Towards a quantization based accuracy and precision characterization of packet-based time synchronization

Abstract

Packed-based precision time synchronization is a fundamental enabling technology of modern distributed measurement and control systems. Today IEEE 1588 and derivative solutions such as IEEE 802.1AS can be considered stable technologies with wide scale support from manufacturers, and new technologies built upon them also, such as dependable, real-time communication over Ethernet (Time Sensitive Networking, TSN). However, properties of such solutions in their synchronized state; when error signals tend to be small and close to the minimum value for their numerical representation, is not well-understood, especially in complex systems configurations using several transparent clocks connecting the master clock to slave clocks. In the paper a quantization and sampling based initial approach is introduced for such packet-based time synchronization systems taking into account implementation specific details, primarily the used timestamping approach and effects of finite word-length arithmetic units used in computer systems. Based on the model some preliminary analyses results are also shown that verify that this model is applicable.