Fuzzy Logic Based Active Power Filter

Have you done a Laplace characterization of your supply control and feedback system? If you have then just do a Routh test of the entire closed loop system.

Hi Redfred. I thought I'd never see those words again.

Your solution may not be helpful, since the point of using fuzzy logic to dodge having to determine the systems characterstic equation. (I'm not saying that's a good thing) and instead substitute a series of rule based responses.

There's a interesting discussion about fuzzy logic by Bob Pease called "What's All This Fuzzy Logic Stuff, Anyhow?"

Do you honestly think that the rules found to work in the analog world can now just be ignored because you can tinker in software until you stumble into a software loop that follows the analog rules?

Easy Redfred, I was agreeing with you.

I read the Bob Pease posting on fuzzy logic. His reasoning is very difficult to refute. I remember when it was all the rage and then it just seemed to fade into the background. No doubt many doctoral candidates rode that wave for a while. Lotsa papers published on that.

Shamsmemon,

Welcome to real life. You cannot instantaneously feed back the result of a signal on this same signal and expect your simulator to be happy. The time delay lets the simulator digest the previous sample and generate the correction to be applied to the input before performing the next sample. Without this, your are calculating with a signal that is constantly affected by the result of your calculation.

In real life it is very likely that this feedback combined with the components delays will turn into a nice high frequency oscillator. For these reasons, real active harmonics filters are not done that way. (Each valid method is proprietary to different manufacturer)

You have to come up with your own.

Quite true. By doing a step response measurement of an existing or simulated power supply control system and then translating this into a Laplace transfer function, one will be including all of the delays that are part of the system. Combining this with a Routh test will permit an exact knowledge of the possible precision, response time and stability of the power supply. This will also identify the feedback control system requirements of gain and delay to obtain these supply characteristics without oscillating. More importantly this will tell the engineer if the desired supply characteristics can be obtained from the existing supply configuration.

I agree with your technique but what Shamsmemon is describing is impossible to do. This is why Matlab refuses to even try. Adding a sampling time removes the mathematical impossibility but does not solve the feedback problem.

If Shamsmemon wants to solve his problem, he will have to come up with a model with real time delays. Then he will be in position to follow your suggestion and maybe come up with a stable topology.

Active filters are much harder to design than it sounds once real life components are used. This applies to low power noise cancellation devices as well as power harmonics mitigation at 50-500A. I have been there.