Multiple negative feedbacks...

It's not necessarily bad, just complex to analyze.

Agustin Ochoa and I have been trying, for many years, to find a way to do this via simulation, sort of a multi-loop Middlebrook method... but have so far failed :-( ...Jim Thompson

Examples? Electromechanical control systems with multiple loops driven by multiple sensors but with one actuator are not at all uncommon. It's less so in circuits, but one example is where you drive a transistor from an op-amp, with feedback on the base (or gate) lead and more feedback on the emitter (source) or collector (drain) lead.

Generally it's done when you have a good high-frequency sensor (or node) that's not accurate at DC that you use for the inner loop, and the 'real' thing that you're controlling that is, for some reason, slow compared to the loop dynamics that you want to achieve.

Examples? The loops that I know of fall readily enough to linear control loop analysis -- if they're reasonably linear.

At the device level things aren't so neat. Auggie and I have been trying to find a simulation method for years :-( ...Jim Thompson

Maybe I misunderstand this but pretty much all my SMPS design have multiple loops. One for voltage, one for current, sometimes another one from a reflectometer bridge. Sometimes two or three act at the same time. Where'd be the problem?

Do you mean loops inside loops, or merged errors, like a constant-voltage/constant-current power supply?

Several circuits have been posted here recently that use slow DC feedback and fast AC feedback, from different circuit nodes. Like the opamp here...

ftp://jjlarkin.lmi.net/Isrc2.JPG

John

Do you mean within a device, or do you mean a circuit that's built up of a number of transistors, and is operating over a wide enough range that small-signal analysis is laughable?

It would seem that this would be a rich source of publication topics in things like the IEEE Circuits & Systems journal.

You may just find that there isn't a good way to do it -- nonlinear differential equations are just plain hard, which is why they're generally done by simulation or ad-hoc approximations rather than any attempts to solve them by hand.

IIRC the optimization of a system with multiple feedback loops is mathematically intractable problem. There is no simple and neat solution for this case. Everything depends on the particulars. So they avoid discussion of such systems in the textbooks.

VLV

When it comes to loop analysis by simulation, Auggie and I are the authorities ;-) ...Jim Thompson

Yep, except for PhD's who think they know it all ;-) ...Jim Thompson

You turned that thread into a flamefest, as you are trying to pollute this one.

John

If it's a linear system, and the optimization criterion is least-squares error, then you can express the system as a state-space problem and find an optimum. Then, if you want, you can put it back into a multiple-loop form.

Or you can optimize on min-max criteria and get an H-infinity controller.

I think that was one of the motivators of the development of state-space theory for control.

But it still all goes out the window if the nonlinearities are significant.

You can find the optimal control for a given configuration, but you can't find the configuration which is the optimal. That's the problem.

VLV

Good point. There's hope for my career yet. As soon as it becomes a crank-turner, then all the design work will go to China and I'll be reduced to making chain-saw sculpture and selling it in my front yard.

Does Larkin ever contibute any "value-added"? No, of course not. He is certainly a sicko and in need of a therapist. ...Jim Thompson

I'm with Joerg on this one. Multiple feedback is common in switchers.

Generally you don't want multiple loops controlling the same output variable if you don't have to do so. For instance. you could cascade multiple gain blocks with individual feedback, but generally putting the whole mess in one loop is the way to go. There are obvious examples when not to do this. For instance, if you have three gain stages, the circuit is difficult to stabilize if all three gain stages are in one loop. [Of course, the LT1028 can do this.]

There are situations where multiple feedback aids behavior, such as in overload recovery. Ladder filters are nothing but multiple feedback.

I don't know if you can do that!

It's a $1000 fine to cut a tree down around my area. The EPA may limit the emissions of your chainsaw. There is a noise ordinance, your saw may be to loud. Zoning laws will not allow construction in your backyard. Zoning laws limit what you can sell from your front yard. Just go take a cool shower, using your government regulated low flow shower head. Mikek :-(