PTB QuickTips: Iterm-related overshoot and PI balance

[@RubyS.1]

I like the format of this type of video. The new release videos are really good too

[@DesignTuneOperate]

Gold as always

[@robmulally]

Amazing thank you

[@JusticeFPV]

Great video Brian! yes, I fell foul to this too! this came from an old Mark Spatz video that recommended raising iTerm to raise the step response line higher. I wonder whether others too have watched the same video and been mistaken.

[@PIDtoolbox]

We all fell for this idea at one point or another. But it’s also not entirely wrong in general. It’s just our miniquads are super responsive so even I term is operating at relative fast time scales, not like P, but relatively fast nonetheless, especially wrt 5inch. If we were looking at shorter time period it’d look like a dc offset so to speak. For some pid systems it’s possible I term works at a much slower freq where the oscillation might be much wider than the window we are viewing it in (500ms window), so at our current scale it’d also look like an offset. I wonder if we used low I gain and we took a wider window and examined more slow turning inputs it might then look like an offset as the copter drifts and remains constantly off set-point, idk. Anyway, we’re all learning from this

[@tomasspetersons4158]

I don't get it. On all racing tunes the I term overshoots so much , cause race tunes have low P and D term. No idea what should I do, as racing lightweight tunes are not recommended to go over 25 D and P around 35-45 ish.

[@tomasspetersons4158]

It seems like it's either lowering I, raising P and D or just raising FF (which should be the way), but the line still overshoots in pidtoolbox, while gyro follow the setpoint really well in blackbox. Not sure what's wrong.

[@VladVortal]

Thx

[@timothyciarlette8250]

Thanks.

[@marlin-fpv]

From my findings I value is best set to P +10-15%

[@duelz9366]

Hi there, so by looking at the step response, how do we know when I-term is correct relative to P?

[@drduc848]

So what does cause the offset? I don’t see it in your videos but it’s pretty much always there on my logs - independent of which frame/controller I’m flying or of my PID controller settings.

[@PIDtoolbox]

FF can do it. Also long flights with very small stick input coupled with noisy gyro. So in essence y offset is artificial and not a real property of our system (at least with the way in which we are testing and measuring). If it’s unavoidable then there’s the y correction button but should be used with caution. Ideally just remove FF for these tests and follow the basement tune method that involves continuous very deliberate stick input that exceeds at least 50-150deg /s for no more than 30sec. 10-15 sec of solid moves is all you need really. Experiment

[@drduc848]

That’s both useful and interesting. I will check but I generally have tuned without FF so it’s probably not that. My gyros are not particularly noisy, in fact mostly fairly clean. However, I tend to do my tuning flights by sitting in my garden and flying in the field next to my house which I cruise to and from, so that may be it. I will have a play and report back (with data 😃)

[@uavtech]

P:i balance FTW! :-)

[@PIDtoolbox]

exactly! gotta be ready for the new sliders

[@hugochiang1072]

PI performs a low frequency amplifier!

[@PIDtoolbox]

Please explain. I’m intrigued

[@hugochiang1072]

@@PIDtoolbox Without math: Iterm turns to have a long time affect on the response. With math: PI's transfer function has 1-pole and 1-zero.