I have been trying to tune the PID of this large octocopter without success (1050mm diagonal with aprox 11kg). I followed carefully all the steps of the tuning page and can get the copter flying ok with the pid values that I got manually. However, as I am still getting small overshoot when levelling and a bit of hot motors, I wanted to try autotune in order to get the supposed best pid values for the drone but the autotune procedure lowers the P term values so much that the drone starts to be very slugish and have a sort of pendulum effetc while stabilizing. P angle goes to around 2 and rate-P to around 0.090 which is the problem. D-term however stays at around 0.012 to 0.015 which I think is not bad but I am not sure.
Currently I don´t have the log files with me but in a couple of days I will share here.
I used inflight-fft harmonic notch and managed to get low gyro noise (at least I think) and have low vibes (x and y below 10 and z around 15 to 20). Rate.Roll.Out and Rate.Pitch.Out values seems also low being around 0.1.
I think HNOTCH is ok but if you could analyse it in the log it would be very helpfull.
I followed the configurations steps very closely and yes AUTOTUNE_AGGR is set to 0.1
An Auto Tue log isn’t the most useful. A post Auto Tune log is. But you can see right away the tune was not good when you see this disparity between Angle Pitch/Roll P:
ATC_ANG_PIT_P,11
ATC_ANG_RLL_P,1.912949 (this is too low for any craft)
So, set the parameters back to pre-roll Auto tune and make a simple hover flight in AltHold like you did to set the Notch filter and post that log.
And, update to latest Stable (4.3.2 or 4.3.3-rc1) for the IMU primary use fix for Cube Orange.
A few hours ago finally I was able to get a better tune (at least I think so).
Previously I was starting autotune with the PID values set by a previous autotune with wrong filter settings and too much vibrations which resulted in too high pid values.
This time I decided to set the PID values to the default and start autotune from there and it finally arrived to decent values.
Here is a log after the new autotune:
Comparing ATT.Pitch with ATT.DesPitch, I still see some small overshoot (same for roll) but maybe it is normal for a drone of this size… let me know if you agree and what else can I do to improve if possible.
I wanted to update to 4.3.2 but I have been delaying that due to other priorities. Now I think I will wait for 4.3.3 stable.
Looking pretty good. You could try lowering the Rate D term for Pitch and Roll a bit. Put it on an in-flight tuning pot with the upper range at .015 and the lower range at half that.
You could also try to re-run autotune. It looks good, but some times a second auto-tune will help dial things in. Just don’t forget to re-set Autotune_axis.