on the right.  You can see that there seems to be a zero where the plant looses phase at around 0.75 Hz.  This seems very similar to the problems we had with CHARD a few weeks ago, (43946)

I tried turning off the top mass offloading for pitch on all 4 optics, and repeated the measurement from 1Hz to 0.6Hz, and it was unchanged.  I took another measurement with the CHARD P gain reduced by a factor of 3, (second attachment) and you can see this has a large impact on the loop shape.  While the shape of the CHARD loop has an important impact on the radiaiton pressure instability in CSOFT, I think that at 2W this shouldn't be important and this is probably just a case of cross coupled loops which can perhaps be fixed by changing the sensing for soft.