Terminology: "open-loop" vs. "closed-loop"

Looks like I should be reviewing that article for clarity. Goody, more work!

In an ideal world you wouldn't have that complicated a setup. In an ideal world you'd just drive your desired signal into the controller, and look at the output from the plant.

But this is the real world, where we have noise, plus offsets and other worse nonlinearities (for a simple example try driving a double integrator with a sine or cosine in the time domain -- you'll see what I mean). So we need a way to get the system to settle to it's normal operating conditions, while still allowing us to take sufficient measurements to find the open-loop response.

What that setup does is allow you to operate the system in closed-loop (which is often the only way you _can_ operate a system to get a decent measurement), yet still collect enough data to determine the open-loop gain and phase.

Essentially you're exciting it with a sine wave (the input to the rightmost summation block), and you're taking the ratio of the two outputs (u_b and y). The gain and phase shift from u_b to y is the ideal- world open-loop response of the plant cascaded with the controller, even if you could never get the system to operate in open-loop mode.

I didn't wade through the article above, but there's various configurations that can be used to measure an open loop frequency response. The old 'HP' has an application note on this. I don't have the title or link right at the moment, but you can search it out if you try hard enough.

dave y.

I learned that method from the HP control systems analyzer that was designed to use it (and reference the thing's manual, in fact) -- the article would probably bear more than a passing resemblance to the application note.

Hi Tim,

Ok, thanks. I had the right idea in mind, then. I was just hung up on nomenclature.

No, I think the chapter is plenty enlightening.

By the way, I've just about got my act together to the point where I can start taking response measurements of my quad copter in the see-saw test jig.

Thanks again,

Rick

P.S. And thanks, Dave y for chiming in.

One thing I kinda remember is some measurement configurations are easier to achieve than others, and it's temping to take the easy way out, but the easy ones have a bias in the measurement. I think it came down to you had to break the loop to inject a signal vs. just measuring the system as is. Wish I could remember what that note was.

dave y.