DIY headphone amp - checking my thoughts/calculations

[NickKUK]

The plate is 280ohm, with two tubes in push-pull 140ohm before feedback.

One option was/is to use a NPN/PNP to follow each output tube and possibly additional tubes - the net effect would be to lower the impedance further.

The simulations are coming along nicely. The current working version is a 6SN7 as a differential amp with a 6AS7 in push-pull and a DC servo on the 6AS7 grid bias. This is actually operating in Class A at the moment. With feedback I still think it would have enough power for average listening without adding an additional stage.

The second simulation I have a driver stage between the two - typically this can really only be used with feedback or causes overdrive distortion by soft clipping without the gain change using global negative feedback. I could make it 1:1 if feedback not used but then it would simply add noise.

This was just a second of simulation with NFB so the FFT hasn't had chance to really drop the noise floor - typically I'd run 5 minutes of simulation time without the first 30 seconds but it's very slow:
Screenshot 2020-12-02 at 11.21.53.png

[sgrossklass]

You need a better window function there... got Kaiser-Bessel? In LTspice I generally let transient sim run for 20 cycles or so.

Note that K240S' are notorious for being sensitive to output impedance... you really want <10 ohms. If you must go OTL by all means (silly and wasteful as it is... tubes just don't work well at the voltage/impedance levels you need for a headphone amp), I'd recommend near Bruno Putzeys levels of feedback.

[NickKUK]

Yup, no window applied.