Function of PD, PI, ID and PID microcontrollers on a system

[Sam Groves]

Well different microcontrollers act differently on a system. A pure P microcontroller reduces the error and increases the final value , a pure I microcontroller nullifies the error and increases the final value and a pure D microcontroller increases the error and keeps the final value the same.

However when combining pure microcontrollers the net result will be somewhere in between right?

Since generally a microcontroller in the s domain can be written down as:K0+K1/s+K2s the net result depends on the constants K0,K1 and K2.Given K0,K1,K2 can we evaluate how the PID would act out?Is it like the RGB color or is it more complicated(K0 has greater weight than K1 or K2)?

 

[DaveE]

Is it like the RGB color

No

is it more complicated

Yes, it's more complicated. I don't think there's a good answer without knowing the nature of the system being controlled. Feedback compensation is tailored for each system to optimize performance based on desired behavior.

BTW, there's nothing special about PID. As a controls guy, I find it kind of annoying that that is what is always taught. You could have two integrator terms, no proportional term, etc. In fact my general advice is to just not use the derivative term unless you both need it and have a really good understanding of stability of feedback systems. It usually causes more problems than it solves.

Sorry, I don't know of a simple reference to explain this. You can search for "feedback compensation" to dig deeper; there's lots of good stuff on the web.

 

[Dullard]

@Dave:
I don't disagree with anything that you said. I think that the emphasis on 'PID' is a result of the fact that for most standard industrial control hardware (PLCs, etc.), 'canned' PID is what is available. Many of the people implementing controls in an industrial setting have little/no understanding of the (mathematical) theory behind the 'knobs' that they're turning when they 'tune' a system. There are some surprisingly good empiricists running around. They lack the classical training that EEs get, but their approach accounts for the practical limitations of feedback (accuracy, latency, etc.) that some EEs never seem to 'get'. Obviously, a trained EE (with actual experience) is ideal.

I agree that 'D' is much less commonly required. Where 'D' is rquired, you absolutely need to know what you're doing - It's not useful in a temperature control for an oven with a 12-Hour time constant. For servo-valves modulating the application of aggregate to multi-colored roof shingles zooming by at several FPS, it's absolutely required. 'D' can always (at least theoretically) 'improve' the response - knowing when it's worth the trouble is the trick.