Does anybody have an idea about the lenght of time constants involved in
PID regulation? I'm specifically referring to Integral action.

Briefly, for Integral action: at each tick, the difference between setpoint and
sensor value is evaluated and such error is added up to formerly calculated
values, probably eliminating the oldest one.

This algorithm should contrast backgroud sources of error (wind?) that
Derivative action doesn't detect (or not much) because those are not
that abrupt (betweek one tick and another, error has a small value).


So, the question is: what's the tick frequency? Can it be trimmed?

My experience is about Ardupilot for Pixhawk but any other information
is welcome.

In this context there is parameter ATC_RAT_RLL_FLTD defined "Roll axis rate
controller derivative frequency in Hz"  that seems to control Derivative
action tick. But nothig like that exists for Integral action.

I find relatively easy to trim for Derivative action parameters. For Integral
is a mess mainly because it is almost impossible to have the same background
conditions.


Thank you.

I had a prod around in the code, but I just don't know enough about Ardupilot. The hardware definitions have the frequency of the MCU, but it sounds like the various clocks are scaled off the back of that. I did find this post that makes it look like it's 480Hz for Pixhawk 1 and 2.

https://discuss.ardupilot.org/t/control-...ency/16914

I'd be very surprised if it's possible to run the different elements of the PID loop at different frequencies - at least not out-of-the-box. It would make a real mess of the scheduler.

(16-May-2022, 10:20 AM)Banelle Wrote: I had a prod around in the code, but I just don't know enough about Ardupilot. The hardware definitions have the frequency of the MCU, but it sounds like the various clocks are scaled off the back of that. I did find this post that makes it look like it's 480Hz for Pixhawk 1 and 2.

https://discuss.ardupilot.org/t/control-...ency/16914

I'd be very surprised if it's possible to run the different elements of the PID loop at different frequencies - at least not out-of-the-box. It would make a real mess of the scheduler.

In the meantime, thanks to lemonyLeprosy indications, I've found one page from Oscar Liang that is very interesting:

https://oscarliang.com/best-looptime-flight-controller/

The key points are:
- the sampling frequency
- the regulation loop frequency
- the various filters set-up

All that combined, make the matter rather hard.

As I pointed out, the relevant issue is to reproduce the same background conditions, of
course. Wind cannot be driven.  You change, you fly .. and you do not know.