It was a big debate in the previous decennium: using external word clocks to influence a digital audio system’s time accuracy. With the introduction of the ‘Precision Time Protocol’ in gigabit networks, the discussion slowly died out and the choice of system clock to influence audio quality has basically disappeared. Many networked I/O racks don’t even have a word clock BNC connector anymore. What happened ?

The external word clock debate was about influencing ‘jitter’. Jitter is a distortion of a word-clock’s pulse regularity, resulting in a very small timing difference in a system’s A/D and D/A conversions. These timing differences in turn generate very small noise artefacts in the audio signal – generally below -80dBfs. Not clearly audible, but it’s there.

Synchronising a device’s internal word clock timing to a more accurate external one influences the low-frequency part of the internal word clock’s jitter. For recording and playback systems (e.g. CD player) this has a direct effect on the system’s timing accuracy and jitter noise level, with many reports about the audible effects. However, in live systems, the lowest frequency jitter is automatically cancelled out, leaving only a small possible jitter influence, depending on the system’s latency and word clock PLL design.

This ‘low frequency jitter cancellation’ is easy to understand by imagining an input A/D convertor circuit running on a word clock with a certain time deviation, with it’s digital signal immediately played out by a D/A convertor circuit connected to the same word clock with the same time deviation. Because the A/D and D/A time deviations happen at exactly the same time, they cancel each other out and no jitter occurs. In reality, networked systems have a combined latency of a few milliseconds, but for very low frequencies (generally below 20 Hz) the cancellation is still significant.

Gigabit Ethernet audio network protocols – such as Dante, Ravenna, Q-Lan, and the interoperability standard AES67, use the PTP “Precision Time Protocol” to synchronise parts of an audio network: mixing consoles, DSP units, recorders, AD and DA convertors. The PTP protocol uses only a few timing packets per second to let all devices in the network know what time it is, each device generates an internal word clock based on this information.

This low frequency of timing packets implies that an external word clock selected as sync master in the system only has effect on timing inaccuracy below this so called ‘corner frequency’ of a few Hertz. Timing inaccuracies in this frequency range are not called jitter anymore, but ‘wander’. Also, the effect of an external word clock will be way below the ‘jitter cancellation’ frequency. Hence, the influence of external word clocks in Gigabit Ethernet networked audio systems on jitter has virtually disappeared and is not a topic of debate anymore.

This doesn't mean that external word clocks don't make sense - they can still be applied for example to synchronise an audio system to a video source. But they have virtually no effect on the audio quality any more.

It has to be noted that the implementation of the PTP protocol in most devices on the market is excellent, supporting synchronisation accuracy in the range of microseconds – similar to non-networked digital systems. At the same time, it has to be noted that word clock jitter generated inside the networked devices is still there – it’s just very low and virtually independent of the system’s PTP synchronisation performance.