IEEE 1588 style synchronization over wireless link

Abstract

We present a technique for synchronizing clocks over a wireless link between a pair of resource constrained nodes. A Kalman filter is used as a pre-processor to a PI controller to mitigate the effects of packet-losses and attenuate noise spikes. This Kalman filter directly tracks the skew, which is the rate of change of offset, between a pair of nodes. The PI controller accepts this skew as input and disciplines the clock on the follower node. Experimental results demonstrate the performance of this technique over a single hop. In the future, this technique can be extended to multihop systems and improve determinism in networked embedded systems.

A Survey of Clock Synchronization Over Packet-Switched Networks

“…One technique to estimate the relative clock rate is the use of a Kalman filter [108], [109]. Kalman filters are widely used in telecommunications to estimate unknown variables given a set of observed measurements and random noise variables.…”

A servo design for slave clock in IEEE 1588 synchronization networks

“…In [7], a clock model was established based on a step function to describe the skew under the assumption that the clock drift over a synchronization interval is zero. Different from [7], we take oscillator jitter into consideration, which can be described by zero-mean and finite-variance random noise.…”

“…Different from [7], we take oscillator jitter into consideration, which can be described by zero-mean and finite-variance random noise. So we put forward a clock model based on Sine function and a random noise component that distributed with Gaussian distribution to generalize the model proposed in [7].…”

“…In [6], a PI controller was introduced in with a conventional filter to track the offset, but this approach is susceptible to packet losses. In [7], an improved method that tracked the skew instead of offset was proposed, by which small errors in skew estimating are not magnified during its conversion to offset. But the merely use of proportion and integration will make the rate of convergence slowly.…”

“…But the merely use of proportion and integration will make the rate of convergence slowly. Moreover, the Kalman filter in [7] lacks of external control during the process of states alternate.…”