[Dr-Tom]

Will this work???

I use a Nelson Pass B1 pre-amp/buffer... and was thinking of making a head-amp for my new AKG701...

Can I make a push-pull amp by using two TPA6120 chips to generate a inverted+non-inverted signal for each channel from the single ended B1 buffer?

There should be no need for an input buffer (apart from the B1), meaning that the design will be quite simple even if the numerous recommendations from datasheet/posts are taken into considderation

The B1 has an output impedance around 1kOhm

Will this work better than an "active ground" channel?

/Dr-Tom

 

[jcx]

sounds like you're describing Bridged operation which is good for getting increased signal swing from limited Vsupply - some recent discussion shows some people express strong subjective preference for so-called "Balanced"(really Bridged output) amplification despite the theoretical advantages not obviously reaching audible thesholds

in the case of TPA6120 driving K701 its likely not necessary if you use +/-12 to 15 Vsupply for the TPA

another audiophile hack that output power capability of the TPA6120 allows is full-range output stage Class A operation

I've used the TPA in multiloop composite amps similar to Jung Multiloop topology

 

[Dr-Tom]

I think I'll just have to build and test it...
single ended B1 -> non-inverted
single ended B1 -> inverted
single ended B1 -> non-inverted+inverted

The concept of an active ground channel is tempting compared to the standard solution of just "dumping" the signal to ground

 

[jcx]

as an experienced instrumentation/measurement electronics designer I find the "fashion" for active ground here quite curious

dual supplies with a gnd plane on the pcb is likely to give better results even without carefully thinking about the current loop return paths and adding "star gnd" concepts by some judicious slotting of the gnd plane and taking care with power entry/routing

as for building and testing - I've measured differences/limitations in crosstalk and expect distortion measurements could show differences - but I really doubt the measurable differences in good implementations of either could be verified in controlled listening tests