That's a bit of a strange way of looking at it, but it might just be terminology. The point of class AB is that the output operates in class A at small power levels to preserve small signal linearity, then switches to class B when the signal is large enough to switch on the complimentary transistors that are biased off in the quiescent condition. I think that qualifies as class B operation, and helps to keep the dissipation low.
That is incorrect, strictly speaking may be... I posted the graph, you can see clearly the difference between class B and class AB. I do see your point about the "signal is large enough to switch on....." but it is a moot point because a) no matter how large the signal is, it always cross 0 right, and no matter how small the signal is, including 0 V, in class AB, there is always bias current that is high enough to make sure the transistor will conduct even when the signal just starts to cross 0.
Having said that, I would agree with you that
for large signal excursions, only one transistor is active for each half of the waveform, acting like a class B amplifier but that's the confusing part by nature, because it is about the "excursions" and I supposed that's what you referred to, but that does not really qualifies it as class B operation because for class B, the output transistors only conduct during half (180 degrees) of the signal waveform, period! Whereas in class AB, the bias will be high enough to ensure there is always an overlapping conduction between both output transistors in the push pull configuration, that is the conduction angle is always >180 degrees, not just = 180 degrees.
There is a difference between just 180 degrees (class B), and
between 180 and 360 degrees).
I have seen forum posters (even manufacturers (one that I remember seeing) interpreting class AB the way you do, but it is a misconception, or their own weird way of defining class AB and class B, resulting is creating a term that is misleading, one that we should not help to spread further (unintentionally of course
).
I think what Wirrunna was asking is when does the output stage stop working in class A? Very few audio manufacturers advertise that figure, and those that do make a play of it when their class A operation is relatively high.
One of the tests that Amir did a little while ago was to measure the difference between class A and AB with an amp that offered the choice. I'll see if I can find it. As I recall there was absolutely no measurable advantage with class A. In addition, if you look at the distortion vs power plots for the LA90, there's no great jump in distortion at low power levels, as with most good amps. It suggests that crossover distortion isn't something to worry about.
That varies with different designs, it could range from virtually never operate in class A, to a large % of rated output. For example, my A21 supposedly would run up to 8 W, 8 ohms in class A, whereas my 4B SST would do about 20-25 W in class A, both were told by the manufacturers. I remember seeing some Emotiva and Passlab class AB amps that claimed to run in class A up to much higher level (in % of rated) than the two I mentioned. So again, it is up to the amp designers, no fix rule.
The rest of it, I generally agree with you.