I love debates with sane people
@mrspeakers: I'm sticking to DACs for the moment, where load is not so much of an issue. With amplifiers, obviously the load can affect the sound, but in a well designed amp I don't think it should. The AMB M^3 measures uniformly well into a variety of headphone impedances, for example.
@Upstateguy
Regarding the power supply issue, the power supply is a potential source of audio signal degradation. It can affect the sound, but should not unless it is crap. Should it audibly affect the sound, this will be apparent in the measurements - even inaudible anomalies caused by the power supply are very easy to measure - I remember Ti Kan on the AMB forums mentioning their presence - but far, far below the levels of audibility.
To address your other points:
NOS DACs are nothing like as prominent as ordinary DACs IMO. As I said earlier, NOS DACs are to ordinary DACs as tube amps are to ordinary SS amps. They do something audible to the sound - in the same way that applying equalisation or other forms of digital or analog signal processing (tone controls, digital noise reduction ect) will change the sound. These are clearly measurable things - but as I said in my previous post, no-one is arguing that every DAC will sound the same. If I design a DAC based on some misguided audiophile mythology that measures abominably but makes great square waves - and it obviously sounds different to something that is well engineered - okay, okay, some people may like what NOS DACs do - I'm not doing well at considering my dislike for many of them. But making a DAC that does something of questionable sense to the sound is not what the "DACs generally sound the same" people are actually getting at. They are annoyed by the well measuring cheaper components which inexplicably sound worse than the latest high-end product (according to Stereophile, anyway).
We HAVE established the effect of every conceivable component in the DAC because, in the end, the effect of every single component will be represented by the overall measurements. You can marvel that modern electronics and the designers that work with it can take huge numbers of components and make them behave in extraordinarily predictable ways, but in the end if you have a DAC that measures far, far beyond the threshold of audibility, why should it concern you what "effect" the power supply has? Every component has an effect, but these effects are run in simulations, balanced and accounted for - to switch to the world of amps for a moment, the Beta22 was designed almost entirely using computer simulation. Additionally, these are tiny changes - circuit boards are optimised to reduce crosstalk and the like quickly, and largely by computer - but the resulting differences would probably be inaudible anyway. These designs are done to an extremely high standard - on one level it is miraculous that faced with the huge array of components even good designers can consistently churn out stuff that measures ludicrously well, but with the aid of science and modern technology, it is none-the-less the honest truth.
In the end, if the overall measurements say everything is inaudible, demanding how each individual component affects the measurements is academically interesting, but does not have any bearing on what the overall measurements show - an effectively perfect product.
Your comments about the resolution of the chip: The latest Sabre chip almost certainly has more resolution, but not more audibly. The reason DAC chips keep on improving is firstly for practical reasons - pushing ahead with techniques to get better and better measurements will trickle down to cheaper designs - for example, the manufacturer of a TV with a digital TV receiver would doubtless be interested in a DAC chip to process the sound part of the digital TV signal that requires less and dirtier power to do its job as well as the one they were previously using. Secondly, there is a demand for better chips - gotta have something to sell the audiophiles - and the engineers that design them would rather give genuine, if completely and utterly inaudible beyond all shadow of a doubt, improvements.
The reason I admittedly tried to avoid that question is because I respect Tangent - his website is full of invaluable information for the audio DIYer. However, the time comes when you have to bite the bullet for your beliefs, so I'll say that if an opamp is properly served by the design (eg taking into account its power requirements and varying input impedance) and it measures above a certain level - see previous posts so we don't get back to the defining an exact level problem, it sounds exactly the same as other well-measuring opamps. Besides, the general consensus on opamp sounds varies from forum to forum IMO to some extent, although this may be inherent bias on my part.
Regarding your question about mosfets, I feel it's time to use an analogy, which is probably going to be awful, for which I apologise, which will hopefully explain my outlook on this idea of how there are so many factors involved, differences that are audible must result. Imagine you have several monstrously complex equations, all of which perform roughly the same mathematical task - say to work out the strength of foundations required to support a building. Both equations are fed with huge numbers of values to their variables; some variables are the same across both, some specific to one equation. In the end, they both turn out exactly the same number, give or take a few thousands of a gram of load-bearing ability. (Wow, this really is a bad analogy, but I'll soldier on regardless). Both approaches, due to the advanced scientific knowledge behind each of them, have effectively both arrived at a solution that is to all intents and purposes, perfect. This can be compared to two well designed DACs producing effectively audible output. Simply pointing to the different values some of them are fed with, or their different approaches, does not disprove the fact that they both did their job supremely well. We can marvel at how different their approaches are and how they have dealt with incredible complexities, but the facts remain the same.
Regarding your Benchmark vs DACmagic request for proof - going to use another analogy, sorry. I have two objects which I have measured to all intents and purposes to be identical ( the human eye being unable to detect 0.0005% differences in diameter on a 10cm wide cylinder, for example). If someone suggests they are perceptibly different, the burden of proof is upon them, not me, to demonstrate that they are correct.