I'm curious, it seems in the current discussion one of my questions is creeping in again. What are the exact mechanisms by which an R2R DAC sound better than a D-S DAC? I have very little doubt that some of the better R2R designs we have been hearing about in this thread are actually more musical, better sounding DACs as compared with many of the garden variety D-S DACs, but it does seem reasonable that the real difference in the sound could be mostly attributed to the digital reconstruction filters and or the analogue stage of the DAC (as well as the digital input stage). So if this were at all plausible then couldn't a D-S DAC that was equally well designed in these key metrics sound equally as good?
I guess a related question of mine would be is it the R2R design that allows the digital filters to work as implemented in say the Theta's and the Yggy? If I have misunderstood this discussion my apology.
The basic bottom line is that the purpose of a DAC is to convert a list of numbers into an analog voltage. Therefore, assuming that whatever method you use does that well enough, how it actually works is really not terribly important. The circuitry involved in a "basic R2R ladder DAC" is theoretically simpler than that of a Delta-Sigma DAC. In analog circuitry, there are certain drawbacks to making circuits more complicated - basically, since the analog signal picks up noise and distortion whenever it passes through a part, a more complex circuit has "more places for things to go wrong". However, this really isn't the case with digital circuitry. (A modern computer has a lot more transistors in its CPU than a 25 year old one, and it runs faster, but it still delivers numbers that are "just as clean".)
I've only ever heard two of what I would consider to be "credible technical arguments" about specific reasons why oversampling DACs in general would be "inferior" - and neither specifically applies to only Delta-Sigma DACs.
First,
ANY type of oversampling filter will introduce
SOME ringing. It's a sort of side effect of how digital filters work, and is unavoidable with any current filter design - you can have super accurate frequency energy response or super accurate time response but not both (the ringing you see is there because some of the energy that belongs in the impulse has been "smeared" and appears at the wrong time; if you simply wipe out the smeared energy, then the total amount of energy is no longer perfectly correct). Now, pre-ringing and post-ringing sound different, at least according to some people, and it would be nice to minimize the amount of ringing to a bare minimum - or make sure none of it is in the audible frequency range - so this is one area where extra effort may indeed yield improvements (or, at least, differences). One current idea is that post-ringing is less audible than pre-ringing because it is masked better by the actual signal, so many current filters use some math tricks to "shift" the pre-ringing to post-ringing (you get less or no pre and more post). SOme people prefer the way these filters sound - other's don't.
Second, jitter produces distortion - because, when there is jitter on the clock, then you have "correct samples at slightly incorrect times" - so, when you convert these samples into analog audio, you get errors. The amount of error you get for a given amount of jitter depends on how much jitter you have as a percentage of the clock period. So, if you have a file sampled at 44k, with 1 ps of jitter on each sample, and you upsample it to a 10x clock "locked" to the original clock, you may well end up with a data stream that is sampled at 440k, but still has 1 ps of jitter on each sample (assuming you use a simple clock, locked to the original, and no jitter reduction or filtering mechanism). Since that 1 ps of jitter is now a larger
PERCENTAGE of your new higher sample clock, it may produce a larger amount of distortion when your new signal is converted. This means that, in order to maintain an equivalent level of performance (in terms of jitter caused distortion), for any oversampling DAC, including a Delta-Sigma one, you must maintain a lower level of jitter. (In other words, all else being equal, a Delta-Sigma DAC may well be a lot more sensitive to jitter, and so you have to assure a lower level of jitter to avoid problems. (Or, to say it in reverse, if you have jitter that you can't eliminate, it will probably cause a Delta-Sigma DAC to distort more than it would a simple NOS DAC.... however, jitter isn't that difficult to reduce or eliminate, and oversampling offers more than enough benefits to offset this "limitation".)