Your definition is probably something like this :
Class A, B, D, G, H to T Audio Amplifiers
The class B configuration employs 2 transistors, each of which conducts for exactly half of the signal cycle. In the quiescent state, no current at all flows through the transistors.
An externally hosted image should be here but it was not working when we last tested it.
Underline is an "exact" configuration which, I think, is impossible to maintain.
Either, there is at least a very small idle current.
Or, if there is no idle current at all, a dead zone appears. The output then misses part of the input signal, the dead zone means that both halves are working in class C.
Self's classification clears up this old ambiguity about class B which puzzled me as soon as I was involved in electronics.
Last edited:
Yes, this is well defined and accepted by the rest of the world. What he is talking is just Class AB. Just changing it for the sake of being different like the Class XD?
Exactly. I agree with Self on this one. My approach to classification is as follows (from my PhD thesis):
In all cases, the output signal being considered is a sinusoid with no DC component.
- Class-A: Quiescent current is set to at least one half of the peak output current.
- Class-AB: Quiescent current is set to significantly more than zero but less than one half of the peak output current.
- Class-B: Quiescent current is set close to zero at a level that minimises full-power distortion.
- Class-C: Quiescent current is zero.
- Class-D: All previous classes have been dealing with output stages that accept a voltage signal input whose instantaneous value corresponds to the instantaneous value of the audio signal and the output stage switches a maximum of once per cycle. Class-D on the other hand requires the input signal to be a suitably modulated binary voltage, where the positions of the rising and falling edges of the voltage waveform represent the audio signal. The output stage devices then operate in a switching mode, being either off or fully on (saturated) and switching at a frequency much higher than the highest signal frequency. The analogue music signal can then be recovered using a low-pass filter. Class-D amplifiers are sometimes erroneously thought of and referred to as “digital” amplifiers; a good explanation of their true analogue nature can be found in [1].
The definitions here of classes A to C specify quiescent current levels rather than the more usually used conduction angles of 100 % for class-A, 50 % – 100 % for class-AB, 50 % for class-B, and less than 50 % for class-C. For output stages with static bias settings, the quiescent current and conduction angle definitions are equivalent. One benefit of the quiescent current definition is that it avoids classifying “non-switching” or “sliding bias” output stages [2] as class-A even though neither half of the output stage in such amplifiers ever fully turns off. Whilst the definition here of classes -AB and -B appear to be universally acknowledged, upon closer inspection they are not. Whilst class-B is often defined as each output device conducting for 50 % of a cycle, the corresponding push-pull schematic is often given with no bias. However, such a configuration is class-C, as each output device has no quiescent current and conducts for less than 50 % of a cycle; the stage will exhibit significant distortion. In order for each output device to conduct for exactly 50 % of a cycle, some bias voltage is necessary. However, not only would a precise 50 % for all signal and load conditions be very difficult to achieve, it would also result in worse distortion performance than an optimally biased stage. In the case of an amplifier rated at 100 W into 8R, an optimally-biased BJT emitter-follower output stage requires a bias level of around 45 mV [3] p154. With a peak output voltage of 40 volts, emitter resistors of 0R22, and a load of 8R, conduction over a cycle is given by:
((π + arcsin(0.0225/40 * 8/0.22 ))/(2π))*100
To two significant figures, this evaluates to 50 %, demonstrating that the presence of bias does not imply a significant violation of the classic definition of a class-B output stage. Stating that the presence of any quiescent current, regardless of its magnitude, implies class-AB operation lumps optimally-biased output stages with those that conduct significantly more than 50 % of the time and would relegate class-B to being a physically unrealisable anomaly. The definition class-AB is best reserved for those output stages that are attempting to deliver class-A performance up to a certain significant power level* and class-B performance above that power level.
* In the example amplifier for the above equation, class-A operation extends to around 670 mW into 8R, less than 1 % of the rated output power of 100 W. A suggested level of “significant power” is 3 % of rated output power.
[1] Putzeys, B.; Veltman, A.; van der Hulst, P.; Groenenberg, R.: All Amplifiers are Analogue, but Some Amplifiers are More Analogue Than Others, 120th Convention of the Audio Engineering Society, 2006.
[2] Tanaka, S.: New Biasing Circuit for Class B Operation, Journal of the Audio Engineering Society, Vol. 29 No. 3, pp148-152.
[3] Self, D.: Audio Power Amplifier Design Handbook, Fourth Edition, Newnes, 2006. ISBN 0 750 68072 5
It's class B, loaded with a current source (which varies with output voltage level).
Just a comment... no brilliant insights; I mentioned this before and it wasnt appreciated but here is a good example: Higher performance with less parts/complexity is what I always look for as a hall-mark of superior designs. But it takes a LOT of sophistication to achieve it.
-RNM
I want to clarify, When I mention Class B, I change topic from the Class XD, nothing to do with XD anymore.
I keep reading Self talking about using Class B. But if you look at the examples throughout the book, he even give idle current and some are quite high. that is no the definition of Class B be it as % of on in each cycle or definition of "very little" idle current. I don't remember what page, but I remember he talking about idle current over 100mA. I would not call it Class B by any stretch.
I keep reading Self talking about using Class B. But if you look at the examples throughout the book, he even give idle current and some are quite high. that is no the definition of Class B be it as % of on in each cycle or definition of "very little" idle current. I don't remember what page, but I remember he talking about idle current over 100mA. I would not call it Class B by any stretch.
Interesting, so if I understand correctly you define class B as what is usually known an optimally (in the B. Oliver sense) biased class AB. While I don't necessary disagree with such a definition, I believe it comes against the current common knowledge.
I don't think class C was ever considered in a push-pull configuration, but it comes from the tank output, zero biased, RF power amplifiers used for CW amplification. Therefore, it may very well be that "class C" doesn't make any sense in the context of push-pull audio amplifiers. This would save the common class AB definition, as any push pull stage that is biased >0.
Now, the next distinction is how the distortions are depending on this class AB >0 bias. It is obviously decreasing from zero bias, to the Oliver point, by lowering the dead zone. However, above the Oliver point, the distortions are increasing again, but mostly because a completely different issue, that is gm doubling around the crossover. Further increase the bias, and the distortions are decreasing again, as soon as the class A is reached.
Of course, the transitions between the modes described above are not sharp, but gradual.
It is necessary to remember that classification applies to one device only first :
class A : conduction during of the whole cycle.
class B : conduction during exactly half the cycle (impracticable).
class C : conduction during less than half the cycle.
And then, apply it to push-pull configurations.
As en example,
many people think of the current dumping Quad 405 amp as having its output power transistors being biased in class B. Considering one device only at a time, it is clear that it is in conducting state during less than half a cycle : it clearly works in class C.
class A : conduction during of the whole cycle.
class B : conduction during exactly half the cycle (impracticable).
class C : conduction during less than half the cycle.
And then, apply it to push-pull configurations.
As en example,
many people think of the current dumping Quad 405 amp as having its output power transistors being biased in class B. Considering one device only at a time, it is clear that it is in conducting state during less than half a cycle : it clearly works in class C.
Current dumping amplifiers have two output stages working in parallel, one in Class-C that provides the majority of the output current, and another in Class-A that provides error correction. I like to use Self's suggestion of using a "+" to denote a series combination and a "•" to denote a parallel combination. Hence, current dumping is Class-A•C
This definition of class B collides with Mr. Dymond's class B definition, isn't it?
Class C can be used in push-pull audio amplifiers but they need some "assistance". In fact, it's a bit the contrary, they work by "assisting" a small power stage which drives them to provide enough current when needed.
I think that, apart from the already quoted Quad 405, Class C has been used used in Phase Linear and some other high power amps (maybe Stasis).
They had an array of power devices which delivers current only when sufficiently driven by a small power push-pull. On this schematics, look at the low value of R38 and R39, the voltage across them bias the power transistors only on large signals :
Forr said:
Not really, because Harry Dymond speaks about push-pull configurations and I speak about conduction of one transistor alone to which class B applies... theoretically... in text books... not in reality.
Sure.
Quoting the Quad 405, I restricted the mention of class C to the push-pull of the power transistors.
I think Douglas Self has made an excellent work with the detailed classification he proposed.
I guess I'm totally confused, all I see is a quasi-comp output stage, including something that reminds of the Baxandall diode. The superdiode delivers around 4Vbe's, so why would the output stage zero biased?
P.S. The very place I believe you got the schematic from recommends a 25mA bias for the output stage devices: http://synthetizer-sche.chez-alice.fr/power%20ampli/phase%20linear/repair.htm
Last edited:
Hi Harry,
There is certainly a muddy region between class B and class AB, partly because real-world devices don't go from conduction to cut-off instantly.
Of course, the class B vs. class AB distinction dates back to the tube days, where those devices also do not cut off instantly as a function of bias voltage. There, you could look on a scope and see pretty clearly when the bias was reduced to the point where notch distortion appeared. Bias just barely above this point was class B. Moderately more bias than this was class AB. Indeed, even tube amplifiers have static crossover distortion due to transconductance droop, just as MOSFETs do.
With BJTs, where there is an optimum bias, things are a bit different. The optimum bias is fairly small in the overall scheme of things, so we are closer to the gray region between what one would call class B and class AB. However, we regularly talk on this forum and elsewhere about the size of the "class A" region of our output stages. This is why I lean toward the term class AB. It is also true that amplifiers with a significant number of output pairs have a larger "class A" region.
It is also true that in the tube days, class B was considered to be a lower-quality, more distorting bias condition for output stages, so it pains me a bit to paint optimally-biased BJT output stages with that brush.
I don't have his paper in front of me, but I wonder whether Barney Oliver referred to it as class B or class AB when he came up with the concept of the optimally biased push-pull solid state output stage long, long ago.
Cheers,
Bob