It’s not slight shaking, it’s a lot of shaking, at least to my taste.
But your case is not the typical “junk frame flying with defaults” thing with trivial advice. In particular, as video is involved - and apparently not in a commercial sealed package that can fly on a internal combustion engine plane and still manages all vibrations on its own - you need to get deeper.
First, on vibration objectives. If you want to have a smooth crisp clear footage, you have to become stricter on vibrations. Vibes of “below 30” is a requirement for the flight controller to still see the gravity vector. For video, you want them much lower than “below 30”, I would say that “at most 5 in hover” is a better objective, especially because they typically get somewhat higher when not hovering.
Your vibrations are not extremely bad, though the vibe profiles with and without the gimbal are remarkably different, and the “without” profile is worse. Which indicates to me that it’s not the gimbal which adds more vibrations, but that the machine was tuned with the gimbal present, and it feels unhappy without it. In which exactly configuration did you AMC your machine?
There is at least one indicator that the machine is tuned suboptimally. The value of ATC_RAT_YAW_FLTE is 1.0, which is its minimum, which means autotune of YawE essentially failed. One of the possible causes is poor yaw authority. From my observations, in such cases the result of autotune really depends on the initial values (a term “attractor” is not totally inappropriate for this process), and this is something one has to perform some guesswork about.
Related to that, in the flight with the gimbal, P and D contributions in pitch and roll are very low (the I contribution frequently outdoes them), which on its own is not yet too bad - but again in my experience with reasonably well-built machines this is an indicator that you are filtering too much, and the controller does not care about high-frequency noise - which may be reaching your gimbal, and which may have been compensated by reducing the amount of filtering applied and achieving tighter control.
In the flight without the gimbal, the contributions are still not extreme, but unlike the gimbal case, there is much more D-term than the P-term. This is what made me think you were tuning with the gimbal attached. Removing mass from a tightly tuned machine typically makes it unhappy about that. If you need both operation modes, either tune for the lightest one, or for both separately and switch tunings.
On “filtering too much”: I see you have configured the filters based on propeller diameter only, which is a sensible start, but it does not necessarily reaches best results. This is beyond the scope of default AMC procedures, so - while still being methodical - you need to learn more about your machine.
One way is to perform system identification flights, and to tune the machine according to what you actually see. @amilcarlucas has processes for this and I believe he can consult you, perhaps not for free though.
Another way is to start becoming an expert yourself, at least an expert in your particular machine. For this, you need to start paying attention to details, and to have enough courage to get beyond default parameter suggestions.
One such detail can be seen clearly in the noise plots. There is something in roll, which is seen as an uncompensated red-and-blue peak spanning 25-40 Hz here:
As not too much of different flight regimes were tested, I cannot say it with 100% confidence, but most likely it is a physical resonance. Before trying to fix this in software (with
FILT1_* and
ATC_RAT_RLL_NEF), you should try to eliminate the mechanical source, which will give better results. (By the way, it is not in the gimbal, because on the gimbal-less flight it is still present).
And returning to filtering too much, I think it is almost unavoidable that you eventually have to try increasing INS_GYRO_FILTER and the ATC_RAT_*_FLTD values and to see what this makes to your machine. The defaults are fail-safe recommendations that one starts with, but does not necessary stick to them. Unfortunately, this will require a bit more than three flights to nail down. But the results may be very exciting.
) this shaking behaviour?
