Quote:
Originally posted by theaudiohobby
You do not have to look far,, see my previous post for the link and read the followup reactions.The history and here is a previous thread that the later pages cover a few of the key papers of both sides of the discussion. |
Thanks for the links to those threads. I had totally missed them when they came around the first time. They were very interesting reading.
I'm somewhat concerned that most of the data used to support DSD cited in those threads were from white papers from manufacturers, particularly dCS, rather than peer-reviewed scientific work (but read on below...).
Quote:
Edit - Wodgy, sorry I misunderstood your post, however the previous thread does indeed cover some key papers that answer the Lip****z contention, however I have yet to locate a key AES response paper that directly address some of his key contentions, the paper was reported in the press at the time, however I do not think it is in public domain. |
Yeah, the Stereophile "rebuttal" is outdated and not really on topic. I was able to find a few follow up papers on both sides of this topic, some from the manufacturers. (I'm only interested in peer-reviewed stuff though.)
It seems that everyone on both sides of the issue agrees that single-bit ADCs are insufficient even if your final goal is DSD (you posted this yourself when you mentioned that good DSD quantizers had to be multibit). Where there is considerable disagreement is whether multibit DSD ADCs can generate an adequate DSD signal. There is a rebuttal paper from Philips published at the AES that gives empirical results showing that multibit DSD ADCs are indeed good enough, but they do not rebut the claim in the final part of the Lipschitz and Vanderkooy paper that DSD can never be perfected, even with multibit techniques (see the Appendix of the original Lip****z paper for the multibit discussion). The math is merely mentioned there, but Lip****z and Vanderkooy have two follow up papers where they rebut the criticisms and work through all the math properly: "Toward a Better Understanding of 1-Bit Sigma-Delta Modulators, Part 2" and "Toward a Better Understanding of 1-Bit Sigma-Delta Modulators, Part 3." Later, some researchers at the University of London came up with a good paper that demonstrates how to construct audio signals that will have
audible distortion when encoded as DSD, even using the best multibit techniques:
http://www2.elec.qmul.ac.uk/%7Ejosh/...ons/AES114.pdf
When you step back, there's no contradiction here. The manufacturers are saying that the technology is adequate and perhaps even superior to PCM under certain circumstances and implementations which may be common or more practical, while the scientists are saying that DSD isn't perfect and is indeed techically inferior to PCM assuming perfect implementations. The sticking point which doesn't totally sway the discussion towards PCM entirely, is pointed out in a Philips paper that I found at work but can't find now for some reason, is that DSD is somewhat more information-rich in the time-domain (but weaker in the noise and frequency response areas). So the debate will probably go on forever given the technologies as they stand today. I will concede now (and I didn't realize this earlier) that both SACD and DVD-A have their own advantages and disadvantages, though IMHO DVD-A is still probably the winner overall.
From my understanding of all this, if there was some mythical 24/392 PCM format (sampling about as twice as fast as DVD-A), there would be little basis for debate at all, and that it would be hard to argue that DSD had any advantages at all there, even given comparable or somewhat higher sampling rates, but alas such a mythical format doesn't exist, and the debate will probably continue. I'm exiting now because I have other stuff to do tonight.