Adaptive minimum filter that achieves precise time synchronization on IP networks

Abstract

Time synchronization protocols, such as NTP (Network Time Protocol), are frequently used on IP (Internet protocol) networks. These protocols estimate the clock time offset by measuring the Round Trip Time (RTT). In this estimation, errors may occur by delay variation, which is included in the measured RTT values. To avoid this, a conventional minimum filter technique using a fixed sample size has been developed previously. For the conventional filter to work efficiently, it is necessary to select the appropriate number of samples from which the minimum value is selected, while considering the tradeoff between the filtering effect and synchronization stability. This paper investigates an alternate filter, called an adaptive minimum filter, which controls the number of samples depending on the variation of the RTT values. This filter achieves a sufficient filtering effect as well as superior clock control stability. A computer simulation is used to evaluate the time synchronization protocol that employs the adaptive minimum filter. The simulation results show that the synchronization protocol with the adaptive minimum filter achieves better time accuracy and stability than that with the conventional minimum filter.