Yes, that is indeed an incorrect picture- when you drive, you don't veer all over the road crashing into thing (unless you're my ex-wife). That's because your reaction and correction is rapid compared to the variation that you're controlling. The feedback signal at audio is essentially instantaneous- stability criteria are based on phase shifts within the circuit, and all of that was worked out long before you or I were born.
A basic text in control theory would be beneficial to an intelligent guy like you. I'd love to see if understanding that ends up leaking over into your excellent work in transducers.
Sy,
I'll do that. Do you have a reference book I can read that is not more math than verbal concept explanation?
I do understand your better picture of feedback in a driving situation. I just didn't know how to say it in a cogent way. Made me think of those old cars with worn out steering gear and the continuous corrections needed to try and go in a straight line. At least my wife could drive and I could actually close my eyes when she drove, some people make you want to jump out!
I'll do that. Do you have a reference book I can read that is not more math than verbal concept explanation?
I do understand your better picture of feedback in a driving situation. I just didn't know how to say it in a cogent way. Made me think of those old cars with worn out steering gear and the continuous corrections needed to try and go in a straight line. At least my wife could drive and I could actually close my eyes when she drove, some people make you want to jump out!
That's not a bad analogy.
Unfortunately, you really do need math to "get" it to the point where you can really understand and predict stability (as well as other aspects of time and frequency domain response). Bruno Putzeys did a super nice layman's treatment that had some math, but it was very basic and (I think) an incredibly clear explanation that can allow you to cut through the smoke that some people love to blow. Have you read his paper "The F-Word"? If not, it's on the Linear Audio site.
Unfortunately, you really do need math to "get" it to the point where you can really understand and predict stability (as well as other aspects of time and frequency domain response). Bruno Putzeys did a super nice layman's treatment that had some math, but it was very basic and (I think) an incredibly clear explanation that can allow you to cut through the smoke that some people love to blow. Have you read his paper "The F-Word"? If not, it's on the Linear Audio site.
Sy,
I did actually read that paper by Putzey. It and the book by Cordell made it pretty clear that the idea that feedback isn't fast enough in the realm of things to not cause problems ended my wonder about that, I of course intuitively thought it couldn't be fast enough and saw the error of my thinking. I also understand the inviolability of the second law. Otherwise I have a perpetual motion machine to sell with a bridge in NY.
Snup,
Take a look at what it takes to put together a magazine and all the work to do it and I don't think you would consider for a small publisher that the cost is really that high for Linear Audio.
I did actually read that paper by Putzey. It and the book by Cordell made it pretty clear that the idea that feedback isn't fast enough in the realm of things to not cause problems ended my wonder about that, I of course intuitively thought it couldn't be fast enough and saw the error of my thinking. I also understand the inviolability of the second law. Otherwise I have a perpetual motion machine to sell with a bridge in NY.
Snup,
Take a look at what it takes to put together a magazine and all the work to do it and I don't think you would consider for a small publisher that the cost is really that high for Linear Audio.
Where negative feedback can "get into trouble" is that in order to correct a non-linearity it may attempt to force the circuitry to work at higher level of performance then it is intrinsically capable of, hence introducing noise-like artifacts within the circuit. I've spent some interesting time with Spice, driving a valid amplifying circuit using classic feedback, with signals on the edge of what it can comfortably deal with, and you can see the normal, "nice" behaviour breaking down - all sorts of glitches 'magically' appear at various points in the circuit, as the circuit momentarily enters a state of incompetence, courtesy of the feedback now asking too much from it.
The principle of entropy in information theory equates to the one in thermodynamics. Separating signal from noise with no additional information is like free energy, exactly what Bybee claims.
Putzeys article, Waldhauer book
I strongly recommend the Putzeys article mentioned as an introduction.
Another book that I like, and of whose author Cordell has spoken highly, is Waldhauer, Feedback, Wiley, 1982, ISBN 0471053198. Waldhauer has a splendid Preface in which he tells of how his approach sidesteps the conceptual difficulties in which we tend to get mired: "The intuitive understanding of feedback systems at the most elementary level has been made more difficult than necessary by an 'endless chain of dependencies' that seems to arise whenever we attempt to analyze a feedback system."
I'm too lazy to excerpt a lot more (and it is under copyright), but here is a bit: "To express the distinction between the new theory and the old, we use the term 'anticausal analysis' to describe the direction of analysis from output to input."
The book is suitable for self-study. Still in print, evidently, a bit pricey but used copies appear to be fairly plentiful.
I strongly recommend the Putzeys article mentioned as an introduction.
Another book that I like, and of whose author Cordell has spoken highly, is Waldhauer, Feedback, Wiley, 1982, ISBN 0471053198. Waldhauer has a splendid Preface in which he tells of how his approach sidesteps the conceptual difficulties in which we tend to get mired: "The intuitive understanding of feedback systems at the most elementary level has been made more difficult than necessary by an 'endless chain of dependencies' that seems to arise whenever we attempt to analyze a feedback system."
I'm too lazy to excerpt a lot more (and it is under copyright), but here is a bit: "To express the distinction between the new theory and the old, we use the term 'anticausal analysis' to describe the direction of analysis from output to input."
The book is suitable for self-study. Still in print, evidently, a bit pricey but used copies appear to be fairly plentiful.
Barrie Gilbert did this in that article that JC likes to quote on PIM. You assume an ideal output and work backwards to the input that has to be to make it so. I learned it in the 70's but don't know where it came from.
Cool.
Fred W. taught a course at Bell Labs based on (t)his approach. The book is very accessible and applications-oriented as well. I referred Samuel G. to Waldhauer's remarks on the noise of ballasted current mirrors, which he cited in one paper.There is also a very nice philosophical discussion about stuff, including why it wasn't so crazy for people to insist that the world was flat
Waldhauer was a versatile man and a profound thinker. I just learned from the web that he was one of the founders of EAT, Experiments in Art and Technology. I wish I could have known him.If his "acausal" approach can help to dispel the pernicious notion of the "dog chasing its tail", it will be worthwhile for that alone.
- Status
- Not open for further replies.