5W into 8 Ohms per channel.
Simulated: <-120dB (<0.0001%) just below full scale (5W), 1kHz
Measured: <-90dB (0.003%) just below full scale, 1kHz - but at this point, my measurement equipment becomes the limiting factor
Only uses discrete, cheap and easy-to-source components.
Signal transistors used: 2n3904, 2n3906 (~0.05€ each)
Power Transistors used: BD138, BD139 (~0.30€ each)
Diodes: 1n4148 (~0.02€ each).
I might replace them later, because i've found that my batch of diodes has a quite high forward voltage compared to the vbe of my transistors (0.71V vs 0.65V @I_f=7mA). Ideally, it would be the same or a little lower. The circuit can tolerate this +60mV difference, but it is just at the very edge and leads to lower thermal stability; going below 0mV would be much less problematic and leads to increased stability.