The only issue is how fast it can come out of saturation. It won’t be so slow (even with a TIP41) that it would result in cross conduction, but will result in high order distortion. Higher order distortion products tend to be more audible than low order ones. It will make clipping more obvious, when it happens. It’s one reason I really like the EF3. The predriver can be made very small to come out of saturation quick, since it doesn’t have to handle currents upwards of an amp when the outputs start running out of beta.
R23 is the speed up mechanism for getting Q10 out of saturation quickly. Without the resistor it is slower. And the unusual connection (to the emitter of the driver, not the speaker out) is because of the intended fully class B operation. With factory bias, if the cap is connected to output the crossover distortion is intolerable. Depending on output transistor choice, class AB often results in instability. But this connection has other problems too. If the cap gets leaky the amp will snap to the rail. Not good when there is no protection relay!!!!!!! I automatically change this cap out with a new 100uf/100V Nichicon any time I’m inside one. When doing a complementary conversion with MJ15024/5, it gets re-routed to the output, and I bias at 10 mA or more per output device.
You will also notice that R18 is much smaller than R17 so that the bootstrap voltage is limited, not 50% of the rail as is common.
One doesn’t really need half the rail, but doing so results in the highest effective R value (both R’s equal), resulting in the lowest pole frequency for a given cap. With the high values in the PL you can sort of waste it a bit. In the days of EF1, using two 180 ohm was a big deal.
Found a very interesting thread here lots of useful info, I may have to try the protection diode mentioned.
