Audio_Empire wrote:
In article ,
Andrew Haley wrote:
Audio_Empire wrote:
This is all fine and good, and perhaps you can even hear "some*"
difference, but it still doesn't alter the fact that the Oppo in
question still uses the same SabreDac 32 which converts DSD to LPCM
before playing it,
I don't think so: according to their white paper, they feed the DSD
data into their anti-imaging filter, which runs at a very high
frequency. There's no suggestion that they downsample it to LPCM
first; or do you have some other information?
They use the SabreDAC 32. It converts DSD to LPCM. My information comes
from the Oppo Technical guy.
Hmm. Must be true, then. Under the circumstances, I think I'll go by
the ESS white paper. But even if it does, there's no reason to
suspect that doing so will cause any audible damage.
Why would downsampling to LPCM do any harm, anyway? All the damage
has been done already during the DSD encoding.
Damage? What damage?
Single-bit sigma-delta converters are inherently unstable.
Whether it's during analogue to digital conversion or shortening the
wordlength from PCM, you must add dither to linearize the process.
The problem with single-bit sigma-delta (aka DSD) is that it is
impossible to add enough dither without overloading the modulator.
(There wouldn't have been any problem if the designers of DSD had used
two-bit sigma-delta instead of one-bit.) DSD-wide (i.e. 8-bit
sigma-delta) solves the problem, but you still have to go down from
wide to 1-bit to maks a SACD, and the problem recurs. This was proven
by Vanderkooy and Lip****z in their clasic paper. [1]
There was a rather feeble reply from Philips, but no attempt to rebut
Vanderkooy and Lip****z's proof. And it is a real mathematical proof,
the most reliable indication of truth there is.
These days, the highest-quality audio converters use multi-bit
sigma-delta modulators. (The one in the Sabre DAC is probably six
bits wide, although the white paper is rather vague about that.)
For these reasons, I suspect that if you want to convert DSD to
analogue (and remove high-frequency spuriae) there isn't any better
way to do it. As usual, I'm happy to be proved wrong by some real
evdence.
Andrew.
[1]
http://sjeng.org/ftp/SACD.pdf