Here is my take on it. Looking at the pictures it seems as on the nfb-3 only the analogue section of the amplifier is distrete (as nfb-11/nfb-12).
It makes sense, the transistors in the amp section dont have near the same noise rejection as the digital parts, so noise will easily make the signal bad. Secondly, the output amp section needs a fast and liniar power suply - the discrete power suply can give that.
For the digital chips - the dac also have an analogue section and analogue power supply besides the digital, and in my experience the analogue is also very, very sensitive to noise. Dont think of a dac as a digital processor - think more of it as a analogue process. Thats the hard part of it. And think of the digital signal as analogue in its behavior.
With a stable low jitter clock, stiff ground plane, low data jitter and noise free, liniar and fast power suply, the conversion is not so critical. But how do you get a low jitter clock, low jitter data stream from a computer, and how do you - for low money - make low noise power suply. Its not that easy
![Headphone Smile :) :)](https://cdn.head-fi.org/e/headfi/smily_headphones1.gif)
- it can be more cost effective to have a more expensive dac that is less sensitive to clock jitter, data jitter and worse power suply (like the pcm1704
![Headphone Smile :) :)](https://cdn.head-fi.org/e/headfi/smily_headphones1.gif)
). But audio-gd gives us spades of hardware for our money anyway.