Picoseconds Or PPM

máÅçëÉÅçåÇë
=
çê=ééã\=
www.grimmaudio.com
[e-mail]
=
An interesting myth has made its way into the market recently: that a 48000.000Hz clock makes
your audio sound better than a 48000.100Hz one. Okay, that’s not the way they put it. Nobody can
keep a straight face and say that. Instead they say that your audio improves when you lock it to
an atomic standard. For some reason people find that more convincing. Why? Semantics!
=
Accuracy? Stability?
Atomic clocks are “accurate”. Some are only a second off
in a century (about three billion seconds). That’s the sort
of thing they were designed for. Does that imply that
they are only one three-billionth of a second off in one
second? Not at all. When a clock runs one millisecond
late during one second and one millisecond early during
the next, it’s basically correct again. So long as these
short term errors don’t accumulate over time, they do
not disqualify the clock as a timepiece. But you will
agree that it would not merit the name “stable”. A clock
that is one second off every day could still manage to
slice every second into tiny slivers that are just perfectly
equal. Accuracy and stability are two very different
things. People make and use rubidium standards for
audio simply because it is very easy to confuse accuracy
with stability so it sounds all to plausible that accurate
clocks are better for audio. Add to that the tech-appeal of
laboratory equipment and you’ve got a story that
spreads like wildfire.
What does it sound like when an
audio clock is 10ppm off?
Well, it makes the recording sound as though the speed
of sound during the recording were 10ppm faster or slo-
wer. The speed of sound increases by 0.17% per °C (or
some 0.1% per °F). So a static 10ppm error sounds like
the temperature in the recording hall was 0.0057°C
(0.01°F) higher or lower. People who are trying to con-
vince you that those 10ppm matter are actually saying
that a temperature difference of 0.0057°C during a mu-
sical performance makes all the difference.
Seriously: the best trained ears can detect pitch diffe-
rences only down to about 700ppm. Red book specifies
that CD players be no more than 200ppm off and the
CC1 is typically accurate to 2ppm to insure any thinka-
ble downstream device will lock. Tighter accuracy specs
serve no purpose as audio is concerned.
What does it sound like when an
audio clock jitters?
Well, to keep the analogy: it sounds like the speed of
sound is constantly changing, modulating the sounds
and spatial cues that it carries. This ties in well with
subjective reports that improved jitter makes it much
easier to pick out placement and ambience without ha-
ving to strain one’s ears. We can debate what the smal-
lest amount of audible jitter is, but there can be no doubt
that jitter matters infinitely more than whether a clock
can be used to keep time for a century. If sound quality
matters, the battle to win is reducing jitter. It happens
to be much harder than getting absolute precision.
How do rubidium clocks work?
They operate on the fact that if you shine the light from
a rubidium lamp through a cell filled with rubidium gas,
you get a 0.1% increase in absorption if at the same time
you submit the gas to an electromagnetic field oscillating
at 6834682610.904324Hz. So you make an oscillator
operating at nearly this value and you constantly wiggle
(ie. intentionally jitter) the frequency around to home in
on this tiny dip. The wiggling is needed because if your
oscillator wanders out of the dip, you don’t know which
way it went.
The oscillator already has to be quite good so a very high
quality crystal oscillator is used followed by a multiplier