UGF=5Hz now, the separation between different DOFs is more than 20db. You can increase the BW more but watch for gain peaking.
------
I decided to diagonalize based on PZTs (don't remember how it was done at EY). If you want to do it based on QPDs, just swap input and output matrix.
Sensing matrix was measured by turning on the servo with 3-8Hz band stop filter, then stop it holding the output, then injecing at 3.1, 4..1, 5 and 6.5Hz for PZT1_PIT, PZT2_PIT, PZT1_YAW and PZT2_YAW. Inverted the matrix to make a new input matrix such that POS=PZT1 and ANG=PZT2, and that the filter TF becomes equal to OLTF at low frequency where PZT response is flat.
I removed "calibration" filter bank as it's not useful. Now the filter comprises a simple integrator (FM1 and 2 together), "boost" (FM3) which I disabled as our UGF is much higher, and LF100 (FM10).
Integrator is scaled such that the absolute value is 1 at 1Hz. With just the integrator and LF100, filter gain is equal to the UGF.
The first plot shows the OLTF of all four DOFs when the filter gain was set to 10. Since the phase margin at 10Hz is already getting thin, I changed the gain to 5 after this measurement (UGF=5Hz).
The second plot (right bottom) shows that the loops are separated from each other by more than 20dB (look at red and blue for PIT modulations at 3.1Hz and 4.1Hz, green and brown for YAW at 5 and 6.5Hz).
When you need more BW for some reason, disabling LF100 will allow you to push to the gain of 50 or so. Since PZT starts to fall off from 20Hz or so, gain of 50 doesn't give you of about 30Hz UGF instead of 50.
If you need to have even higher BW for whatever reason, Stefan says that the MadCityLabs control box lets you switch to high BW mode which he disabled at some stage because the 2" PZT was oscillating with that setting.
I made EY QPD servo similar to EX (PZT based diagonalization, UGF5Hz without boost).
Black trace shows the OLTF when gain is set to 50 without LF100.
