These are good points. BTW, I have always thought of class C only in the single-ended RF amplifier with tank sense.
Cheers,
Bob
I have to take the book out and find the part. 6th edition, P278, table 10.2. first line said for Re=0.1ohm, Iq should be 216mA!!! This is assume complementary EF pair.
In what world this is Class B?
This is close to the idle current I am using. I am designing with 5 stages in parallel, 200mA each and using Re=0.12ohm to satisfy Oliver's condition. I have idle current of 1A, I design this to get 8W of pure Class A output power for my 4ohm speaker. AGAIN, in what world this is Class B?
This is out right Class AB where you have conduction at idle, that you have a small region of Class A, then beyond that, the amp goes into Class B.
Come on guys. there is no grey area in this.
In what world this is Class B?
This is close to the idle current I am using. I am designing with 5 stages in parallel, 200mA each and using Re=0.12ohm to satisfy Oliver's condition. I have idle current of 1A, I design this to get 8W of pure Class A output power for my 4ohm speaker. AGAIN, in what world this is Class B?
This is out right Class AB where you have conduction at idle, that you have a small region of Class A, then beyond that, the amp goes into Class B.
Come on guys. there is no grey area in this.
Last edited:
I think it is useful to retain the term "AB" for the bias deliberately past the Oliver point. The level that D. Self dislikes so much
If you use that term for the usual, close to Oliver point, bias then you lose the distinction.
My modest proposal is that bias close to the Oliver level be called "B+'
That leads to the sequence as bias increases of B, B+, AB, A.
Simple, it preserves distinctions, doesn't conflict with established use of any of the other terms, and B+ makes sense as a term, both as "better than B" and intermediate between B and A.
Best wishes
David
Well, as far as I can tell there are two camps in audio, one which uses class-B more or less as I do, and another which refers to a push-pull with any bias >0 and <1/2 peak output current as class-AB. The former is in the minority especially as I am not aware of any mainstream textbooks that define the classes that way (they merely contradict themselves by stating that in class-B each device conducts over 50% of a cycle, but then draw a schematic with no bias at all meaning each device is actually conducting over less than 50% of a cycle).
There is the extra proviso in my definition that in order to be class-B, the output power level where both halves of the push-pull output stage remain conducting for the whole cycle should be under 3% of the rated output power. Under this definition, a B. Oliver optimally-biassed output stage can become class-AB if the emitter resistors have low value and/or there are a few paralleled outputs.
Don’t A, B and C all originate from the world of single-ended RF amplifiers? Does anyone know who first coined the terms and how they were defined?
I don’t see why class-C can’t be mapped to push-pull. A push-pull with no bias results in both halves of the output stage conducting over less than 50% of a signal cycle and therefore they operate in class-C.
Agreed, hence my attempt to define an unambiguous point at which a stage ceases to be class-B and becomes class-AB.
Admittedly the way my definitions are phrased highlights my preference for BJT output stages. With MOSFETs there is no clear “optimum bias point” and many MOSFET output stages are biased to class-AB in order to get good low-distortion performance.
Surely the term “AB” implies a mix of “A” and “B”? With a BJT output stage, you can bias for very low distortion and have a class-A region that is so small as to be negligible, and under this scenario using the term “AB” makes no sense to me whatsoever.
What is the title of Oliver’s paper? Is it referenced in your book? (don’t have it here to check).
If one push-pull pair is used, this gives a “class-A” region of only 746 mW (into 8 R) which seems fairly negligible to me. If the amplifier is rated at 50 W, doesn’t it seem a bit disingenuous to call it class-AB (implying a reasonable mix of class-A and class-B operation) when the “class-A” region is so small?
It isn’t. This is clearly class-AB.
Indeed.
Under this scenario, what is class-B?
Last edited:
The way his table using Re=0.1 implies it's a single stage. There is no limitation of how many stage the amp can be. If I have 5 stages, it's going to be 1.08A and it's over 18W of class A into 8ohm.
Are you trying to force the idea that one stage of 216mA idle is Class B. But 5 stages of 216mA in parallel is Class AB?
I disagree. You judge by the single stage. 216mA is way away the grey area of barely conducting by any stretch. this is forcing a square peg into a round hole.
How about if I run 300mA per stage? that is 1.44W even for a single stage, 36W per 5 stages. Still class B.
Then how about 1A of idle current by 100V rail.That is 18W of Class A into 8 ohm. But the full power is over 400W. 18W of class A is insignificant compare to over 400W. So by you definition, it's a Class B?
Last edited:
Yes. I don’t see what’s weird about that. One output stage has 5 times higher idle current than the other!
Perhaps you could address my question as to why use the term “AB” when the “A” part is so small as to be negligible? To me it goes against the intuitive interpretation of the term “AB”, and it is in fact this that is trying to fit a square peg into a round hole and not the other way round.
I edited the last post, If I have 5 stages driving 1A, but if my rail is 100V. Full power is over 400W. I do have 18W of Class A. But 18W is insignificant compare to 400W. So is that by your definition is Class B?
You seems to go over backwards to defend Self. The definition is very clean for Class A, B and C in the industry.
OK, I am clear now. We have the industry standard definition of Class B as 50% conduction. But it's too black and white for Self and a few other people. So, Self defined a new "Self Class B" that is defined as conduction from 50% to some TBD % that has "insignificant" amount of Class A to cover the grey area.
So next time when we talk, we just need to define first whether we talk in industry Class B or Self's Class B. Then there will be no debate.
So next time when we talk, we just need to define first whether we talk in industry Class B or Self's Class B. Then there will be no debate.
And how would you go about constructing this fabled “industry” class-B?
It’s me that’s suggested a % cutoff to define the boundary between class-B and class-AB (although I would of course be most unsurprised if someone else has also made the suggestion prior and I’m just unaware).
As I pointed out earlier, many Oliver optimally biassed output stages do actually conduct 50% (to 2 significant figures), so also meet the “industry standard” definition (but not the erroneous “corresponding” schematic shown in many textbooks with no bias). Going to, say, 51% means the class-A region will be very small and I don’t get the logic of classifying a 100 W amplifier with a 800 mW class-A region the same as a 100 W amplifier with an 8 W class-A region.
Last edited:
I too disliked the "new invention" of Self's ClassB, introduced 20 or so years ago, that was so different from the conventional ClassB.
As a result I refer to "Self's ClassB" on this Forum and sometimes tack on the condition equal to "optimally biased ClassAB"
To me optimally biased ClassAB cannot be equal to ClassB
To his credit, D.Self clearly defined what he meant by "Self's ClassB" and that left no confusion for me.
But I know from reading many posts on this Forum that D.Self's use of "Self's ClassB" does confuse many other Members.
It is being argued in the last few dozen posts that Classes C, B, AB and A are not clear cut and that some redefinition is required.
If this goes ahead, then it is only in DIYaudio's Forum that this redefinition will apply.
Then we would have DIYaudio's ClassC, DIYaudio's ClassB, DIYaudio's ClassAB and DIYaudio's other classes whether they are B+, or others. That redefintion would be wrong.
We cannot go out on a limb and leave the electronics world as conventional and DIYaudio as different. Our science is the same the world over.
We need to keep the same naming convention and clarify the operation using that convention.
As a result I refer to "Self's ClassB" on this Forum and sometimes tack on the condition equal to "optimally biased ClassAB"
To me optimally biased ClassAB cannot be equal to ClassB
To his credit, D.Self clearly defined what he meant by "Self's ClassB" and that left no confusion for me.
But I know from reading many posts on this Forum that D.Self's use of "Self's ClassB" does confuse many other Members.
It is being argued in the last few dozen posts that Classes C, B, AB and A are not clear cut and that some redefinition is required.
If this goes ahead, then it is only in DIYaudio's Forum that this redefinition will apply.
Then we would have DIYaudio's ClassC, DIYaudio's ClassB, DIYaudio's ClassAB and DIYaudio's other classes whether they are B+, or others. That redefintion would be wrong.
We cannot go out on a limb and leave the electronics world as conventional and DIYaudio as different. Our science is the same the world over.
We need to keep the same naming convention and clarify the operation using that convention.
Last edited:
-20dB ref maximum power is 1% of maximum power. That is a level that we often listen at.
What's wrong with having a ClassA region that is only a very few percent of maximum, when that would cover the average voltage output at which we listen?
I don’t think that Self invented this definition. Plenty of other authors (in papers published in the Journal of the Audio Engineering Society and other venues) use the definition that Self and I do.
Not redefinition. Merely better/more clearly defined. How can you say devices in a class-B stage conduct over 50% of a signal cycle, and then draw a circuit with no bias at all and say, “that’s class-B”?
Perhaps you could address my question as to why use the term “AB” when the “A” part is so small as to be negligible? To me it goes against the intuitive interpretation of the term “AB”.
Last edited:
In this nomenclature Class B refers to bias that is less than Oliver optimum. This is consistent with the way Bob Cordell uses the term.
This minimizes new definitions, keeps it as standard as possible.
Best wishes
David
I argue that 1% of ClassA is not negligible.
1W of ClassA comes about with just over 250mA of output bias.
1W as an average listening level is not insignificant.
1W is under the 3% of 50W suggested as a limit earlier in the Thread.
Consider the most ordinary op-amps.
The industry usually says their output stage as being in class B.
Does it mean that their output stage has no idle current ? No.
So, with class B meaning there is a small quiescent current, Douglas Self is not in contradiction with the industry.
What is wrong is the way most people explain class B (50% conduction and no more) without mentioning that :
- the conduction of each device of a push-pull stage cannot be strictly maintained at these 50% of a sine signal
- there is an inescapable need for a small idle current to never enter in class C, where a part of the signal is truncated.
The industry usually says their output stage as being in class B.
Does it mean that their output stage has no idle current ? No.
So, with class B meaning there is a small quiescent current, Douglas Self is not in contradiction with the industry.
What is wrong is the way most people explain class B (50% conduction and no more) without mentioning that :
- the conduction of each device of a push-pull stage cannot be strictly maintained at these 50% of a sine signal
- there is an inescapable need for a small idle current to never enter in class C, where a part of the signal is truncated.
I think Forr is right. If I remember my text book right, Class A = 360-degree conduction, Class B = 180<=conduction<360. For an amplifier output stage biased at certain level of standing current, it remains Class A until certain level of output power into certain load impedance. If that certain level of standing current allows the amp stay in Class A throughout its specified output power into specified load we call this amp a Class A amp. Otherwise a Class B amp, though one could specify a power level into a certain load such an amp departs Class A for Class B at. There is no need for Class AB, or Class AAB, e.t.c. as these are marketing terms designed to muddy thing up and speed up sales. Class C means base band information loss and is largely irrelevant to audio application if not irrelevant at all.
Hi Guys
As noted aboive, a typical BJT or mosfet output stage with zero bias will have a conduction discontinuity at the crossover point. At low frequencies and with lots of feedback around the stage, this can be overcome, but THD at higher frequencies will be poor. As a subwoofer amp it would be fine and you would have a true solid-state class-B amplifier with each device conducting for only 50% or the signal (180 degrees). Open loop this would sound awful. Idle power would be zero, so great for a "plate" amplifier
On the other hand, a screen-driven tube output stage can be idled at zero current and exhibit no more distortion than a usual-biased grid-driven stage. You can wire any of the common power tetrodes or pentodes this way but the stage requires power drive rather than voltage drive.
Note that colloquially and in official writings from device manufacturers, "class-B" is generally used to refer to an amp that does not idle its output devices at their full rating. RCA's tube manual suggest a 1/3 idle dissipation in their "Class-B Design Procedure" - pretty hot for some tubes. Similarly in solid-state it seems that for convenience most designers refer to their amps as "class-B" if the idle power is anything from stone cold to quite warm. When you have a big output stage with each BJT idling at say 8W, 16 devices means 128W of idle heat. The output power for such an amp might well be 500-1kW and most would consider it to be either class-B or class-AB. On the other hand, were the amp idling at 128W and its output rating 128W or less, it would be considered class-A by most and likely marketed that way.
Have fun
As noted aboive, a typical BJT or mosfet output stage with zero bias will have a conduction discontinuity at the crossover point. At low frequencies and with lots of feedback around the stage, this can be overcome, but THD at higher frequencies will be poor. As a subwoofer amp it would be fine and you would have a true solid-state class-B amplifier with each device conducting for only 50% or the signal (180 degrees). Open loop this would sound awful. Idle power would be zero, so great for a "plate" amplifier
On the other hand, a screen-driven tube output stage can be idled at zero current and exhibit no more distortion than a usual-biased grid-driven stage. You can wire any of the common power tetrodes or pentodes this way but the stage requires power drive rather than voltage drive.
Note that colloquially and in official writings from device manufacturers, "class-B" is generally used to refer to an amp that does not idle its output devices at their full rating. RCA's tube manual suggest a 1/3 idle dissipation in their "Class-B Design Procedure" - pretty hot for some tubes. Similarly in solid-state it seems that for convenience most designers refer to their amps as "class-B" if the idle power is anything from stone cold to quite warm. When you have a big output stage with each BJT idling at say 8W, 16 devices means 128W of idle heat. The output power for such an amp might well be 500-1kW and most would consider it to be either class-B or class-AB. On the other hand, were the amp idling at 128W and its output rating 128W or less, it would be considered class-A by most and likely marketed that way.
Have fun
Last edited: