Oddly, I consider myself fortuneate in having a less elegant, prior implementation of Bob's dac, the Zhaolu 2.0. Because my Zhaolu 2.0 provides balanced outputs, it uses a typical app note implementation of each phase of a channel feeding its output to an op amp (which provides the balanced outputs) and after, the 2 phases combined in another op amp to provide single ended outputs. [This is the same topology as found in the SB Transporter]
I use the single ended outputs, which allows me to tune the subjective results by using different opamps for dac filtering and balanced to single ended output. For the dac chip (CS4398) output filtering I use 1 dual OPA2107/channel (each phase served by its own amp in the dual), and for the balanced to single ended combining and output to the external world, one dual LM4562, with each of its amps serving its own channel. In total, 3 dual opamps for the analog section of the DAC for both channels.
In reality, this topology is no more complicated than that in Bob's Zhaolu 2.5. My 2.0 has an additional opamp per channel, as each phase, gets its own amp, rather than both phases of a channel filtered and combined in a single opamp, and then buffered. But most importantly, it allows me to use different monolithic opamps for the filtering and single ended output functions, and get what I want subjectively. (nor do my choices compromise objective measures)
This also leads me to the sad conclusion that audio reproduction is far more 'additive' than I'd like from a superficial intellectual perspective. I've not evaluated the newer Zhaolu 3.0 which provides single ended outputs from a single opamp per channel, combining phases, filtering and output in one stage. Perhaps the LT1028, or the new SiGe opamps from TI. But I've yet to encounter a monolithic opamp that 'does it all'.
FWIW,
Paul
