(another) Discrete BJT buffer

[z22]

This unity gain buffer is based on the Schlotzaur configuration in Douglas Self's book (fig. 3.17d), for driving headphones a push-pull output stage is added.

I've built the circuit (ground plane, dead bug style) and I am happy that it sounds good , however I have a few issues:

1. When using fixed resistors for emitter degeneration (R3, R4), there is about 10 mV offset at the output. Is the R3 trimmer a proper method of adjusting DC offset?
2. The compensation cap (C2) was selected by trial and error in LTspice.

[sgrossklass]

1. It is common practice to have a small trimpot in the R3-R4-Q5 junction (wiper to Q5) for offset adjustment.
2. Now let's hope your transistors stick to their respective models

It should sound good, essentially being a beefy Class A discrete opamp. I'd be somewhat worried about thermal stability though, not sure whether 1.2 ohm Rs are enough to keep quiescent current drift in check.

[jcx]

high bias for the diff pair, not using super beta input Q means input base current can be an issue

you really should trim the diff pair degen for minimum 2nd harmonic distortion, not Vos - particularly if the Vos comes from the base current mismatch/unequal +,- input source, feedback DC resistance

R can be added in series with the unity gain feedback (bypass with || C) to match DC R in +,- input signal paths

the current mirror emitter nodes are often used for Vos correction with a high value pot varying the injected offset current

[z22]

Thanks for the comments, I'm new to analog design. I've made several tweaks since I've built it:

1. RC power supply filter to the input stage changed to 100 ohms/4700u
2. Increased the rail capacitance to 11300u, added a 2.2 ohm resistor before that
3. Changed the LTP load to a Wilson current mirror
4. Changed the configuration to inverting, balanced the inputs according to the AD tutorial (I skipped the parallel bypass cap)
5. Common mode chokes at the input and mains lead. RFI is a problem in our area, I can sometimes see demodulated FM from the output.

I think I'll keep the 1.2 ohm emitter resistors as the quiescent current seems to be stable in my build.

I noticed that the power supply quality has an effect on sound (thus the big capacitors), I'll probably try an op-amp based regulator then call it a day.