On Thu, 11 Dec 2003 10:06:03 +0000 (GMT), Jim Lesurf
wrote:

In article , Ian Molton
wrote:


Scenario 2: The two clocks are not a prefect match, the DAC clock is
either slower or faster than the data sources clock.

in this case, if its faster, it will (periodically) drain all its
buffering, no matter how much there is, and will end up stretching bits
to fill the gap (or playing silence, whatever)

This may be the case if there is no mechanism in the receiver to adjust its
medium/long-term clock frequency towards that of the transmitter. However
the S/PDIF stream has clock info encoded into it. Hence a receiver can use
this as it wishes to avoid the phases drifting too far if the designer so
chooses.

Yes, and the Audio Synthesis DAX used a rather cunning system whereby
an essentially free-running low phase-noise oscillator was 'nudged'
every few seconds by a difference signal derived from the datastream.
Essentially a *very* narrow-band PLL, but rather a crafty
implementation.

jitter is simply noise above, and will average out to nothing. If it
DIDNT cancel out, it'd effectvely be a frequency drift of the data
sources clock, which is no longer called jitter (duh).

The above treats this as a matter of terminology. In reality, both the
transmitter and reciver clocks will wander about in an apparently random
manner over all timescales of variation. The object then is to minimise
these wanderings and their effects.

I don't know if there is any official timescale defined in audio for the
longest term that is 'jitter' beyond which it becomes 'drift' of 'static
error'. Hence I'd be inclined to regard these as all facets of the same
problem and treat them accordingly.

Perhaps we could agree that 74 minutes would be a hard limit for
jitter on a CD? :-)
--

Stewart Pinkerton | Music is Art - Audio is Engineering