I ran a few more autotunes today. In order to try and keep the results fairer each one was run on a fully charged 1.5Ah 4 cell (I have 8 identical batteries which I number and always use in order so they wear down evenly, a bit sad I know but hey ) I didn't log a 55Hz sample, just went straight for your suggestion of 45Hz and then 90Hz as it's a multiple.
Filter setup - TauRoll - tPitch - NoiseRoll - nPitch
LPC 45Hz LPO=1 - 20ms - 22ms - 70 - 57
LPC 45Hz LPO=2 - 23ms - 24ms - 53 - 45
LPC 90Hz LPO=1 - 18ms - 18ms - 85 - 86
LPC 90Hz LPO=2 - 16ms - 22ms - 89 - 57
LPC 90Hz LPO=3 - 17ms - 23ms - 96 - 63
I then flew each resulting autotune with the sliders set to 1.0 damping and 1.0%.
Unfortunately I didn't save the PIDs for each individual tune but they didn't vary hugely at all.
The quad felt most "locked-in" on the 90Hz/2 tune.
I wonder whether I am chasing my tail for the moment because I have another problem with this build / firmware which could be upsetting things and causing all sorts of issues. A minute or two into a flight it's throwing a GYROBIAS error on the screen and the pitch axis drifts quite badly on centre stick.
It didn't do that when previously on *flight, maybe I flash the other flight firmware back on and see how it handles. Perhaps the gyro has developed a fault of some kind.
**Forgot to add:
Motor curve fit I have set to 0.8 at the moment. At 0.9 the quad suffered with really bad shakes on high throttle. Do you reckon it's worth going lower?