After recently moving to Ardupilot 4.3 I tried autotune and was definitely impressed with the result. The heli (650 size) flies much better than any tune I could have produced. I am now feeling ready to start auto missions and have initially tested in loiter mode. I have a goal of reaching speeds of 16 m/s but feel there is a slight oscillation starting at higher speeds. I was hoping to get some pointers regarding what the quality of my tune is and how to safely proceed from here.
After autotune I have manually reduced the yaw rate gains slightly, and have added some roll rate D gain (was zero from autotune). Below are logs from a couple of high speed flights. The second have reduced the PSC_VELXY gains to see if the position controller was inducing oscillations - but I am not sure of the effect.
@epoulsen Based on the ATC gains that I have seen for my 600 heli, yours look fairly similar. I would say that the ATC_ANG_PIT_P and ATC_ANG_RLL_P could be raised to 6. I have pushed a fix for the Angle P test that will improve how it determines the proper gain. Along with that I would definitely recommend increasing the ATC_INPUT_TC gain to at least 0.20 to 0.22. The default input TC is too low for your heli. The higher input_tc value will slow down the time it takes to achieve steady state attitude for pitch and roll.
As far as loiter mode, It appears the pitch axis isn’t achieving the desired pitch attitude values. The increase in the pitch angle P gain should help. you could also raise the ATC_RAT_PIT_I gain to no more than your FF gain. That should help the actual pitch values match the desired values. Try loiter mode again and see if the oscillations still persist when you are decelerating.
Thank you very much for taking the time to have a look, it is greatly appreciated. I had a chance to test your suggestions today.
I have changed ATC_ANG_PIT_P and ATC_ANG_RLL_P to 6, and increased ATC_INPUT_TC to 0.22. Observing from the ground the helicopter flew much better with less wobble. Log file is here.
I also tried increasing ATC_RAT_PIT_I to 0.09. Log located here.
I am not entirely sure but I think pitch is tracked slightly better with increased pitch rate I. Here is an example with original (log 1) and with increased (log 2) pitch rate I :
