System-level synchronisation reference compensation for extending time holdover

Phase and time synchronisation have become the fundamental requirements of most telecom and data communications networks today, as numerous end applications use precision timing. Whether consumer, business or industrial applications, the enabling technology requires synchronisation to achieve functionality and performance.

To provide the throughput of 5G, features such as carrier aggregation require stringent phase alignment. Likewise, keeping the air interface of radios within +/-130ns of the same base station is demanding, even when synchronisation sources are available.

Applications that rely on URLLC (Ultra-Reliable Low Latency Communication) – which enables 99.999% availability of services – need very reliable synchronisation in all possible scenarios. Time Sensitive Networks (TSN) such as industrial, financial and automotive networks, data centres, power grids and 5G Virtual RAN deployments, all require high precision synchronisation.

Look beyond standards bodies’ recommendations

The standards bodies provide hierarchical reference models and deployment scenarios for synchronisation. However, actual deployments depend on many circumstances that may not be factored into the standards. For example, for mobile operators, interoperability and co-existence with current networks are critical. Operators generally try to re-use the existing network rather than deploying a new network for new technologies. In such scenarios, the traditional synchronisation mechanism may not work, and atomic clocks tend to be ‘parachuted’ onto the edge nodes of the networks.

The illustration here is a generic implementation of a synchronisation system. The GNSS-based clocking synchronisation is the primary reference clock, and in the event of GNSS signal loss, the PTP based clocks take over. PTP based ’holdover’ can provide phase as well as frequency corrections. There is one additional layer of protection even when the PTP based clock failure supports physical layer clocks. Such clocks can only do frequency synchronisation and are capable of holding the phase. Finally, the fundamental and default holdover comes from the reference clock of the system.

Holdover of servo systems is mainly affected by the following issues

1. Servo initial error: Holdover may occur at any instant of operation of the servo system. At that instant, the error of the system is caused by the initial phase error and frequency accuracy of the servo.
2. Frequency deviation due to temperature variation: Temperature variation is the fundamental short-term impact on oscillators. Higher frequency stability ensures minimal phase movement.
3. Ageing drift: The fundamental building block of the oscillator is a resonance element that builds up the fundamental oscillator. Ageing is the systematic variation that happens to the resonator over time by materials being deposited on the quartz crystal or materials being vaporised from it. This results in slow variation in frequency and long-term deviation in phase movement. Ageing is presented in parts per billion per day (ppb/day) variation.
4. Hysteresis and other system effects: The frequency difference that the resonator accumulates over temperature cycles is called the hysteresis effect. The re-trace impact is the difference in frequency an oscillator exhibits when it undergoes a power cycle. Power supply variations and load variations also cause the frequency to deviate.
5. Random variations: Due to the nature of the resonator, frequency drift can be caused by random effects, primarily measured in the Allan Deviation metric.

The servo initial error is mainly controlled by the servo design. The hysteresis, re-trace, power supply variations and the random variations are related to the resonator characteristic and the basic circuitry design. Of all the different types of contributors to the changes in reference clock frequency, the major contributors are variations in temperature and drift due to ageing. A system trying to achieve short to medium term holdover (4–8 hours) will have a significant impact (90% variation) from these two factors. Therefore, the system designers have been engaging various techniques to improve the holdover performance by addressing these two specific issues. Temperature variation and ageing drift are variations that can be dynamically addressed in the systems. There are a number of methods that can be used to compensate the synchronisation reference caused by temperature variation and ageing.