Michel Hafner wrote:

John Atkinson wrote:

Michel Hafner wrote in message
...
John Atkinson wrote:
The converters were dCS 904s, which use an internal 5-bit flash
converter.

Had a look at this unit. It does not do 24/96 or 24/192 AD. It does 5 bit
AD with high oversampling from which either PCM or DSD versions are
computed. You can chose between different noise shapers for DSD and
PCM. The result is then again converted with a 5 bit DA converter.
That unit has too many variables and is too hybrid to allow conclusions
about what is closer to the mic feed, pure DSD or pure high res PCM.

Not sure I grasp your point Mr. Hafner. If you want to record in DSD,
and you are not Philips or Sony, you have only 3 choices for an A/D
converter: the dCS, the Meitner, and, I assume, the converter in
Tascam's relatively new HS-98 MDM tape machine. I haven't heard
anything about the Tascam's converter; I am told the Meitner and the
dCS sound different from one another, with each having its advocates.
So, if I wanted to compare the mike feed with a DSD representation
of the same, and you feel the dCS is not appropriate, which converter
would you recommend?
My point is that the dCS is not a pure high res PCM system. And I don't
know which noise shaper was used anyway for the comparisons. They obviously
sound all different hence you are given a choice. So my point is that while
the dCS system 'may' sound closer to you via DSD and noise shaper x versus
PCM and noise shaper y both played via 5 bit oversampling DA (I assume, or
was there a dedicated 24 bit DA for the PCM stream?) it tells me not really
if PCM or DSD stores a closer approximation of the analogue signal.
Information theory clearly says that 24/192 is superior as a storage format
than 1 bit DSD at 2.8 MHz. So assuming DSD sounds closer I have to suspect
implementation issues as the reason for this.

I would also like to see how the comparison is done. Is it in real time,
when the listener can switch between the live feed and what comes out of
the D/A, or is it based on memory of what the live feed sounded like?
And is it well-controlled (blind, level-matched, etc.)?

I think a much better test is to simply convert the output of the DSD
D/A into 16/44, 24/96 and 24/192 (maybe also analog tape, too). Then do
a DBT among those outputs. If the PCM's sound identical to DSD, then the
PCM's are transparent. Likewise, one can convert the output of PCM D/A
into DSD using the DSD encoder, and then compare the corresponding
outputs. Wait, I think that's what Norm Strong suggested...