PID control loops - closed loop feedback

[Larry Barello]

Originally Posted by maxlobovsky

as i understand it, the code sums the integral from start to finish, but every iteration of the loop it limits the integral. This seems like a very bad idea. essentially, the effective length of time that sum occurs is changing as the integral is being limited when it grows large. Also, for errors that continue in one direction, the integral error will not change because it will reach its max and keep trying to exceed that. This brings me back to my first question in this post. how long in iterations or in time should the integral sum over?

You must limit the integral term: what happens when your output is at 100% and you STILL have an error term? DO you really want to continue accumulating errors? It is a bad idea and will cause excessive delays in working out the integral term (the official name for this is Integral Windup) when you regain control.