I just finished reading the book about power amplifiers by Douglas Self. The author exhibits a strong "hatred" about all kinds of distortion and his suggestions go to lengths to minimize distortion in every aspect of the design.
What has me baffled is the strong statements promoting Class B amplifiers, compared to Class A and even Class AB.
The main argument against Class A is power dissipation and against Class AB is the increased distortion, compared with Class B. Although admitting there's crossover distortion in Class B, he explains that it can be eliminated by biasing techniques.
Now, for years I've always "known" that Class B = crossover distortion and that it sounds ugly. I've even seen some experiments on youtube showing hard spikes of all odd order harmonics down to the right edge of the spectrum display at 50kHz.
He gives some examples in his articles and book of Class B amplifiers - named blameless for not sporting any avoidable distortion - with measured THD in the level of 0.005 or less at 1kHz.
I would be grateful for some comments and insight on that matter.
What has me baffled is the strong statements promoting Class B amplifiers, compared to Class A and even Class AB.
The main argument against Class A is power dissipation and against Class AB is the increased distortion, compared with Class B. Although admitting there's crossover distortion in Class B, he explains that it can be eliminated by biasing techniques.
Now, for years I've always "known" that Class B = crossover distortion and that it sounds ugly. I've even seen some experiments on youtube showing hard spikes of all odd order harmonics down to the right edge of the spectrum display at 50kHz.
He gives some examples in his articles and book of Class B amplifiers - named blameless for not sporting any avoidable distortion - with measured THD in the level of 0.005 or less at 1kHz.
I would be grateful for some comments and insight on that matter.
Is it too late to get a refund for the book? 
It is going to depend very much on the specific device types and the devices and the surrounding circuitry. There are plenty of great sounding super-amps in all three classes: Naim, Krell, Pass. That is, if you define class B to be at least a few tens of milliamps of bias.
It is going to depend very much on the specific device types and the devices and the surrounding circuitry. There are plenty of great sounding super-amps in all three classes: Naim, Krell, Pass. That is, if you define class B to be at least a few tens of milliamps of bias.
Self defines class B as a push pull output stage biased at the optimum for least crossover distortion as well as gm doubling. Most others appear to refer to such as class AB. Normally this is the bias level that causes the intrinsic emitter resistance to be equal to the emitter resistor value with no signal; the so-called Oliver criterion referred to in the book. Hope this helps.
Sasi, are you looking for good THD figures or do you want a good sounding amp ? This is often not the same.
Self showed that a class AB amp with a bias around 30mA / pair gives the lowest THD. But my ears has a slightly different point of view.
A "Self"-biased amp has a tendency to sound a bit thin compared to an amp that has been biased to some hundred mA. A higher bias gives an impression of solidity, so to speak.
Causes? Who knows. Perhaps our ears are more sensitive to cross over dist when the level is at it's lowest amplitude. And the power rails delivers current more abruptly on a "Self" amp. Then we have the thermal issue. BJT's that turns on and off completely may have some thermal related distortion.
Self showed that a class AB amp with a bias around 30mA / pair gives the lowest THD. But my ears has a slightly different point of view.
A "Self"-biased amp has a tendency to sound a bit thin compared to an amp that has been biased to some hundred mA. A higher bias gives an impression of solidity, so to speak.
Causes? Who knows. Perhaps our ears are more sensitive to cross over dist when the level is at it's lowest amplitude. And the power rails delivers current more abruptly on a "Self" amp. Then we have the thermal issue. BJT's that turns on and off completely may have some thermal related distortion.
Thanks for the feedback.
So I suppose Self considers Class B amplifiers to be those with minimum bias current. It does make some sense, especially when the class A mode in a multi-class design is enabled with the switch in "Winter mode"
our ears - and the scope - are most sensitive to crossover distortion at lowest levels since - for a "pure class B" amplifier there's a +/- 0.6V region of nothingness after which they deliver signal. At 1V nominal output the effect of the zero-zone would be catastrophic. At 30V output it can be argued that the effect would be negligible.
You are right. There is something in the schematics of those Class B amplifiers (the biasing diodes, later to be replaced by a servo bias circuit) that didn't make sense.
So I suppose Self considers Class B amplifiers to be those with minimum bias current. It does make some sense, especially when the class A mode in a multi-class design is enabled with the switch in "Winter mode"
I think it's well understood that the on-off time of the junctions is not insignificant. And I am assuming it goes through a brief non-linear phase before the junction is in its linear part.
I have to go back to the relevant chapters and re-read them with the above in mind.
Actually it's a fascinating book. Thoroughly enjoying to read. The book actually states that Class A amplifiers are supposed to be distortion free and that they are superior in sound - if not wrongly designed. Self is biased towards commercial design and cost is always a strong factor in his choices and recommendations.
No. You have misunderstood what Class B is. Class B means no overlap in signal, not zero bias. Class B is a textbook ideal; no real competent amplifier is actually Class B in the strict sense for the simple reason that it would be impossible to maintain the exact bias point. However, some amps approach this and if done properly this is fine. Self uses 'Class B' in the most sensible form: amps where the overlap is carefully chosen to maintain constant gain through the zero crossing. Using higher bias than this, as most designers do, is Class AB which has gm-doubling and more distortion; however many people seem to prefer this.SaSi said:
hmmm....
Let me see if I understand what you are saying:
Class B - in theory - is a push-pull design where the two transistors operate exactly 50% of the time. This can't be done with unbiased output transistors, since they do have the 0.6V bias need each. So, a "pure class B" amplifier needs some bias to each transistor and that some bias needs to be exactly that 0.6V to each transistor so that they operate at exactly 50% of the time each.
A class AB output stage will have more than that minimum required bias, causing the output transistors to output for more than 50% of the time - how much depends on the level of "overbias" (if I may use that term), which is something that causes an effect opposite to the crossover distortion, having both transistors operate in the "shared" range. There are several graphs in the Self book showing the ill-effects of that.
Did I understand it correctly?
If so, what puzzles me is what some people say about class AB amplifiers, that they operate in class A for small output levels and as the output increases, they operate in class AB. Is that a correct statement?
Let me see if I understand what you are saying:
Class B - in theory - is a push-pull design where the two transistors operate exactly 50% of the time. This can't be done with unbiased output transistors, since they do have the 0.6V bias need each. So, a "pure class B" amplifier needs some bias to each transistor and that some bias needs to be exactly that 0.6V to each transistor so that they operate at exactly 50% of the time each.
A class AB output stage will have more than that minimum required bias, causing the output transistors to output for more than 50% of the time - how much depends on the level of "overbias" (if I may use that term), which is something that causes an effect opposite to the crossover distortion, having both transistors operate in the "shared" range. There are several graphs in the Self book showing the ill-effects of that.
Did I understand it correctly?
If so, what puzzles me is what some people say about class AB amplifiers, that they operate in class A for small output levels and as the output increases, they operate in class AB. Is that a correct statement?
Not 50% of the time, as the signal may be asymmetric. Class B means that the positive part of the signal is handled by one output device, and the negative part be another device. If you assume that an output device is switched off when biased at less than 0.6V and amplifies linearly at 0.6V and beyond then your understanding is correct. Real life is not like this, but this is what you will see described in elementary textbooks.
In real life, devices turn on gradually so there needs to be some overlap. People argue about whether this should be called Class B or Class AB, and whether the mode changes from A to B for higher signals. Given that class definitions come from the ideal world of textbooks it is not always helpful to attempt to strictly apply them to real circuits. The real issue is: what happens to the signal?
In olden days people did not have these arguments. They understood that a push-pull amplifier with little or no standing current was Class B, and one with significant standing current was Class AB.
In real life, devices turn on gradually so there needs to be some overlap. People argue about whether this should be called Class B or Class AB, and whether the mode changes from A to B for higher signals. Given that class definitions come from the ideal world of textbooks it is not always helpful to attempt to strictly apply them to real circuits. The real issue is: what happens to the signal?
In olden days people did not have these arguments. They understood that a push-pull amplifier with little or no standing current was Class B, and one with significant standing current was Class AB.
I do understand your comment about 50% and asymetric signal. I meant the same thing, having sine waves in mind...
Indeed you are correct about the output transistors needing to be biased so that they conduct within their linear range. And some argue that the minimum bias to switch on the transistor isn't 0.6V but rather 0.7V.
I will read the book again having these details in mind.
That sums it up nicely.
Indeed you are correct about the output transistors needing to be biased so that they conduct within their linear range. And some argue that the minimum bias to switch on the transistor isn't 0.6V but rather 0.7V.
I will read the book again having these details in mind.
Call it A or AB, it isn't that important. The thing is that the output devices must have some bias voltage to overcome that +- 0.6V region Sasi talked about. Then we can discuss how much bias we need.
One more thing. I f we bias the stage at 250mA the amplifier actually operates in class A up to around 4V peak ( 8 ohm load ). In a normal living room, a peek signal of 4V actually is quite loud. I would say the level is at a degree that many people would shout to you to turn down the volume. So this 250mA biased amp is actually a class A amp under normal circumstances.
Douglas Self has some reputation among constructors and has analyzed a lot of things, but one thing is strange, he never talked about the subjective impressions he got from various topologies. John Lindsey Hood, the old guy, did that from time to time.
But how can Self be so uninterested in his own subjective impressions? Strange.
One more thing. I f we bias the stage at 250mA the amplifier actually operates in class A up to around 4V peak ( 8 ohm load ). In a normal living room, a peek signal of 4V actually is quite loud. I would say the level is at a degree that many people would shout to you to turn down the volume. So this 250mA biased amp is actually a class A amp under normal circumstances.
Douglas Self has some reputation among constructors and has analyzed a lot of things, but one thing is strange, he never talked about the subjective impressions he got from various topologies. John Lindsey Hood, the old guy, did that from time to time.
But how can Self be so uninterested in his own subjective impressions? Strange.
He is one of those people that hates subjective opinions because in the audio world they have become sort of cult. Like one way conducting wires and acclaims of Tube amplifiers with lots of harmonic distortion sounding better than transistors.
His way of thinking is "if you can't measure it, it ain't there".
His approach to design is to be able to engineer to perfection, and perfection being the measurement threshold of your measuring equipment, provided they are SOTA.
Actually, it's not perfection... it's "good enough to be safely assumed audibly transparent" at a reasonable amount of effort. That threshold for distortion would be at about -80 to -100 dB depending on who you ask. So shoot for <0.001% (10 ppm) across the audio range and you shouldn't have to worry about it. Perfection would mean shooting for sub-1 ppm levels.
As mentioned, Self's class definitions do not always align with common usage, and he does have a point. What's commonly referred to as "Class B" (the bias-less follower) is something with a big dead zone in it - so shouldn't that rather be called class C? That in turn results in definition shifts for B and AB as well - the optimally biased amplifier, traditionally filed under AB, becomes B, and AB now is the overbiased amplifier.
Note that he generally looks at 8 ohm loads and only briefly acknowledges the 4 ohm world. In order to get the same conditions with a different load impedance Rx for a given output power, you have to scale up quiescent current and current handling by a factor of sqrt(8 ohm / Rx) and emitter resistors down by sqrt(Rx / 8ohm). IOW, double up your output, and your amp will be able to handle 2 ohms instead of 8. The tricky part tends to be that you wouldn't generally be running your input stage at a different current, so you have to dig up some more current gain somehow.
As mentioned, Self's class definitions do not always align with common usage, and he does have a point. What's commonly referred to as "Class B" (the bias-less follower) is something with a big dead zone in it - so shouldn't that rather be called class C? That in turn results in definition shifts for B and AB as well - the optimally biased amplifier, traditionally filed under AB, becomes B, and AB now is the overbiased amplifier.
Note that he generally looks at 8 ohm loads and only briefly acknowledges the 4 ohm world. In order to get the same conditions with a different load impedance Rx for a given output power, you have to scale up quiescent current and current handling by a factor of sqrt(8 ohm / Rx) and emitter resistors down by sqrt(Rx / 8ohm). IOW, double up your output, and your amp will be able to handle 2 ohms instead of 8. The tricky part tends to be that you wouldn't generally be running your input stage at a different current, so you have to dig up some more current gain somehow.
This is why the audio world is so fascinating. The battle between subjectivists and objectivists will never end.
The absolute subjectivist standpoint is hazardous since the placebo effect will trick you all the time. And the objectivist standpoint is equally absurd - like constructing amplifiers for THD-meters. But hifi and wine tasting has alwasys some element of placebo, but the degree varies.
Speakers: 20 placebo.
Amplifier: 50%
Cables: 80%
Golden fuses: 98%
The magic little clock( machinadynamica.com ) : 100%
But I think after all the subjective way is the only possible way. But we must be sceptical.
But I must be honest and declare that people like Self is the total antithesis of what I believe in.
The absolute subjectivist standpoint is hazardous since the placebo effect will trick you all the time. And the objectivist standpoint is equally absurd - like constructing amplifiers for THD-meters. But hifi and wine tasting has alwasys some element of placebo, but the degree varies.
Speakers: 20 placebo.
Amplifier: 50%
Cables: 80%
Golden fuses: 98%
The magic little clock( machinadynamica.com ) : 100%
But I think after all the subjective way is the only possible way. But we must be sceptical.
But I must be honest and declare that people like Self is the total antithesis of what I believe in.
that Strawman again - "objectivists" in my experience universally accept Human Listening Test results - certainly includes Self
just require that they be done with Psychoacoustic Controls - Blinding, Level Matching...
I'd guess more "objectivists" own, have read Psychoacoustics, Mastering/Mixing, Physical Acoustics textbooks/articles/websites than your average "subjectivist"
just require that they be done with Psychoacoustic Controls - Blinding, Level Matching...
I'd guess more "objectivists" own, have read Psychoacoustics, Mastering/Mixing, Physical Acoustics textbooks/articles/websites than your average "subjectivist"
Last edited:
Yes, the whole audiophile thing is silly. The better things we get, the more we wan't. Audio junkies, so to speak.
Objectivists are more happy, I believe - they know their "blameless" amps are as good as it gets. They can just relax and listen to music.
I admit I am silly. I don't really believe in blind tests. Why?
Hifi and music is about feelings. In a blind test situation you tend to be uncertain and a bit tense. So the only feeling I will get is uncertainty.
In contrary, if I eveluate an amp at home for a week or so, I will gradually build up a long term impression. Can I live with the stuff?
When testing audio stuff, it's a normal thing to be impressed initially, just because the sound image is fresh and new. So it takes an evening, a day or a week or half a year to decide wether you like the stuff or not.
What do you say about this?
If a "blameless" amp is so good that we cannot tell it apart from any other well designed amplifier, why bothering with constructing and building new ones? This forum is silly since everyone reinvents the wheel over and over. Why not put the soldering station in the drawer, go out and buy some well designed NAD amplifer?
Objectivists are more happy, I believe - they know their "blameless" amps are as good as it gets. They can just relax and listen to music.
I admit I am silly. I don't really believe in blind tests. Why?
Hifi and music is about feelings. In a blind test situation you tend to be uncertain and a bit tense. So the only feeling I will get is uncertainty.
In contrary, if I eveluate an amp at home for a week or so, I will gradually build up a long term impression. Can I live with the stuff?
When testing audio stuff, it's a normal thing to be impressed initially, just because the sound image is fresh and new. So it takes an evening, a day or a week or half a year to decide wether you like the stuff or not.
What do you say about this?
If a "blameless" amp is so good that we cannot tell it apart from any other well designed amplifier, why bothering with constructing and building new ones? This forum is silly since everyone reinvents the wheel over and over. Why not put the soldering station in the drawer, go out and buy some well designed NAD amplifer?
I've no problem with personal rituals that increase enjoyment
just the arguments that the rituals, debunked beliefs, strawman distortions of "objectivist's" rationals, practice, reasoning deserve equal consideration when evaluating audio electronic's performance, especially as captured by measurements and intellignetly related to the existing body of Psychoacoustic Science
I comment on amp design because of my interest, expertise from a EE career, I have also taken advantage of opportunities to learn about Psychoacoustics, Transducers, Room iteractions...
to make a real Listening difference, were I motivated enough, I would try to build speakers after learning enough to weigh, modify, reproduce say Earl Geddes designs
and indeed just buy decent consumer audio PA - despite knowing how to design build and measure sub ppm distortion Amps
just the arguments that the rituals, debunked beliefs, strawman distortions of "objectivist's" rationals, practice, reasoning deserve equal consideration when evaluating audio electronic's performance, especially as captured by measurements and intellignetly related to the existing body of Psychoacoustic Science
I comment on amp design because of my interest, expertise from a EE career, I have also taken advantage of opportunities to learn about Psychoacoustics, Transducers, Room iteractions...
to make a real Listening difference, were I motivated enough, I would try to build speakers after learning enough to weigh, modify, reproduce say Earl Geddes designs
and indeed just buy decent consumer audio PA - despite knowing how to design build and measure sub ppm distortion Amps
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.