re-newed interest in Wollcott like designs
I have been reading thru the lengthy Cordell Interview - Error Correction thread on the SS forum recently, and have gotten a little more "iducated" on the Hawksford error correction idea.
Now I see that it IS formally EQUIVALENT to the older idea of a positive feedback loop within a negative feedback loop. (just move all the feedbacks back to a single input sum-er in the diagram) The implementation details are a little different (Hawksford EC correcting the output stage directly without additional high gain stages). Both approaches require a further global NFB loop around them with some extra gain to control stablility.
The Wollcott design is of course a specific PFBK within NFB design implementation among many. Seeing how the addition of just the inner loop to get Hawksford EC can drop distortion products by 30 dB (for nearly free no less!) in the Cordell design case, one has to look at the PFBK within NFB designs in a new light.
The gears are turning, lights flashing. Looks like a continuum between the two approaches is possible. Many possibilities. May be some interesting variations using low-pass band limiting of the feedbacks to eliminate those nasty re-entrant "distortion of distortion" products. A big premium on making the inner positive feedback stage(s) very linear is obvious. Both schemes pay off as something like the square of underlying stage linearity. (ie. 1% -> 1% of 1% or .01% resultant)
Don
I have been reading thru the lengthy Cordell Interview - Error Correction thread on the SS forum recently, and have gotten a little more "iducated" on the Hawksford error correction idea.
Now I see that it IS formally EQUIVALENT to the older idea of a positive feedback loop within a negative feedback loop. (just move all the feedbacks back to a single input sum-er in the diagram) The implementation details are a little different (Hawksford EC correcting the output stage directly without additional high gain stages). Both approaches require a further global NFB loop around them with some extra gain to control stablility.
The Wollcott design is of course a specific PFBK within NFB design implementation among many. Seeing how the addition of just the inner loop to get Hawksford EC can drop distortion products by 30 dB (for nearly free no less!) in the Cordell design case, one has to look at the PFBK within NFB designs in a new light.
The gears are turning, lights flashing. Looks like a continuum between the two approaches is possible. Many possibilities. May be some interesting variations using low-pass band limiting of the feedbacks to eliminate those nasty re-entrant "distortion of distortion" products. A big premium on making the inner positive feedback stage(s) very linear is obvious. Both schemes pay off as something like the square of underlying stage linearity. (ie. 1% -> 1% of 1% or .01% resultant)
Don
Re: re-newed interest in Wollcott like designs
That's why arxiv is a preprint archive. Let's see these results reproduced and other, non EM-gravitational link effects eliminated from being the cause. If this were a stock on the market, I wouldn't invest in it -- at least not at this point. The claims are extraordinary, and so should be the evidence, to paraphrase you-know-who.smoking-amp said:
Levitating moments: http://xxx.lanl.gov/abs/gr-qc/0610015
"That's why arxiv is a preprint archive. Let's see these results reproduced and other, non EM-gravitational link effects eliminated from being the cause. If this were a stock on the market, I wouldn't invest in it "
Yeah, I know.
But I want a flying saucer for Christmas, darn it!
Don
"That's why arxiv is a preprint archive. Let's see these results reproduced and other, non EM-gravitational link effects eliminated from being the cause. If this were a stock on the market, I wouldn't invest in it "
Yeah, I know.
But I want a flying saucer for Christmas, darn it!
Don
I think the EC idea is still a very good one. It is a little more clear now that it must be enclosed within an outer global NFB loop for stability or some feedforward used in addition to lower its pos. fdbk gain.
What has changed is my appreciation of getting gain from positive feedback for "next to nothing". EC, when error nulled, corresponds to infinite gain - conventional neg. fdbk.
Hence its ability to nearly completely null out error. (mathematically, the distortion in the subtractors limits this total null however).
But the positive feedback gain loop must be quite linear so as not to worsen faster than the effective correction rate, so it must be applied to situations where the signal dynamic range is small. The Hawksford design operates on the output gate drive signals in a floating near unity (closed loop) gain situation. (small signal with lots of bandwidth to correct crossover dist.)
Conventional positive feedback loops would usually be placed early in the signal chain to keep dynamic range small. The pos. fdbk loop in any case must be short, to keep phase under control in the audio band.
One must also check for problems under clipping conditions, where one or more loops might be de-activated. (a concern likely with the Wolcott design).
My interest now is with "active" ultralinear output stages. The error sensed correction signal can be fed back a'la Hawksford style to the grid drive(s) and some "sort of feedforward" applied to the output screen(s). (this "sorta feedforward" still needs to be checked on a simulator as to reducing the need for pos. loop gain. Too busy trying to get Visual Basic to control some rs232- Xantrex power supplies at the moment)
Don
What has changed is my appreciation of getting gain from positive feedback for "next to nothing". EC, when error nulled, corresponds to infinite gain - conventional neg. fdbk.
Hence its ability to nearly completely null out error. (mathematically, the distortion in the subtractors limits this total null however).
But the positive feedback gain loop must be quite linear so as not to worsen faster than the effective correction rate, so it must be applied to situations where the signal dynamic range is small. The Hawksford design operates on the output gate drive signals in a floating near unity (closed loop) gain situation. (small signal with lots of bandwidth to correct crossover dist.)
Conventional positive feedback loops would usually be placed early in the signal chain to keep dynamic range small. The pos. fdbk loop in any case must be short, to keep phase under control in the audio band.
One must also check for problems under clipping conditions, where one or more loops might be de-activated. (a concern likely with the Wolcott design).
My interest now is with "active" ultralinear output stages. The error sensed correction signal can be fed back a'la Hawksford style to the grid drive(s) and some "sort of feedforward" applied to the output screen(s). (this "sorta feedforward" still needs to be checked on a simulator as to reducing the need for pos. loop gain. Too busy trying to get Visual Basic to control some rs232- Xantrex power supplies at the moment)
Don
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.