Sorry if this extends the debate on se or balanced but for those with tech interests and why chord follows se output only. This pertains to making the dac amp as quiet as possible reduce distortion and increase accuracy. This has no reflection on good dacs that provide balanced just the intricacies and challenges involved. Remember and to quote maybe a harsh metaphore in this instance, garbage in garbage out, here is rob watts quote pertaining to Hugo architecture.
Yes single ended is best, but it exposes other problems. I am talking here about the DAC itself. Conventional hi-end chip DAC's are fully differential which means that the DAC supplies a negative signal current OP and a positive signal current OP. This means that the analogue OP stage needs 3 op-amps - two single ended (SE) current to voltage converters (I to V) plus a differential to SE output op-amp. Then if you want headphone drive, then another stage is added. So you can see the analogue section is quite complex, with lots of active and passive components involved, all reducing transparency. Whereas Hugo is entirely SE, which means that it is one beefy OP stage, with one global feedback path - so we only have a single active stage and 2 resistors and 2 capacitors in the direct signal path. The OP stage handles both analogue filtering and I to V conversion. To make the analogue filtering easier, then the problem is transferred to the digital domain, which is why the noise shapers run at at 2048FS and it is digitally filtered up to 2048FS. Thus the raw DAC OP has very low out of band noise, which means that simple analogue filtering is needed.
Now the reason chips are always differential is that it cancels substrate noise, and other chip common mode noise, and of course with discrete DAC's we don't have this problem. But the other reason is that it hides the effect of the reference supply - the noise of the reference is cancelled at small signal levels (but not at high signals so this is a source of noise floor modulation which dramatically upsets SQ) and it also hides the OP impedance behaviour of the reference. So going SE means that you have to take great care with the reference supply, but you have to do this anyway if you want to eliminate noise floor modulation which is something that conventional chips are not very good at doing.
So Hugo's success is in part down to it's very simple and direct OP stage - but to do this I have had to be very careful with the references, and digitally filter to much higher levels than normal. Going balanced just adds to complexity, thus reducing transparency.