Are you talking about re-entrant distortion (covered more extensively by Baxandall)? If you want to translate this to the tube world, Morgan Jones pointed out where the re-entrant distortion issue is important and where it isn't in his treatment in Valve Amplifiers 4th edition.
If you're talking about "feedback only works on sine waves," then my suggested experiments will demonstrate quite nicely that this is not the case. Not to mention that fighter jets aren't falling out of the sky (usually).
The experimental proof of the pudding is the inaudibility of stable high feedback amps in the signal chain when tested for sound alone. This is a solved problem. The twist for the sorts of amps you sell is that the frequency response variations encountered due to the high-ish source impedance can in some cases make up for frequency response variations elsewhere (e.g., in the speaker design). See, for example, the nice synergy between the Arpeggio speakers and the high-ish source impedance of the Bulwer-Lytton amplifier, over in the Articles section of this site.
If you're talking about "feedback only works on sine waves," then my suggested experiments will demonstrate quite nicely that this is not the case. Not to mention that fighter jets aren't falling out of the sky (usually).
The experimental proof of the pudding is the inaudibility of stable high feedback amps in the signal chain when tested for sound alone. This is a solved problem. The twist for the sorts of amps you sell is that the frequency response variations encountered due to the high-ish source impedance can in some cases make up for frequency response variations elsewhere (e.g., in the speaker design). See, for example, the nice synergy between the Arpeggio speakers and the high-ish source impedance of the Bulwer-Lytton amplifier, over in the Articles section of this site.
.... then perhaps I may modestly qualify, having started designing/building amplifiers since about 1952?
True - but then that is a rather unqualified statement! Not to step on toes (at least not intentionally), but most high quality amplifiers of note used negative feedback way since the Williamson of about 1947. Leak, Quad II, Radford, Dynaco (not too knowledgable about the American scene of yore), et al....? Negative feedback is not only the only but also the most powerful way to reduce amplifier distortion to below audible levels - but let's not open that can of worms here; rather OT.
Sorry, worthy Sir,
but again I cannot support that as a general statement; it depends quite heavily on the circuit design. Yes, I have also read most of Crowhurst's works and learnt from them (cannot recall that exact one - as you said, it is rather long ago ....
) So I am sure he was correct in whatever way he came to that conclusion (he usually was), but again - as said. What with the powerful modern spectrum analyser programmes available, it is easy to show that with proper design, e.g. over 30dB nfb is required before 're'-generation of distortion sets in or before distortion reduction does not more or less follow the feedback factor. And then only really down to about the 11th harmonic etc. etc. But as said, this is quite dependant on starting with a proper open loop performance. (Example drawn from one of my designs, graphs here somewhere but will make this post too long. And no, Baxandall was not wrong, but he used a very specific case.)There, I'm way off topic now (a Moderator fingering a yellow card?), and are certainly not looking to draw swords with you (I don't even own a sword). Only, since you opened the door ....
"The fed-back signal always arrives late ....."
If I may perhaps dare setting that into some perspective for newbies, since it is an often quoted view as SY said:
To one's in-born cause-and-effect mindset this is a powerful concept, and true - until again one goes from a qualitative to a quantative manner of thinking. I would say a moderately proper amplifier design can be expected to be down by no more than some 3 dB at 20 kHz open-loop, i.e. before negative feedback (nfb). Such an amplifier with say 20 dB of nfb could easily be flat to > 60 kHz. The phase shift could typically be some 5 deg. at 20 kHz. That would then mean in terms of 'lateness', that the fed-back signal arrived some 3% of a half-wavelength late. (Rounding heavily here to just convey the concept; a half-wavelength chosen because there feedback would have no effect, either negative or positive.) One might then say 97% of the fed-back signal would still be doing its distortion-cancelling work (calculating to 19,73 dB of effective nfb instead of 20, if one wants to split hairs) etc. - of that order. (OK, the first harmonic to be cancelled would be no. 2 at 40 kHz; said amplifier has a phase shift there of still some 6 deg. only.) But we are now way out of the audible region.
I hope this puts the 'lateness' fallacy in some perspective. (Figures used are from the Williamson amplifier characteristics, approximated.)
If I may perhaps dare setting that into some perspective for newbies, since it is an often quoted view as SY said:
To one's in-born cause-and-effect mindset this is a powerful concept, and true - until again one goes from a qualitative to a quantative manner of thinking. I would say a moderately proper amplifier design can be expected to be down by no more than some 3 dB at 20 kHz open-loop, i.e. before negative feedback (nfb). Such an amplifier with say 20 dB of nfb could easily be flat to > 60 kHz. The phase shift could typically be some 5 deg. at 20 kHz. That would then mean in terms of 'lateness', that the fed-back signal arrived some 3% of a half-wavelength late. (Rounding heavily here to just convey the concept; a half-wavelength chosen because there feedback would have no effect, either negative or positive.) One might then say 97% of the fed-back signal would still be doing its distortion-cancelling work (calculating to 19,73 dB of effective nfb instead of 20, if one wants to split hairs) etc. - of that order. (OK, the first harmonic to be cancelled would be no. 2 at 40 kHz; said amplifier has a phase shift there of still some 6 deg. only.) But we are now way out of the audible region.
I hope this puts the 'lateness' fallacy in some perspective. (Figures used are from the Williamson amplifier characteristics, approximated.)
Johan, you might want to look at Baxandall's analysis. He used as a model a square-law FET with relatively high distortion (purely second). In that case, with moderate feedback levels, the re-entrant distortion of higher order harmonics rears its head. The limitation of this simple model is that it only is strictly true for relatively high open loop distortion and pure square law devices (neither of which obtains here). MJ covers this quite nicely in VA4. There's a lot more detail as well in the Self/Baxandall articles and letters reprinted in "Baxandall and Self on Audio power."
The inaccuracy of the first-order model is one reason for the extremely low and benign distortion signatures of cathode followers, which essentially have 100% feedback.
edit: "inapplicability" is probably a better word than "inaccuracy"
The inaccuracy of the first-order model is one reason for the extremely low and benign distortion signatures of cathode followers, which essentially have 100% feedback.
edit: "inapplicability" is probably a better word than "inaccuracy"
SY,
Thanks! Yes, I was referring to that model of Baxandall. Also thanks for reminding me about the Self-Baxandall articles; did read those as well but might very well go back there once more (not to self-rub it in, but age, memory and such ....)
Further, I probably should buy a copy of MJ-4 as the book is called (just some pecuniary embarrassment at present, what with our devalued currency ... but OOPS! that be politics
)
Regards
Thanks! Yes, I was referring to that model of Baxandall. Also thanks for reminding me about the Self-Baxandall articles; did read those as well but might very well go back there once more (not to self-rub it in, but age, memory and such ....)
Further, I probably should buy a copy of MJ-4 as the book is called (just some pecuniary embarrassment at present, what with our devalued currency ... but OOPS! that be politics
)Regards
What puzzles me about the "power paradigm" is that surely almost all loudspeakers these days are designed to be driven according to the "voltage paradigm." If such a speaker is instead driven according to the power paradigm, it will surely produce a totally skewed sound pressure level as a function of frequency?
Does that mean that amplifiers operating according to the "power paradigm" are suitable only for driving those exceptional loudspeakers that are designed with such a paradigm in mind?
Chris
Why would that be? It sounds as if you must be imagining that the feedback at the present instant is going to be making use of the signal from the previous cycle of the sinewave, or something like that. This is completely wrong. The timescale of any delay in the feedback loop is tiny compared to the period of any sinewave that the amplifier will be called upon to handle.
The feedback is using the signal essentially at the present instant. It doesn't care what the signal was in the past, one period of the sinewave oscillation ago. Nor does it care what the signal will be one period into the future. So whether the signal being processed is repetitive or not is completely immaterial.
Chris
Yes, the article is here. So basically a select group of hi-efficiency full range drivers used with transconductance amplifiers to reach hi-fi nirvana.
Or more realistically, one more set of design options. In reality, there's a range of options, with current source and voltage source just being the endpoints of a spectrum.
Feedback causes additional (re-entrant) harmonics because of the non-linear transfer characteristic of the main amplifier path. Whether even or odd, depends on the non-linearities. Crowhurst understood that, Cordell, Self etc understand that, Peter Baxandall understood it. Won't you join their ranks?
It has, of course, nothing whatsoever to do with delays.
For further info, see my sig line
Jan
PS A cathode resistor also causes neg. feedback of course. Gobs of it, actually.
Last edited:
By the way, for those who rather do some rational thinking instead of 'believe', this might help:
http://www.diyaudio.com/forums/blogs/jan-didden/454-feedback-how-late-time-same-time-all-time.html
Jan
http://www.diyaudio.com/forums/blogs/jan-didden/454-feedback-how-late-time-same-time-all-time.html
Jan
Check out a Stereophile amplifier review on their web site. They show the amplifier's frequency response into a dummy reactive load.
The amps with higher output impedance will have deviations of several dB from flat at the speaker terminals.
I really have the feeling that y'all are missing my point entirely!
First, to correct a misconception, we are not talking about 'current drive' as Nelson Pass has talked about. No paradgm ever developed around that concept.
Note that I use the word 'Paradigm', as I explained in the article for those of you who apparently have not read it and are simply reacting (here is the link again: Atma-Sphere music systems, inc.) for a reason.
The reason is that if one operates in a paradigm, all thinking outside that paradigm will appear to be blasphemy, and incorrect.
Here are some examples of the Power Paradigm: The 600 ohm balanced line standard. The efficiency measurement in loudspeakers (as opposed to sensitivity). Nearly all headphones...
I don't argue that feedback works to reduce distortion, increase bandwidth, etc. etc. and those such as SY responding to such are not looking at what I am saying, likely because they did not read the article.
While propagation delay does indeed play a role, and we can discuss that if you really think it important, it is beside the point.
What the point is is that in adding feedback, it causes the amplifier to violate a fundamental rule of human hearing/perception. This rule is about how the human ear/brain system perceives sound pressure.
In a nutshell the ear uses higher ordered harmonics to detirmine sound pressure. Now if y'all read Norman Crowhurst, you know that adding feedback to an amplifier circuit can cause the creation of harmonics up to the 81st(!) as well as intermodulations at the feedback node. In fact this harmonic texture becomes the noise floor of the amplifier.
(Now it happens that the amps I work with have no feedback and do have bandwidth well past 300KHz; by adding feedback we have extended the bandwidth well past 3 MHz without stability problems; IOW I am very familiar with wide bandwidth issues.)
It is these harmonics that Crowhurst mentions that are at the heart of the problem, and why the Power Paradigm is still around after all these years. If you read the article then you already know of speakers that are examples of Power Paradigm technology. I will reiterate them in a moment.
The presence of trace amounts of odd ordered or higher ordered harmonics is translated by the ear/brain system into brightness and harshness. Right at this point, while not the only factor by any means, we have one of the principle issues of why the endless debate of tubes/transistors continues ad nauseum.
Admittedly these are trace amounts of distortion! However the ear/brain system is more sensitive to these distortions than almost anything else, as it pertains directly to one of the most fundamental rules of human hearing: how we perceive sound pressure and so is more sensitive than most distortion analyzers. This is the point where one has to decide as a designer what is 'negligible'.
I've been at this a while and so far I've not found a way around this problem (except to leave out the feedback), that is if one's goal is to get the resulting reproduction to sound like real music instead of a good HiFi. This results in a higher 'output impedance' (which incidentally is a different term in the Power Paradigm than what is used in the Voltage Paradigm).
Now for those of you that didn't/won't read the article, here are some loudspeakers that operate on the Power Paradigm: most horns, most full-range single driver loudspeakers, most ESLs (as their impedance curve is not based on a driver in a box) and most magnetic planars (for the same reason). In addition, likely one of the most famous loudspeakers, the AR-1 (world's first acoustic suspension loudspeaker, which is why it is so famous) was based on Power Paradigm concepts- it was designed for an amplifier with a 7-ohm output impedance.
There are plenty of box speakers based on the Power Paradigm as well. Merlins, Coincident Technology (of Canada), Audiokinesis, Classic Audio Loudspeakers (a horn hybrid), Zu Audio (though technically a full-range single driver speaker supplemented with a tweeter) and quite a few more.
It was MacIntosh and ElectroVoice that led the charge in the 1950s to develop a set of rules wherein the voltage response of the amplifier and its interaction with the loudspeaker would be predictable, leading to (assumed) flat frequency response, which was one of the few human hearing rules we knew back then; many of the human hearing/perceptual rules were yet unknown. By the late 1970s, the resulting Voltage Paradigm had pretty well taken over. Nowadays we know that the brain has tipping points which can cause the processing of music to move from the limbic to cerebral cortex; we knew nothing of that 50 years ago! The Voltage Paradigm is really based on the idea of flat frequency response and low distortion, and ignores most research into human physiology that gone on since the 1970s, mostly because it never got updated.
But in the 1990s we saw a resurgence of SETs and horns (both examples of Power Paradigm technology). This occurred because designers were looking for a way to get away from the mechanical sounding consequences of using negative feedback (brightness, and the way it can cause tubes to sound bright and harsh like most solid state).
Again this is not about tubes/transistors- I am simply being pragmatic. I don't need a proof of this BTW, one does not have to know anything technical to know that tubes do something right- otherwise they would be long gone, historical relics of a bygone age. They survive 50 years after being declared obsolete though because it is easier to design an amplifier with them that sounds like music.
Now there are solid state designers, notably Nelson Pass (who is notable enough that there is an entire section of this very website devoted to him and his efforts) who has also found that negative feedback has problems. And guess what? He makes some of the best sounding solid state amps made.
Here is a link about distortion from his website- give it a read:
https://www.passdiy.com/project/articles/audio-distortion-and-feedback
First, to correct a misconception, we are not talking about 'current drive' as Nelson Pass has talked about. No paradgm ever developed around that concept.
Note that I use the word 'Paradigm', as I explained in the article for those of you who apparently have not read it and are simply reacting (here is the link again: Atma-Sphere music systems, inc.) for a reason.
The reason is that if one operates in a paradigm, all thinking outside that paradigm will appear to be blasphemy, and incorrect.
Here are some examples of the Power Paradigm: The 600 ohm balanced line standard. The efficiency measurement in loudspeakers (as opposed to sensitivity). Nearly all headphones...
I don't argue that feedback works to reduce distortion, increase bandwidth, etc. etc. and those such as SY responding to such are not looking at what I am saying, likely because they did not read the article.
While propagation delay does indeed play a role, and we can discuss that if you really think it important, it is beside the point.
What the point is is that in adding feedback, it causes the amplifier to violate a fundamental rule of human hearing/perception. This rule is about how the human ear/brain system perceives sound pressure.
In a nutshell the ear uses higher ordered harmonics to detirmine sound pressure. Now if y'all read Norman Crowhurst, you know that adding feedback to an amplifier circuit can cause the creation of harmonics up to the 81st(!) as well as intermodulations at the feedback node. In fact this harmonic texture becomes the noise floor of the amplifier.
(Now it happens that the amps I work with have no feedback and do have bandwidth well past 300KHz; by adding feedback we have extended the bandwidth well past 3 MHz without stability problems; IOW I am very familiar with wide bandwidth issues.)
It is these harmonics that Crowhurst mentions that are at the heart of the problem, and why the Power Paradigm is still around after all these years. If you read the article then you already know of speakers that are examples of Power Paradigm technology. I will reiterate them in a moment.
The presence of trace amounts of odd ordered or higher ordered harmonics is translated by the ear/brain system into brightness and harshness. Right at this point, while not the only factor by any means, we have one of the principle issues of why the endless debate of tubes/transistors continues ad nauseum.
Admittedly these are trace amounts of distortion! However the ear/brain system is more sensitive to these distortions than almost anything else, as it pertains directly to one of the most fundamental rules of human hearing: how we perceive sound pressure and so is more sensitive than most distortion analyzers. This is the point where one has to decide as a designer what is 'negligible'.
I've been at this a while and so far I've not found a way around this problem (except to leave out the feedback), that is if one's goal is to get the resulting reproduction to sound like real music instead of a good HiFi. This results in a higher 'output impedance' (which incidentally is a different term in the Power Paradigm than what is used in the Voltage Paradigm).
Now for those of you that didn't/won't read the article, here are some loudspeakers that operate on the Power Paradigm: most horns, most full-range single driver loudspeakers, most ESLs (as their impedance curve is not based on a driver in a box) and most magnetic planars (for the same reason). In addition, likely one of the most famous loudspeakers, the AR-1 (world's first acoustic suspension loudspeaker, which is why it is so famous) was based on Power Paradigm concepts- it was designed for an amplifier with a 7-ohm output impedance.
There are plenty of box speakers based on the Power Paradigm as well. Merlins, Coincident Technology (of Canada), Audiokinesis, Classic Audio Loudspeakers (a horn hybrid), Zu Audio (though technically a full-range single driver speaker supplemented with a tweeter) and quite a few more.
It was MacIntosh and ElectroVoice that led the charge in the 1950s to develop a set of rules wherein the voltage response of the amplifier and its interaction with the loudspeaker would be predictable, leading to (assumed) flat frequency response, which was one of the few human hearing rules we knew back then; many of the human hearing/perceptual rules were yet unknown. By the late 1970s, the resulting Voltage Paradigm had pretty well taken over. Nowadays we know that the brain has tipping points which can cause the processing of music to move from the limbic to cerebral cortex; we knew nothing of that 50 years ago! The Voltage Paradigm is really based on the idea of flat frequency response and low distortion, and ignores most research into human physiology that gone on since the 1970s, mostly because it never got updated.
But in the 1990s we saw a resurgence of SETs and horns (both examples of Power Paradigm technology). This occurred because designers were looking for a way to get away from the mechanical sounding consequences of using negative feedback (brightness, and the way it can cause tubes to sound bright and harsh like most solid state).
Again this is not about tubes/transistors- I am simply being pragmatic. I don't need a proof of this BTW, one does not have to know anything technical to know that tubes do something right- otherwise they would be long gone, historical relics of a bygone age. They survive 50 years after being declared obsolete though because it is easier to design an amplifier with them that sounds like music.
Now there are solid state designers, notably Nelson Pass (who is notable enough that there is an entire section of this very website devoted to him and his efforts) who has also found that negative feedback has problems. And guess what? He makes some of the best sounding solid state amps made.
Here is a link about distortion from his website- give it a read:
https://www.passdiy.com/project/articles/audio-distortion-and-feedback
Quite the wall of words. It reduces, though, to "re-entrant distortion," which has been treated more rigorously since Crowhurst's days, and "source impedance," ditto.
Just so as not to be bogged down in "tube vs transistors," here's measurements on a low/zero feedback solid state amp that you seem to view with favor (see figs. 8 and 9) :
Pass Labs XA60.5 monoblock power amplifier Measurements | Stereophile.com
And a high feedback solid state amp (see figs. 6 and 7):
Halcro dm58 monoblock power amplifier Measurements part 2 | Stereophile.com
Much more high order products in the former. The re-entrant distortion approximation is simply invalid when the open loop distortion is relatively low and the transfer characteristic isn't purely quadratic. This is stuff that was figured out after Crowhurst's initial papers (I think he wasn't the first to develop this idea in any event). If you have experimental data showing the contrary, I'd encourage you to post it.
The "power paradigm" requires speakers engineered for it, which hardly any are. If you look at the impedance curve of a real-world commercial ESL, you'll see that there will be a large frequency response deviation with "power paradigm." One can design an ESL for that sort of drive, but I don't know of anyone who has actually done so.
Just so as not to be bogged down in "tube vs transistors," here's measurements on a low/zero feedback solid state amp that you seem to view with favor (see figs. 8 and 9) :
Pass Labs XA60.5 monoblock power amplifier Measurements | Stereophile.com
And a high feedback solid state amp (see figs. 6 and 7):
Halcro dm58 monoblock power amplifier Measurements part 2 | Stereophile.com
Much more high order products in the former. The re-entrant distortion approximation is simply invalid when the open loop distortion is relatively low and the transfer characteristic isn't purely quadratic. This is stuff that was figured out after Crowhurst's initial papers (I think he wasn't the first to develop this idea in any event). If you have experimental data showing the contrary, I'd encourage you to post it.
The "power paradigm" requires speakers engineered for it, which hardly any are. If you look at the impedance curve of a real-world commercial ESL, you'll see that there will be a large frequency response deviation with "power paradigm." One can design an ESL for that sort of drive, but I don't know of anyone who has actually done so.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.