Ok, thanks! I understand it now but there is another detail: The book says that R4 and R3 improve thermal stability since there is a voltage drop of few tenths across them that can be controlled with R2. But wouldn't it be better that there wasn't any quiescent current at all?
No sure what you have understood here.
R3 and R4 are emitter resistors that provide a small amount of local emitter feedback to each output transistor. This has nothing to do with with the push-pull arangement and is basic to the emitter follower (common collector) configuration that Q2 and Q3 are in.
If you do not understand this I suggest you go back and revise that part of your notes it is really basic.
In order to be just at the edge of conduction each of Q2 and Q3 need to be biased the equivalent of one forward pn junction - the base emitter junction. Hence the two diodes.
Using a diode instead of a resistor thermally stabilises the transistor since the diode junction reacts the same as the transistor junction to temperature change, unlike a resistor.
The purpose of R2 is twofold.
Firstly the voltage across it provides the extra bias to compensate for R3 and R4.
Secondly adding a small base bias on top of this puts Q3 and Q4 just slightly beyond the edge of conduction into actual conduction. This is done to eliminate the unwanted highly non-linear transfer characteristic at the origin called crossover distortion.
You may be confused with the use of the term bias.
R2 and the diodes provide
voltage bias to the bases of the two output transistors.
This results in a small
collector bias current in both transistors. It is this bias current that is meant when referring to setting the amplifier bias.
go well
