Thomas, while it is true that the analogy of ATRAC3:MP3 :: SACD
VD-A is not perfectly valid in every way, it does demonstrate one thing -- (file) size isn't everything. This is because DSD and PCM are, in themselves are different methods for
representing the sound, not for processing the sound. Granted, though, much of the problem with DVD-A is the longer chain it has to go through during decoding.
I agree that algorithm was a poor choice of words. PCM and DSD really aren't algorithms because an algorithm is a method of solving a mathematical problem. But you understand what I mean -- if you're approaching the problem of representing the sound in different ways, even using only 61% of the memory that 24/192 PCM audio does, SACD's DSD can still in principle be as good as DVD-A.
First, while the noise shaping can be bad for frequencies above 30 kHz, we can't hear that high anyway, so it's not really an issue. Some super-tweeters, as I understand, can reach up to about 30 kHz, but none are even only in the 20-kHz-and-over range.
Second, the issue of converting PCM to DSD is not a non-issue -- DVD-A is made that way, and this unfortunate chain of processing is a downside of DVD-A.
Ugh, I've got to go. I'll finish this message soon.