[Giacomo, Matt, Lisa, Cheryl, Keita, Kiwamu, Paul]

With the mode cleaner locking easily again, we were able to measure the open loop transfer function at the IMC servo board. Three such transfer functions are shown in the first attached plot here ('IMC_OLTFs.pdf), with these settings common to each transfer function:

• Servo fast gain = -2dB
• Servo slow gain = 0dB
• Servo common mode 'compensation filter' 40kHz/4kHz ON
• MC2 M3 input filter gain = 1
• MC2 M3 lock filters: [150:4], [CLP100], gain = -1000
• MC M3 input filter gain = 1
• MC2 M2 lock filters: [0.1:1], [0.01:0.1], [ELP70], gain = 0.005

And these settings specific to each transfer function:

Low gain 'lock acquisition' mode

• Common mode gain = 0dB
• Common mode boost filter 1kHz/20kHz OFF

High gain mode

• Common mode gain = 24dB
• Common mode boost filter 1kHz/20kHz ON

High gain mode + notch filter 

• Common mode gain = 24dB
• Common mode boost filter 1kHz/20kHz ON
• Analog notch filter placed between demodulator I phase output and servo input (see attached plot 'Notch_TF.pdf' for notch filter transfer function)

And now a bit of explanation:

From low gain mode to the high gain mode, we engaged the first common mode boost filter on the servo board, and ramped the gain up in steps, all the while looking at the OLTF up to maybe 500kHz. From this we expected that we should be able to push the UGF up quite a bit past 100kHz, but when we got past 24dB common gain, the loop became unstable (despite apparently having enough phase margin at the UGF. We looked out to a little higher frequency and saw the feature at 730kHz, which at 24dB was just poking above 0dB and apparently causing the instability. To cure this problem we tried implementing a notch + low pass filter at the servo input. This succeeded in squashing the 730kHz peak, but also lost us a few precious degrees of phase margin at the UGF. As a result, we still saw instability above 25dB common mode gain. The notch filter is no longer in the circuit, but will reside in Cheryl's office for a while in case someone finds it useful later on, or wants to modify it e.g. to make the notch Q higher. We discussed possible causes for the 730kHz feature - it may be related to the feature in the FSS OLTF measurement shown in entry 5093, though we couldn't be sure of the frequency axis in the attached plot. We have also reduced the common mode gain for now to 20dB, to keep the phase margin around 45deg. This leads to our new 'standard' locked high gain mode:

'Standard' High gain mode

• Common mode gain = 20dB
• Common mode boost filter 1kHz/20kHz ON

Maybe once we get an autolocker script running we can have the above settings as the final state (for now).

The notch filter design, made by Matt, is attached, along with a picture of the box itself, in case anyone wants to use this in future and wonders what it looks like.