TPA6120A2 with balanced crossfeed?

[vcelkamaja]

It seems TPA6120A headphone amp is ideally driven by balanced input such as opamp buffered balanced DAC output. How can be then crossfeed used (I mean passive corssfeed between opamp and TPA6120A2)? All crossfeed designs I have seen (such as Meier) are always unbalanced and mixing unbalanced R to L channel with passive filter. Does anybody ever designed balanced crossfeed circuit?

[peranders]

Notice that the TPA6120a is a current feedback amp which makes it a bit special compared to ordinary voltage feedback. An another thing is the high input bias current of 12 uA. This can create a huge output offset.

[vcelkamaja]

There will be a opamp "buffer"similar as shown in TPA6120A2 datasheet (with PCM1792 DAC). The corssfeed passive circuit should be in front of this opamp buffer so I think it has no negative impact on TPA6120A2 itself.
What is not clear to me is how to design balanced crossfeed to both R+ / R- rails (and L+ / L-).

[qusp]

one thing is for sure, in order to not ruin CMRR and distortion performance you would have some pretty serious matching of passive components to do for the filter network

[jcx]

much balanced interface equipment does diff-to-single ended conversion internally for processing, then back to balanced for output if required

pure balanced path doesn't seem to have any technical justification - just "eye candy", appeals to the most "inconvenient, most expensive implementation must be better" strain of Audiophiledom

my view of balanced output DACs is that they are forced by digital noise, substrate coupling to use balanced output to get decent numbers – a problem with MHz clocked 100k+ digital transistors on the same die

in fact it is worth being suspicious of balanced outputs – did they provide sufficient RF common mode filtering internally or are they just exporting the clock feedthru, shaped noise to the next piece in the chain?

[chris719]

Originally Posted by jcx

my view of balanced output DACs is that they are forced by digital noise, substrate coupling to use balanced output to get decent numbers – a problem with MHz clocked 100k+ digital transistors on the same die

in fact it is worth being suspicious of balanced outputs – did they provide sufficient RF common mode filtering internally or are they just exporting the clock feedthru, shaped noise to the next piece in the chain?

What do you think of the effects on CMRR caused by separate op-amp I/V before diff-to-SE conversion? Does this make a case for using a fully differential op-amp like OPA1632 for I/V conversion?