PID Loops

[MSimon]

I have an article up at ECN on PID loops.

http://www.ecnmag.com/article-pid-loops-100509.aspx

Simon

[TallDave]

A good read.

I remember something of yours on that a while back. I had always wondered how those problems were solved.

[BenTC]

Finally you have the D term. It is the rate of change of the error multiplied by a the constant Kd. The purpose of this term is to keep the system from changing too fast.

I always thought that the D term sped things up, amplifying small changes, which is problematic in typical process environments with noisy pressure and level variables as things are pumped around. In my usual domain of mineral sands mining we only use PI and avoid D like the plague.

From "changing too fast" do you mean minimising overshoot? About the only use for D we have is temperature control, when it has a very large deadtime - and also since temperature is also a smooth/slow variable.

[93143]

He means "changing too fast" in the most basic sense. D stands for derivative. Of course if your sensor data is noisy and unfiltered, the derivative is going to be a huge mess and the D term may not have the desired effect.

And yes, it is supposed to add damping.

[KitemanSA]

Herr Professor MSimon,

In order to aid the understanding of generally, but not specifically, knowledgeable individuals, please include a written out version of any acronym you use, as follows:

"I want to take a look at Proportional–Integral–Derivative (PID) control loops and how they are typically handled in software."

As a mechanical engineer, the version here is quite understandable. Yours left me racing for Wikipedia. Just a suggestion.