Slave to Rhythm Primer - PLL and Clock Basics

Mark that in this graph we are not looking at
audio frequencies, but at jitter (or ‘fluctuation’)
frequencies. The blue curve shows the
attenuation of input jitter. The red curve
shows the attenuation of the local oscillator’s
jitter. Below the cut-off frequency the external
oscillators jitter dominates, above cut-off
the local oscillators’. A cut-off frequency at
4 kHz or higher can be found in AES/EBU
receivers and general word clock inputs. If he
regards the quality of his local clock highly,
the designer of a PLL can decide to put the
cut-off point much lower, for instance at 10 Hz
or even further down. This makes for a ‘slow’
PLL with a very narrow bandwidth.
A quick way of seeing if a PLL is slow or fast
is to see how long it takes to achieve lock.
Usually, slow PLL’s also take a while to lock.
The CC1 takes about 40 seconds to lock and
has a 0.1Hz bandwidth. Typical AES/EBU
receiver chips lock within a few samples and
have a bandwidth of around 10kHz.
If the local clock is very clean, a narrowband
PLL is the best choice because all but the
lowest-frequency jitter in the external sync is
rejected. A converter designed along those lines
will sound stellar under all conditions. If the
external sync is very clean, a wideband PLL
is the best choice because the local oscillator’s
own errors will be corrected. This is the case
where a good external sync like the CC1
improves a budget converter, or even a pricey
one, beyond expectations.
If the designer guesses wrong however, a too-
fast PLL might end up forcing an otherwise
fine local oscillator to reproduce faithfully
every bump and hiccup in the external sync
signal. Equipment constructed along these
lines sound good in master mode but will only
improve in slave mode if the external sync is
stabler than the internal oscillator. An unstable
external sync actually makes it sound worse.
Alternatively a too-slow PLL might not correct
a local oscillator of suboptimal quality. In that
case, jitter performance is bad regardless of the
quality of the external clock. And here lies the
rub: a slow PLL will always make a converter
sound the same, but not necessarily good. If a
converter is insensitive to external jitter, that
alone is no indication that its internal jitter
is low. A slow PLL shuts the door to external
jitter, but also to any improvement to be had
from external clocking with a very stable
source.
In short, one cannot expect an external clock
to work miracles everytime. If the PLL of the
receiving device is slow, the sound quality will
be independent of the quality of the external
clock, for better or for worse. If the PLL is fast,
real improvements can be had.
By example, the graph below shows the result
of measurements on a well known DAW
converter. The jitter performance, measured
at the converter chips’ clock pin, improves
substantially at jitter frequencies below 200 Hz
when slaved to a CC1.
4f
1f
100p
30p
10f
10f
100f
100f
1p
1p
10p
10p
Jitte
r
(s)
Jitter density (s/sqrtHz)
20 20k50 100 200 500 1k 2k 5k 10k
Hz
DAWslaved to CC1
DAW(master)
CC1