When clocking an audio interface using an external device, such as a stand‑alone mic preamp with A‑D conversion, is it better to use the dedicated word-clock (normally BNC) connectors, or will the clock signal be just as good coming down the S/PDIF, AES3 or ADAT cable that is carrying the audio signal?
Sam Wood, via email
ASOS Technical Editor Hugh Robjohns replies: Theoretically, recovering the clock from a dedicated clock source via word clock should be easier and better — with less intrinsic interface jitter — than trying to extract an embedded clock from AES3, S/PDIF or ADAT interfaces.
The reason is that word clock is a really simple square-wave signal running at the sample rate. Cable losses will slow the transitions, so what falls out the end of the cable will probably be a little more rounded than what went in. But the important point is that the periods between each transition from high to low (and vice versa) will remain absolutely consistent. It's those transitions that define the sample points.
By contrast, AES3, S/PDIF and ADAT all embed the clock pulses within the audio stream by using a type of modulation process. The result is that the transitions depend on both the audio data and the clock data, so the transition rate varies. The inherent capacitance within the cable reacts with those transitions and the result is small, but varying, time shifts of the transition point. This process is called 'interface jitter', and it is an intrinsic part of that kind of digital audio interface. The diagram above, courtesy of Prism Sound, illustrates the effect.
Fortunately, interface jitter can be removed fairly easily using phase‑locked loops (PLLs) or other, often proprietary, techniques these days. Moreover, it doesn't actually matter in the case of transferring digital audio from one digital machine to another, anyway, provided all that is required is to identify each data‑bit value reliably.
Where interface jitter becomes important is in the D‑A conversion stage, or when clocking an A‑D converter from an external master clock. In both of these situations, the signal is being converted between the analogue and digital domains, and precise, jitter‑free sample-clock timing is, therefore, absolutely essential.
That's why it's best to run an A‑D from its internal crystal clock, rather than an external master clock, if possible, and why good D‑As go to so much trouble to isolate and remove interface jitter.