I desgined two new PRCL controller:
- a better roll-off above 100 Hz, designed for PRCL runnign with a UGF of 30 Hz. This is called LP100
- a completely new controller with better rolloff, more gain below 1 Hz, similar gain at 3 Hz, less gain between 3 and 10 Hz where the PRCL signal is likely not sensing real displacement. This is called Dcntrl
First plot compares the open loop gains. The new filter banks are in PRCL2
I chnaged PRCL gain to have a UGF at 30 Hz and started testing the two new controllers.
I successfully enaged the LP100 controller, reducing the PRCL control above 100 Hz.
I caused a lock lss when agaging the Dcntrl. I made the mistake of having this filter at the end of the PRCL filter chain, after an integrator. This new controller increases the DC gain by 10, so I ended up kicking the PRCL control signal. To test it again, the Dcntrl filter module should be moved to PRCL1 at the very input, before the integrator.
The second test of the better PRCL controller succeeded. I mvoed the "Dcntrl" filter bank to FM1 of PRC1, right in front of the integrator. I also engaged the "reg3.4" filter module in PRCL2, to have 10 db more gain at the 3.4 Hz peak, since that's the dominant feature in the PRCL error point RMS.
The new controllers have the effect of
- improving the PRCL suppression at 3.4 Hz and below
- increasing the measured residual error signal between 4 and 20 Hz. However I think that's not real PRCL motion, so we don't need suppression there
- reducing the PRCL control signal everywhere above 30 Hz
The stability margins for thsi controllers are a bit tighter than the nominal controller, so we need to keep a close eye to the dropping PRCL gain over time
The new controller is now autometd in guardian in the state LOWNOISE_LENGTH_CONTROL. Tested once during the last lock acquisition