Node Synchronization in a Wireless Sensor Network Using Unreliable GPS Signals

This paper presents our findings in using pulse measurements from a jittery one pulse per second (pps) global positioning system (GPS) clock, to synchronize the real-time clock (RTC) in each node of a wireless sensor network, when the timing jitter is subject to a empirically determined bimodal non-Gaussian distribution. Specifically, we 1) estimate the RTC phase and align it with an estimate of the true time phase, 2) calibrate the frequency of a 19.2 MHz low-cost temperature compensated crystal oscillator (TCXO) that drives the one pps RTC, and 3) track and compensate TCXO frequency variations due to environmental and aging effects. In our GPS driven synchronization methodology we adopt a statistical signal processing framework to estimate the 2% percentile in the bimodal distribution, perform a long-term frequency calibration with fractional frequency adjustment, and track the changes in the TCXO frequency to within three tick per second over a nominal 19.2 MHz frequency with an adjustment made every four hours.