From a technical standpoint, the paragraphs George quoted from the MSB literature all make sense to me, as opposed to being the kind of marketing techno-babble that seems all too common in audio-related white papers and other literature.
An additional theoretical advantage of using sign-magnitude architecture in a ladder DAC implementation, besides the one that is stated, is that the "offset binary" architecture they seem to be implying is used in other ladder DAC implementations will cause all of the bits to toggle (i.e., to change from 1 to 0 or 0 to 1) **simultaneously** at or near the critical 0 volt crossing. Which in turn can contribute to noise issues at that crossing, and hence degrade resolution at very low signal levels. With sign-magnitude, only a few bits would toggle at or near the 0 crossing.
I'd have to add, also, that providing 20.8 to 28.5 effective bits using a ladder approach, depending on model, as shown
here, and not just for the D/A circuit but for the unit as a whole, is quite an amazing achievement. Primarily because of the incredibly tight +/- tolerances that would have to be met by the resistors in the ladder.
Which leads me to the one concern that is raised in my mind by a quick look at their writeups. Their focus seems to be on achieving the best possible performance at very low signal levels, i.e., near the 0 crossing. Which arguably makes a lot of sense. But consistent with that, I see in the figure I linked to that their measurements comparing their results with those of delta-sigma DACs were taken with the test signal at -90 db, a very low level. So what I wonder is how well the much superior performance of their approach that is depicted in the graph would hold up at moderate to high signal levels, if in fact it would hold up at all.
Just some thoughts to keep in mind. Best regards,
-- Al