This project would feel perfectly at home
here, too.
An AD843 seems like a bit of an odd choice for a headphone amp. Sure it's fast and all, but rather on the noisy side of things at 19 nV/sqrt(Hz), not to mention spendy. (Both the AD8610/20 and the popular OPA(2)132/134 show much less voltage noise.) While the PimetaV2 is not exactly a low-noise design to begin with (resistors alone contribute about 12 nV/sqrt(Hz), minimum), this opamp's input noise would easily dominate here. At standard gain, expected output noise level is 18 µV within a 20 kHz b/w, or -95 dBV. Expect perceptible background noise with (mostly) IEMs in the 120+ dB/V category. Plus, you'll need adapters for two single opamps where one dual goes, which add parasitics and hopefully ship with decoupling caps.
I'm not sure why R1 = 4k32 was chosen. This places the point of minimum input impedance distortion at absolute minimum volume, which makes little sense and requires an exotic resistor value. 3k9 would have been just fine. (It makes more sense for the ground channel where impedances are, in fact, balanced, though obviously one could equally swap the 4k32s with 3k9s there.) I'd also prefer a lower-impedance pot than the standard 50k, which may add up to 12k5 of source impedance... something like 20k or 10k, like the
O2 does.
