Quick tip: Do P-I balance last! 

You’re welcome my friend! Enjoy

That's reasonable for smaller rigs, and not that different from what I do, except that if you move P and I together, in particular on 5+ inch rigs, you would be potentially trying to dampen iterm-related overshoot. By pulling iterm out of the equation, and manipulating the damping slider, then you know for sure that any overshoot in the step resp curves is PD-related.

first, it's not necessarily the case the adding weight means you have to change PIDs. It depends on the distribution of weight more than just weight. So for example adding a heavier cam might suggest that pitch only might need different gains. But in either case, I'd be careful with this simple idea. In practice what usually happens is changing weight also has a big effect on vibration, and this dictates what has to be adjusted, which sounds like what happened in your case. So yea, even though I made that assumption in a previous video, it's not that simple, and I wish is was. WRT your second point re dc offset of the step response, the characteristics of PIDs on the step response for our quads is manifested as specific changes to the initial part of the step response, usually before about 200ms in the response. Even Iterm is represented as a broader slow overshoot in the step resp, but still operating within the 200ms of the curves. The dc offset in the curves is generally not tied to anything we are interested in. Usually when there is a dc offset it is related to noisy gyro and generally low signal to noise (when there's only small input and noisy output). So in that case I'd use the 'y correction' feature. The "smoothing" feature can also help (if there's good input but gyro is noisy). These are kind of bandaid solutions so I use them sparingly just to help with interpretation in these cases. NOW, that said, in extreme cases where Iterm is set to 0, you may see a slow offset issue where it looks fine during the first 200ms or so but begins to slope off from 1. In those cases I think this is real drift from low iterm and should be exacerbated by poor cog. I havent tried this test where I intentionally imbalance weight distribution. Anyway, it's tricky because the step resp algorithm is all done in the Fourier domain, so it's really tied to the spectral composition of the input and output signals, that's the beauty in it. I dont know how much of this long winded rant makes sense to you, but I hope it helps.

Ideally you'd set MM before iterm for the reason I stated here. To reiterate, as you increase the MM youare increasing the responsively of the copter, and so this reduces the set point the gyro latency, and when this latency is reduced, then iterm does no accumulate as easy, so then when you turn up iterm after you have MM sufficiently high, then you can tune Iterm higher (more optimally) before you see it accumulate and overshoot in the step resp. To take this a step further, which will probably confuse you but what the heck, when you engage a sufficient amount of FF then you could theoretically even push iterm higher again, if you wish. so one could argue that maybe iterm should be tuned last. That's valid. Only challenge is the step resp plots can be misleading when using FF. No method is perfect. If you can work within these limitations though, it's pretty good.

@@PIDtoolbox great. Thanks for the in depth reply. Makes perfect sense to me now. No method is perfect as you say but this one seems to be the closest to it at this stage. 😊 In the end it comes down to "feel" too, well for me anyways. But starting off with a good tune helps a lot.

it's all dependent upon the authority of your rig. If it's a lifter then there will always be enough lag to allow iterm to accumulate. Only thing that can help there is reducing iterm, lowering iterm_relax_cutoff, and increasing FF. you're always working within the limits of your rig.

@@PIDtoolbox thanks.

You’d have to move the other sliders up while maintaining the proper proportions. But if latency does not asymptote at 2.0 you should look to see if oscillation is creeping in anyway. The latency gain may also be minimal after that point anyway and one should be careful not to over interpret tiny decreases in latency of

@@PIDtoolbox I think I may indeed be getting some oscilations. I think I also pushed the PD slider up too much so I'll back that off a bit and redo the MM tests.