[sdiy] Digital audio demo on Youtube

[rburnett at richieburnett.co.uk]

> Actually, nearly all audio DACs from the last decade - and certainly
> all 24-bit audio DACs - are sigma-delta converters which literally
> cannot hold a constant value. So, it's not that they've solved the
> ZOH problem, but rather that they've created an ultrasonic
> oscillation problem in its place.

They have solved the reconstruction problem within the audio band with 
digital interpolation, and they've pushed what hash there is in the 
output well up into the RF spectrum with the use of noise shaping.  It 
is then trivial to remove the RF hash with a simple 1st-order RC lowpass 
filter that has it's cutoff set well above the audio spectrum where it 
won't harm the precious audio frequency response.

> I personally think that it's irrelevant whether the DAC used ZOH, and
> therefore irrelevant that even the intermediate analog signals might
> be stair-stepped. The real point is that the final reconstructed
> signal has no stair stepping. How you get there shouldn't affect your
> perception of the final analog signal.

That is it precisely.  The stair steps don't exist in the digital 
domain, and they don't exist in the final analogue output either.  They 
might exist somewhere inside the audio interface as a "mid-way point" 
between the digital sampled data and the smoothly varying audio output, 
but so might miniature hamsters with a bank of variable resistors, i've 
never checked ;-)  Stair steps are not required or necessarily helpful 
in understanding how the modern audio CODEC works.

-Richie,