In article , David Looser
wrote:
"Jim Lesurf" wrote


And as I've pointed out there is no reason in principle why that can't
work fine.

In theory it can work fine, but a high-performance analogue filter for a
non-oversampling converter is neither simple nor cheap.

Indeed. Hence the earlier points I made.

The devil is then in the details when it comes to assessing one
implimentation versus others and the nominal spec for CDDA. I can
think of potential advantages and snags of either approach.

And your potential advantages for the non-oversampling approach are?

As ever, that depends entirely on the implimentations being compared and
the purpose/use/etc. However there is one example so obvious that I'm
puzzled that you don't seem to have realised.

If you don't employ oversampling or resampling then you don't risk problems
with some intersample peaks being above 0dBFS. Oversampling and resampling
systems need to allow for values as much as 3 or 4 dB above any individal
values in the source stream to avoid overflows. Using a non-oversampling
DAC followed by analogue filters with enough headroom means the digital
system doesn't need this extra 'bit' or so above 0dBFS.

Of course, correctly implimented oversampling systems will allow for the
needed headroom. Just as correctly implimented non-oversampling ones with
use appropriate analogue filtering. So the exchange is that oversampling
reduces HF spurons at the risk of intersample problems, whereas
non-oversamplng reduces intersample problems at the risk of HF ones. Yer
pays yer money and yer takes yer choice. Ideally getting something you are
happy with either way.


I've been happy enough with oversampling DACs like the Meridian (and
now the DACMagic) for some years. Still enjoy the results they give.

Are they multi-bit oversampling designs, or, like the majority of audio
DACs these days, of the 1-bit variety?

Afraid I can't recall if the 263 and 563 I use are 'one bit' or 'low bit'.
Either way they are moderate order delta-sigma oversampling as I recall.
The DAC magic I think just resamples to a higher rate whilst keeping a fair
bit depth per sample. Does this by applying a TDA type method I think to
get the 'intersamples' with the chosen filter properties.

Pure '1 bit' needs to be used with caution, as does high order delta-sigma.
Too easy to get into chaotic trails or problems with levels that approach
near max. IIRC That was (is?) a basic snag with DSD/SACD. I think Philips
simply arranged for DSD streams to avoid going above the central range to
minimise that cropping up in use. But given what producers do with heavily
clipped CDs I wonder if any SACDs have problems. Can't say as I only have a
few (hybrid) SACDs and just listen to their CD layer. Although comparing
the layers is one of the many items on my "must get around to that" list.
:-)

Slainte,

Jim

--
Please use the address on the audiomisc page if you wish to email me.
Electronics http://www.st-and.ac.uk/~www_pa/Scot...o/electron.htm
Armstrong Audio http://www.audiomisc.co.uk/Armstrong/armstrong.html
Audio Misc http://www.audiomisc.co.uk/index.html