Overview and challenges
The synchronization of large spatially distributed clocking systems is the basis for most timing-related services. The International Atomic Time (TAI) is based on a worldwide distributed network of atomic clocks that synchronize repeatedly. It serves as the basis and pace keeper for the Coordinated Universal Time (UTC) standard.
Digital synchronization protocols such as the Network Time Protocol (NTP) and the Precision Time Protocol (PTP) are used to synchronize networks and support the distribution of UTC timing-services to societies and most of their infrastructure. Also linked to TAI and UTC are the GNSS services that provide timing, positioning and navigation services.
Applications to which no permanent connection to internet or GNSS based timing services is available, or to those that require more precise timing, alternative means of synchronization are needed. In principle the set of technological synchronization solutions available, depends mainly on environmental properties, the networks’ spatial extension and required time-resolution, i.e., frequency. As synchronization relies on the exchange of information, it is subject to signaling time delays that can cause an equivalent time-offset between spatially distributed clocks. When sending signals at speeds close to that of light, a distance of only 30 centimeters causes a signaling time delay of about a nanosecond. This is of the order of the time-scales of processes that operate in the Gigahertz regime and can hence affect their collective dynamics. Timing errors larger than the period of the clocks may be introduced. The difficulties involved in measuring signaling delay induced time-offsets range from time-dependent properties of the signaling to the increasing uncertainties at higher levels of time-distribution hierarchy.
An alternative approach to hierarchical time-distribution and synchronization concepts mentioned above is mutual synchronization. It relies on the mutual information exchange between the oscillators and has been shown to scale efficiently in terms of the quality of synchronization as network size increases [Lindsey1985]. It had been found that the synchronization properties scale advantageously with network size. Especially the phase noise reduction as system size grows can be of great importance in large distributed systems. At the same time, the complexity involved in designing the hardware backbone of such solutions increases [Lindsey1985].
The main difference with this approach is that cross-coupling time-delays do not translate into phase-differences as in hierarchical approaches, but instead the system responds by adjusting the frequency of the synchronized state. Hence, synchronization with constant or no phase-differences within a spatially distributed systems can be achieved without delay-compensation techniques or the necessity to monitor the associated signaling times. In order to achieve time-synchronization on top of that, it is necessary to reset the counters that count the oscillators’ cycles via, e.g., Einstein synchronization once frequency synchronization has been achieved in the system.