Ovidiu wrote:
You can find my many hoax breaking analyses in the earlier posts of this thread. Try reading from about page 70 onwards.
If you have a theory that claims otherwise then I am all eyes and ears. Please post it for critique. But after reading your last sentence I will be compelled to spread its torn pieces all over this thread if it doesn't bear out what you claim it does. Be warned!
You can find my many hoax breaking analyses in the earlier posts of this thread. Try reading from about page 70 onwards.
If you have a theory that claims otherwise then I am all eyes and ears. Please post it for critique. But after reading your last sentence I will be compelled to spread its torn pieces all over this thread if it doesn't bear out what you claim it does. Be warned!
traderbam said:Ovidiu wrote:
You can find my many hoax breaking analyses in the earlier posts of this thread. Try reading from about page 70 onwards.
If you have a theory that claims otherwise then I am all eyes and ears. Please post it for critique. But after reading your last sentence I will be compelled to spread its torn pieces all over this thread if it doesn't bear out what you claim it does. Be warned!
Brian,
This is definitely not an academic forum and I'm certainly not going to start scribbling differential equations in ASCII. If I will eventually decide to publish my results I will certainly let you know.
Meantime, I would appreciate if you could answer the question and substantiate your statements. Please consider my opinion that an error does not qualify as "ignorant claim" unless you consider the author as ignorant in the matter. To save you a frantic search through this thread, no, I have made no statements regarding HEC and the matter. I'm only trying to defend a principle here.
one of these things is not the same
the Hawksford/Cordell error correction was shown to be equivalent to ordinary negative feedback by the early 60's - just another example of Hawksford's rather loose observance of academic polity in not citing earlier relevant material
http://www.diyaudio.com/forums/showthread.php?s=&postid=1070802&highlight=#post1070802
fig f is the "error correction" scheme's signal flow graph
I thought it was clear that Black's feedforward and Hawksford's "2nd" feedforward error correction were a completely different animal than negative error feedback with "FF" requiring 2 power output stages and a noninteracting power summer:
http://digilander.libero.it/paeng/feedforward_concepts.htm
it would be a feat to establish that this FF scheme is equivalent to conventional negative feedback given that it hasn't been shown in the last 70 yrs
Hawksford's sticking 2 fundamenatly different "error correction" schemes together in his general error correction diagram doesn't in any way establish their equivalence
conflating all 3 as a "challenge" to the equivalence of the Cordell Mosfet amp's error correction and negative feedback is simply nonsensical
the Hawksford/Cordell error correction was shown to be equivalent to ordinary negative feedback by the early 60's - just another example of Hawksford's rather loose observance of academic polity in not citing earlier relevant material
http://www.diyaudio.com/forums/showthread.php?s=&postid=1070802&highlight=#post1070802
fig f is the "error correction" scheme's signal flow graph
I thought it was clear that Black's feedforward and Hawksford's "2nd" feedforward error correction were a completely different animal than negative error feedback with "FF" requiring 2 power output stages and a noninteracting power summer:
http://digilander.libero.it/paeng/feedforward_concepts.htm
it would be a feat to establish that this FF scheme is equivalent to conventional negative feedback given that it hasn't been shown in the last 70 yrs
Hawksford's sticking 2 fundamenatly different "error correction" schemes together in his general error correction diagram doesn't in any way establish their equivalence
conflating all 3 as a "challenge" to the equivalence of the Cordell Mosfet amp's error correction and negative feedback is simply nonsensical
FWIW, I think there's some irony to the fact that Edmond's output stage (left one) is almost device for device identical to one used in some popular line drivers for DSL (millions installed).
So some of you might have it right under your noses (so to speak).
BTW there are passive error neutralization techniques that work and can be shown to not use/be NFB and are "invisible" to the main loop frequency response.
Edit - Of course the bi-polar line drivers do not use MOSFETs, but same circuit with power NPN's and PNP's. Power in this case still smallish but the modulation of quiescent current to peak current is >200x.
So some of you might have it right under your noses (so to speak).
BTW there are passive error neutralization techniques that work and can be shown to not use/be NFB and are "invisible" to the main loop frequency response.
Edit - Of course the bi-polar line drivers do not use MOSFETs, but same circuit with power NPN's and PNP's. Power in this case still smallish but the modulation of quiescent current to peak current is >200x.
Re: Re: Re: hec != hoax ?
Thanks, Edmond. I appreciate your sending me the paper so I could read it.
It turns out that this paper is not about Hawksford Error Correction.
The paper is about feedforward error correction and some early active error feedback schemes that disguised themselves as forms of feedforward. Not included in these was HEC.
First, there is no reference in the paper to Hawksford.
Secondly, it does not appear to anticipate or describe the HEC architecture.
If you are referring to Figure 3(a), this is NOT a Hawksford arrangement. Look closely, there is no positive feedback loop. The two forms of Active Error Feedback described in this paper by Vanderkooy are NOT the same as HEC.
The paper by Vanderkooy and Lipshitz that I was referring to came after my paper, and specifically referred to HEC. It showed that HEC also could be viewed as an implementation of negative feedback.
Cheers,
Bob
Edmond Stuart said:
Thanks, Edmond. I appreciate your sending me the paper so I could read it.
It turns out that this paper is not about Hawksford Error Correction.
The paper is about feedforward error correction and some early active error feedback schemes that disguised themselves as forms of feedforward. Not included in these was HEC.
First, there is no reference in the paper to Hawksford.
Secondly, it does not appear to anticipate or describe the HEC architecture.
If you are referring to Figure 3(a), this is NOT a Hawksford arrangement. Look closely, there is no positive feedback loop. The two forms of Active Error Feedback described in this paper by Vanderkooy are NOT the same as HEC.
The paper by Vanderkooy and Lipshitz that I was referring to came after my paper, and specifically referred to HEC. It showed that HEC also could be viewed as an implementation of negative feedback.
Cheers,
Bob
Re: Re: Re: Re: hec != hoax ?
Hi Bob,
That's right, the paper mainly deals with error feed forward, but on page 7 the authors made some remarks about error feedback (cited on post 2846).
Indeed, Figure 3(a) is not HEC, so what? I can't see any reason why their general remarks about error FB should not be applicable to Hawksford's error FB, the more so as the authors have shown (okay, in another paper) that also HEC could be viewed as NFB.
Maybe I missed your point and/or we are at cross purposes.
Cheers, Edmond.
Bob Cordell said:
Hi Bob,
That's right, the paper mainly deals with error feed forward, but on page 7 the authors made some remarks about error feedback (cited on post 2846).
Indeed, Figure 3(a) is not HEC, so what? I can't see any reason why their general remarks about error FB should not be applicable to Hawksford's error FB, the more so as the authors have shown (okay, in another paper) that also HEC could be viewed as NFB.
Maybe I missed your point and/or we are at cross purposes.
Cheers, Edmond.
Re: Re: Re: Re: Re: hec != hoax ?
Hi Edmond,
You asserted that Vanderkooy and Lipshitz discussed HEC before I published my paper. That was wrong. That is all.
Vanderkooy and Lipshitz did not write anything about HEC, to the best of my knowledge, prior to my paper at the AES Convention.
We agree that, subsequent to my paper, Vanderkooy and Lipshitz further analyzed HEC and also showed that it could be viewed as NFB. You are certainly free to view this as an elaboration on their earlier work.
Cheers,
Bob
Edmond Stuart said:
Hi Edmond,
You asserted that Vanderkooy and Lipshitz discussed HEC before I published my paper. That was wrong. That is all.
Vanderkooy and Lipshitz did not write anything about HEC, to the best of my knowledge, prior to my paper at the AES Convention.
We agree that, subsequent to my paper, Vanderkooy and Lipshitz further analyzed HEC and also showed that it could be viewed as NFB. You are certainly free to view this as an elaboration on their earlier work.
Cheers,
Bob
Re: one of these things is not the same
Hi jcx,
Thanks for reminding us of this piece. I agree, the signal flow graph in (f), which they call "model feedback" is the generalized version of HEC.
It appears that HEC is the special case where aPo * H * G = 1.
Notably, I think that is the only signal flow graph shown there that has a positive feedback loop.
To be honest, had I seen and read that in, say, the early '70's I'm not sure that I would have been inspired to evolve that into the HEC circuit as Hawksford did, both in terms of the special case coefficients and the output stage circuit topology.
As a Reviewer for the JAES, I often hold people's feet to the fire for not properly citing prior work, but I must admit that I would have missed this one if I was reviewing Hawksford's original paper. So I would not hold this too strongly against him. I still think he made a very honest, clever and effective contribution to the audio state of the art with his disclosure of HEC.
Cheers,
Bob
jcx said:
Hi jcx,
Thanks for reminding us of this piece. I agree, the signal flow graph in (f), which they call "model feedback" is the generalized version of HEC.
It appears that HEC is the special case where aPo * H * G = 1.
Notably, I think that is the only signal flow graph shown there that has a positive feedback loop.
To be honest, had I seen and read that in, say, the early '70's I'm not sure that I would have been inspired to evolve that into the HEC circuit as Hawksford did, both in terms of the special case coefficients and the output stage circuit topology.
As a Reviewer for the JAES, I often hold people's feet to the fire for not properly citing prior work, but I must admit that I would have missed this one if I was reviewing Hawksford's original paper. So I would not hold this too strongly against him. I still think he made a very honest, clever and effective contribution to the audio state of the art with his disclosure of HEC.
Cheers,
Bob
Re: hec != hoax ?
Hi Edmond,
Thanks for doing this. I also had a hunch that 1 ppm could be reached by HEC + NFB alone, but I have not yet done a sim of it, just some back of the envelope of my individual results of my VAS and EC OPS.
Can you share the schematic and details with us?
I assume that you did not use TMC, is that correct?
Was the THD-20 100W into 8 ohms with a single output pair biased at 150 mA? Or was it 200W into 4 ohms with three output pairs biased at a total of 450 mA? Or something else?
Can you show us the fourier spectrum?
Can you show us the 100 kHz square wave response?
I assume you used the same rail voltages throughout; what were the rail voltages and at what peak voltage did the amp clip?
Thanks,
Bob
Edmond Stuart said:
Hi Edmond,
Thanks for doing this. I also had a hunch that 1 ppm could be reached by HEC + NFB alone, but I have not yet done a sim of it, just some back of the envelope of my individual results of my VAS and EC OPS.
Can you share the schematic and details with us?
I assume that you did not use TMC, is that correct?
Was the THD-20 100W into 8 ohms with a single output pair biased at 150 mA? Or was it 200W into 4 ohms with three output pairs biased at a total of 450 mA? Or something else?
Can you show us the fourier spectrum?
Can you show us the 100 kHz square wave response?
I assume you used the same rail voltages throughout; what were the rail voltages and at what peak voltage did the amp clip?
Thanks,
Bob
jcx wrote:
Why didn't Hawksford make it clear in his 1981 paper that his scheme is equivalent to ordinary NFB? I mean, that is rather a glaring omission. Did he not realise at the time? Was it deliberate?
In the intro of his 1981 paper it is interesting that he talks about some of the problems that NFB loops pose as if to suggest that he has a non-NFB alternative. He then goes on to talk about error cancellation in reference to the combined FB/FF model, but drops this when talking about the FB only system. In his conclusion section he says "The technique should find favour among designers who adhere to the low-feedback school of design, as corrective feedback is only applied when distortion in the output stage is generated."
That last suggestion is ignorant of the way a control system works as a NFB system maintains control even when the input and output signals are equal. Whether Hawksford was mistaken or was trying to make his paper more tantalizing, I don't know. But I imagine the subsequent adoption of HEC circuits was primarily motivated by the erroneous expectation that it was somehow different from ordinary NFB and had properties like "cancellation" or "error only feedback". The opportunity to have a so-called "null adjustment" may have further fuelled the desire to believe there was something novel going on. And we all know how important novelty is in marketing audio. But these ideas do not stand up to rigorous inspection.
Admitedly, describing HEC as local OS NFB incorporating a boostrap to generate forward gain doesn't sound as sexy. Much more exciting to use terms like "error correction" and to claim it is "an error cancellation technique like feed-forward as opposed to an error-reduction technique like negative feedback" (Bob Cordell) and to re-inforce this by the misnomer of a "null adjustment".
Exciting but erroneous.
Like Edmond, when the tinsel is stripped away, I see no compelling functionality benefit of the HEC topology from either a performance or a convenience point of view. Marketing, certainly. I haven't noticed any compelling arguments for HEC topology that aren't tinsel-related. I'm all ears.
Edmond has no time for the tinsel either. He has quickly knocked up a NFB OS stage without a bootstrap/PFB loop which has equivalent, if not superior performance. It still has too may transistors for my liking but that's another story.
So I suppose I would attribute the aforementioned conceits as the basis for terms like "HEC hoax" that have arisen recently.
Brian
Well informed comments as usual, jcx.
Why didn't Hawksford make it clear in his 1981 paper that his scheme is equivalent to ordinary NFB? I mean, that is rather a glaring omission. Did he not realise at the time? Was it deliberate?
In the intro of his 1981 paper it is interesting that he talks about some of the problems that NFB loops pose as if to suggest that he has a non-NFB alternative. He then goes on to talk about error cancellation in reference to the combined FB/FF model, but drops this when talking about the FB only system. In his conclusion section he says "The technique should find favour among designers who adhere to the low-feedback school of design, as corrective feedback is only applied when distortion in the output stage is generated."
That last suggestion is ignorant of the way a control system works as a NFB system maintains control even when the input and output signals are equal. Whether Hawksford was mistaken or was trying to make his paper more tantalizing, I don't know. But I imagine the subsequent adoption of HEC circuits was primarily motivated by the erroneous expectation that it was somehow different from ordinary NFB and had properties like "cancellation" or "error only feedback". The opportunity to have a so-called "null adjustment" may have further fuelled the desire to believe there was something novel going on. And we all know how important novelty is in marketing audio. But these ideas do not stand up to rigorous inspection.
Admitedly, describing HEC as local OS NFB incorporating a boostrap to generate forward gain doesn't sound as sexy. Much more exciting to use terms like "error correction" and to claim it is "an error cancellation technique like feed-forward as opposed to an error-reduction technique like negative feedback" (Bob Cordell) and to re-inforce this by the misnomer of a "null adjustment".
Exciting but erroneous.
Like Edmond, when the tinsel is stripped away, I see no compelling functionality benefit of the HEC topology from either a performance or a convenience point of view. Marketing, certainly. I haven't noticed any compelling arguments for HEC topology that aren't tinsel-related. I'm all ears.
Edmond has no time for the tinsel either. He has quickly knocked up a NFB OS stage without a bootstrap/PFB loop which has equivalent, if not superior performance. It still has too may transistors for my liking but that's another story.
So I suppose I would attribute the aforementioned conceits as the basis for terms like "HEC hoax" that have arisen recently.
Brian
traderbam said:jcx wrote:
Well informed comments as usual, jcx.
Why didn't Hawksford make it clear in his 1981 paper that his scheme is equivalent to ordinary NFB? I mean, that is rather a glaring omission. Did he not realise at the time? Was it deliberate?
<snip>
That last suggestion is ignorant of the way a control system works as a NFB system maintains control even when the input and output signals are equal.
<snip>
Brian,
You could call a forum member "ignorant" and all you would risk is most likely a temporary ban granted by this forum moderators. But this time, you conveniently (lacking any direct opportunity to reply) call a distinguished professor, having all the credentials one could collect in a lifetime of university work, "ignorant" and suggest that he could commit a fraud in his work.
Such statements are not only disqualifying yourself, but are also staining the reputation of this forum, in which enthusiasts meet industry leaders and experts. I suppose that everybody you quoted and praised in your support is disgusted by this kind of low attitude.
I wouldn't accuse Hawksford of malfeasance, only over enthusiasm that should have been tempered by greater diligence in searching out prior works - missing the occasional obscure decades old paper is excusable, "text book" level control theory less so
I'm quite happy to have Bob's implementation of "error correction" available - I just cringe at the posts imputing magical potential to it when its equivalence to single loop error feedback lets us use Bode's negative feedback theory to find very well determined limits
I'm quite happy to have Bob's implementation of "error correction" available - I just cringe at the posts imputing magical potential to it when its equivalence to single loop error feedback lets us use Bode's negative feedback theory to find very well determined limits
WRT Brian's sketch in #2871, I see this uses same topology as a conventional DC servo. Except that the reference is not GND but the input, and the "DC" extends into the MHz. We could read "DC" as "Difference Component", though, and the integrating servo tries to null it... and to judge the stability of a DC servo I always use bode plots...
- Klaus
- Klaus
Ovidiu wrote:
You misquote me. I have not called any individual ignorant. I called the claim ignorant. I did say that Bob Cordell made a mistake in his 1984 paper.
I wrote:
Let's talk about the reputation of this forum. I do not support your suggestion that this forum should become politically correct. That the analysis of ideas and the criticism or praise of ideas should be in any way prejudiced by the "qualifications" or "reputation" of the author of those ideas.
You may be figuratively awestruck by the title of "professor", I'm not, but I reserve the right to expose fault if I find it. A title does not bestow perfection on the holder or impunity from criticism of their claims. Especially in the audio community.
I think the reputation of this forum is better served by it being a foum of equals and a place of uncensored critical debate.
You misquote me. I have not called any individual ignorant. I called the claim ignorant. I did say that Bob Cordell made a mistake in his 1984 paper.
I wrote:
Let's talk about the reputation of this forum. I do not support your suggestion that this forum should become politically correct. That the analysis of ideas and the criticism or praise of ideas should be in any way prejudiced by the "qualifications" or "reputation" of the author of those ideas.
You may be figuratively awestruck by the title of "professor", I'm not, but I reserve the right to expose fault if I find it. A title does not bestow perfection on the holder or impunity from criticism of their claims. Especially in the audio community.
I think the reputation of this forum is better served by it being a foum of equals and a place of uncensored critical debate.