Hi, Mikeks,
From your link here http://www.stereophile.com/solidpoweramps/683/index6.html
Halcro
I'm interested in residual structures. In Fi 4 and Fig 5.
Compare with http://www.stereophile.com/solidpoweramps/306chord/index4.html
Chord fig5
Halcro has high order residuals, Chord relative low order ones.
Is this typical of EC amps?
From your link here http://www.stereophile.com/solidpoweramps/683/index6.html
Halcro
I'm interested in residual structures. In Fi 4 and Fig 5.
Compare with http://www.stereophile.com/solidpoweramps/306chord/index4.html
Chord fig5
Halcro has high order residuals, Chord relative low order ones.
Is this typical of EC amps?
Thanks Bob.
I certainly don't want to flog a dead horse with this theory thing. But I have found it best to get the theory rock-solid before dealing with the implementation. When I've done it the other way round I usually go down blind alleys or start believing in magic.
Knowing EC and NFB are the same thing on paper releases the mind from the shackles of established conventions. It begs the question of whether Hawksford is the best way to implement a NFB loop in an output stage.
I am starting out with a prejudice against the Hawksford topology because I believe it is overly complex: it requires an extra loop and an adjustment process. I am also prejudiced against positive feedback loops because I believe they amplify the errors of the gain device in the loop. On the other hand, I recognize the odds are that I am wrong in my prejudices because Hawksford is popular among industry savants.
I would like to ask your leave, Bob, to put your o/p stage under the spotlight. I'd like to propose that we (this thread) characterize its feedback system in detail and then explore alternatives. Even if the circuit can not be bettered, the journey should be educational.
Brian
I certainly don't want to flog a dead horse with this theory thing. But I have found it best to get the theory rock-solid before dealing with the implementation. When I've done it the other way round I usually go down blind alleys or start believing in magic.
Knowing EC and NFB are the same thing on paper releases the mind from the shackles of established conventions. It begs the question of whether Hawksford is the best way to implement a NFB loop in an output stage.
I am starting out with a prejudice against the Hawksford topology because I believe it is overly complex: it requires an extra loop and an adjustment process. I am also prejudiced against positive feedback loops because I believe they amplify the errors of the gain device in the loop. On the other hand, I recognize the odds are that I am wrong in my prejudices because Hawksford is popular among industry savants.
I would like to ask your leave, Bob, to put your o/p stage under the spotlight. I'd like to propose that we (this thread) characterize its feedback system in detail and then explore alternatives. Even if the circuit can not be bettered, the journey should be educational.
Brian
Hi Brian,
I like your idea. I'm sure I'd learn a lot. The same for others too.
Bob, would you mind conducting a class? You have my complete attention.
-Chris
Edit: Sounds like a new thread to keep the message clear. I am really very interested.
I like your idea. I'm sure I'd learn a lot. The same for others too.
Bob, would you mind conducting a class? You have my complete attention.
-Chris
Edit: Sounds like a new thread to keep the message clear. I am really very interested.
lumanauw said:
Hi Lumanauw
The aa.pdf is a modification in this amplifier here:
http://www.diyaudio.com/forums/showthread.php?postid=1061760#post1061760
It looks like Hawksford but the signal input is different from the original concept. It needs a complete voltage amplifier as driver, and the gain is near unity.
The modification works… the THD disappears as a miracle. The only thing I see on residual, is noise…
More food for thought:
Error Correction in Power Amplifiers
The engineering community is presently putting much effort into designing low distortion amplifiers with techniques more sophisticated than ordinary feedback. To guide this effort, we present a detailed exposition of feedforward error nulling techniques, and we analyze a recent commercial design, illustrating significant improvements. The paper thus consists of: 1) A brief history of feedforward error correction, why it was eclipsed by feedback, and why the time is now ripe to exploit its possibilities for total error nulling. 2) An analysis of Blacks' feedforward configuration and how it relates to more recent circuit concepts such as MacDonald's active error feedback, Sandman's error take-off, Walker's "current dumping," and Pass's "stasis" principle. 3) An illustration of the only commercially available feedforward circuit, the Quad "current dumping" amplifier. 4) Significant improvements of the latter scheme using practical amplifiers, and generalizations of the bridge system incorporated in this concept. 5) An incorporation of error correctioninto class D switching amplifiers with resulting relaxed design criteria.
Preprint Number: 1511 Convention: 63 (April 1979)
Authors: Vanderkooy, John; Lipshitz, Stanley P.
I believe this is the one Bob referred to:
Is Zero Distortion Possible with Feedback?
A number of recent papers have suggested that a combination of positive and negative feedback loops can result in an amplifier in which total distortion cancellation occurs, so that the amplifier in principle produces zero distortion. Such a property is known to be achievable through the use of error feedforward, but can feedback also achieve the same end? We analyze the configuration in question and show that the apparent distortion cancellation is the result of ignoring the requirements for the physical realizability of the gain blocks and summers used. When these are properly accounted for, we find that stable feedback can only produce finite distortion reduction, as expected, except possibly at discrete frequencies.
Preprint Number: 2170 Convention: 76 (September 1984)
Authors: Lipshitz, Stanley P.; Vanderkooy, John
Jan Didden
Error Correction in Power Amplifiers
The engineering community is presently putting much effort into designing low distortion amplifiers with techniques more sophisticated than ordinary feedback. To guide this effort, we present a detailed exposition of feedforward error nulling techniques, and we analyze a recent commercial design, illustrating significant improvements. The paper thus consists of: 1) A brief history of feedforward error correction, why it was eclipsed by feedback, and why the time is now ripe to exploit its possibilities for total error nulling. 2) An analysis of Blacks' feedforward configuration and how it relates to more recent circuit concepts such as MacDonald's active error feedback, Sandman's error take-off, Walker's "current dumping," and Pass's "stasis" principle. 3) An illustration of the only commercially available feedforward circuit, the Quad "current dumping" amplifier. 4) Significant improvements of the latter scheme using practical amplifiers, and generalizations of the bridge system incorporated in this concept. 5) An incorporation of error correctioninto class D switching amplifiers with resulting relaxed design criteria.
Preprint Number: 1511 Convention: 63 (April 1979)
Authors: Vanderkooy, John; Lipshitz, Stanley P.
I believe this is the one Bob referred to:
Is Zero Distortion Possible with Feedback?
A number of recent papers have suggested that a combination of positive and negative feedback loops can result in an amplifier in which total distortion cancellation occurs, so that the amplifier in principle produces zero distortion. Such a property is known to be achievable through the use of error feedforward, but can feedback also achieve the same end? We analyze the configuration in question and show that the apparent distortion cancellation is the result of ignoring the requirements for the physical realizability of the gain blocks and summers used. When these are properly accounted for, we find that stable feedback can only produce finite distortion reduction, as expected, except possibly at discrete frequencies.
Preprint Number: 2170 Convention: 76 (September 1984)
Authors: Lipshitz, Stanley P.; Vanderkooy, John
Jan Didden
Hawksford EC need an adjustment (need to turn a pot) to set the correct operation point.
At what frequency is this? If it is set for 1 frequency (1khz, for example), will it work the same sharpness for 10hz up to 200khz test?
At what frequency is this? If it is set for 1 frequency (1khz, for example), will it work the same sharpness for 10hz up to 200khz test?
traderbam said:Clever guys, Vanderkooy and Lipshitz.😉
maybe they read I. M. Horowitz "Synthesis of Feedback Systems", 1963
7.15 Systems with Combined Positive and Negative Feedback; Zero Senistivity Systems
{1st 2 sentences: }
“ The literature on feedback systems is replete with claims of extraordinary benefits available from systems with minor positive feedback loops. The fallacy in these claims is usually due to lack of consideration of sensitivity (or loop transmission) as a function of frequency.”
{latter: }
"Positive feedback around active elements is a legitimate tool of active network synthesis, and it should be regarded as such and no more. For example if a very small sensitivity at a specific frequency is desired, then positive feedback is a convenient means of realizing a pole of L(jw)" very close to the jw axis…
…However it has never been shown that the gain-bandwidth available from an active device can be increased by positive feedback"
I agree; cheers Julian.
anatech said:
I'm not sure y'all realize how much of a burden you're asking Bob to take on here. In a free-form discussion, people contribute as their time allows. Asking for more than that is selfish in my view.
Mikeks, you have mail.
jcx said:
The system made from optimal subsystems can't be optimal, it is an axiom.
If one transistor can't benefit from a positive feedback, the whole system can, for example gain bandwidth of amplifier+speaker can be increased by positive feedback by current applied.
janneman said:
...Again, this magical Bridge...
Jan, 2 non-linear functions continue each other, but they do not compensate errors of each other! (Especially, the second one that is exponential and depends on a frequency and complexity of the load!)
Strictly speaking, the first one is assumed to be linear (as may be linear class A driver), the second one is non-linear, but it does not equally increase own value and decrease driver's gain, anyway it means the same crossover distortions that happen on powers where they are significantly less audible.
"Don't look for a black cat in the dark room if it is not there!"
Nothing is wrfong in shifting distortions to less audible powers and frequencies, it is what I do in my amps, it is absolutely valid approach since our perception is subjective and keeping in mynd it's nature we can make better sound understanding what we are doing instead of appyying significant forces whered it is not needed.
If you measure Bob Cordell amp's distortions you will see that it has much-much less distortions than Walker's amp with that "Magical Bridge".
ingrast said:
The difficulty is that positive feedback in a system increases distortion even as it increases the gain required to reduce it: a case of proverbially losing on the swings that which is gained on the slides.
Wavebourn said:
It is exceeding unlikely that a single transistor can be made to generate more gain with positive feedback
mikeks said:
I was surprised to find this example of positive feedback in a commercial amplifier. I have not done any analysis to confirm the author's claims, however it is very interesting:
http://www.diyaudio.com/forums/showthread.php?postid=779911#post779911
This discussion has me thinking back to the regenerative receiver and Armstrong's invention:
http://www.columbia.edu/cu/alumni/Magazine/Spring2002/Armstrong.html
Pete B.
lumanauw said:
Chord's 5th, 7th and 15th order IM products are down only 101 dB. I would not say that their products are of only low-order. In fairness to Halcro, keep in mind that the low amount of low-order IM products does not mean that the remaining high-order products are high. One just has to keep it all in perspective.
I would not tend to generalize that EC circuits have higher-order THD and IM products than non-EC circuits.
Bob
traderbam said:
Sure, I'll probably learn as much as anybody else in the process!
I certainly won't be insulted if you find something better (but remember, the devil is in the details, so don't claim you have something better until you can demonstrate measured performance 🙂.
Best regards,
Bob