I do want to address a few things specifically to Dogmatrix here....
While it makes sense at one level that it should be difficult to sell any audio product these days with "audible noise or distortion"... the reality is actually the opposite in much of the audiophile market. Specifically when it comes to electronics, any two products that were "audibly perfect" would perforce sound exactly the same. The reality is that, while many audiophile products are sold based on other distinctions, real or imagined, many do in fact sound quite different from each other. In some cases these differences are due to obvious differences in well-known characteristics like noise and distortion. Some of these differences may be unintentional or incidental - for example, no speaker is audibly perfect, but most speaker manufacturers have a "house sound". This is simply a way of saying that, accepting that perfection is out of reach, their designers and their customers have agreed to prefer certain flaws in deference to others. (There is no such thing as a perfectly flat speaker. So, accepting that, would you prefer a small bump
here or a slight dip
there? And, do you prefer your speakers a little bright or a little laid back?) In other cases, the designers may have deliberately chosen to make their product sound different, so there is some way to differentiate it from other similar products. (It's much simpler to convince potential customers that your product "sounds
better" if there is a clearly audible difference you can demonstrate.)
To pick the example of the DAC you mentioned - one of the NOS DACs from Metrum....
There is a very well known design constraint on NOS DACs. In order for a DAC to reproduce 16/44k digital audio content properly, without using oversampling, it must include a filter which nominally passes the entire audio frequency band with a flat frequency response, but strongly attenuates frequencies above 20 kHz, to eliminate spurious images. Such a filter is difficult to design and manufacture and involves some serious compromises. (The main reason why oversampling is so prevalent in modern DACs is that it avoids this issue.) Therefore, when designing an oversampling DAC, the designer must either accept a significant deviation from a flat frequency response inside the audio band, or accept other serious compromises.
I owned a Metrum Octave for a while....
When playing 16/44k content, its frequency response is -3 dB at 20 kHz, which is a significant and audible high frequency roll-off.
(I also happen to have access to an AP analyzer... and the THD wasn't especially low either...)
Therefore, while that DAC was audibly quite different, and many people rate it as "sounding very nice"...
There are obvious measured differences that may well account for that audible difference.
The part about manufacturing tolerances is a bit of a red herring... based on a bit of truth...
(It's not that R2R DACs
inherently have better manufacturing tolerances... but that they
require them in order to deliver even acceptable performance.)
In order to design an R2R DAC, and achieve reasonably good performance measurements, certain parts must be specified with extreme accuracy.
(Even a tiny mismatch in the resistor values in the ladder network will result in significant distortion and nonlinearity.)
From an engineering perspective, one of the benefits of the D-S DAC topology is that it can achieve similar or better performance using less accurate lower cost parts.
In technical terms - the D-S topology is "largely immune to inaccuracies that would have a major effect on an equivalent R2R design".
(In other words, in order to build an R2R DAC that works equally well, you must build it out of
MUCH more precise and
MUCH more expensive parts than a D-S DAC.)
(In the vernacular.... R2R DACs are very fussy, if you want even decent performance, but you can make a D-S DAC using really cheap parts, and it will still work very very well.)
Imagine a design for a paper clip which HAD to be made out of platinum to work as well as - but not better than - the cheap steel one you're using now.
Would you describe that platinum paper clip as "sophisticated" or "a terrible waste of money"?
As it turns out, while D-S DACs, for a given cost, perform FAR better than R2R DACs, there are some minor ways in which R2R DACs do deliver superior performance.
Outside of audiophile circles, R2R DACs are still often used in instrumentation and video applications...
However, both of those are applications where very high operating speed and very rapid settling times are critical requirements...
And where very good linearity, low noise, and low distortion, are less critical...
A video DAC must operate at incredibly high speeds... but linearity of 10-12 bits is usually adequate.
In contrast, an audio DAC operates at relatively low speed, but its linearity and noise floor are critical.
In other words, most engineers agree that R2R DACs work much better for video, but that D-S DACs are superior for audio.
And you'll see this stated in many engineering texts - and the spec sheets for many DAC chips.
(The reason so many R2R audio DACs are using "instrumentation DAC chips" is that their manufacturers DO NOT recommend using those chips for audio applications!)
An R2R video DAC, that can operate at hundreds of mHz, with linearity of 10-12 bits, can be had for a few $$$.
(And there are probably a handful of them in your TV or video recorder.)
A D-S DAC with those specs would cost a small fortune - if it could be had at all.
HOWEVER, an D-S DAC that can operate at 384k, with better than 24 bit linearity, costs only a few dollars.
(Which makes it a much better match for the requirements of audio equipment.)
They were sinad measurements specifically so well below auditory perception , it would be hard to sell a dac these days with audible noise or distortion . My understanding is the difficulty is not in the concept of r2r but manufacturing tolerance .
While inaudible factors are an unfair criticism of a product in an audio market free market competition means they are non the less a factor in a products success or failure .
