Research Of Clock Synchronization Over WIFI Networks

There are many technologies for clock synchronization, the traditional one is deploying GPS at the base station, but it cost a lot and is not safe. In pace with rapid development of WIFI technologies and applications, WIFI becomes the main wireless access network and an important complement to 3G networks. As a result, the demand for Clock synchronization becomes urgent over WIFI networks. But WIFI is not used for real-time business currently. Consequently, it can not meet the demand for station synchronization. Therefore, researches on the technologies over WIFI networks have important and practical significance.Clock synchronization includes syntonization and phase synchronization, and many synchronization standards have been formulated. IEEE 802.1as is setted by IEEE standard organization in 2011, and it has the best performance to rise euipment's synchronization precision. In addition, it not only simplifies the equipment model in IEEE 1588v2 but also expands to support 802.11 link and MoCA, and this paper focus on the 802.11 link.The core mechanism of IEEE 802.las is the timestamping. The accuracy of the protocol implementation depends mainly on the accuracy of the timestamps used for synchronization. The closer to the real sending or receiving time, the more precise can be reached. In this paper the timestamp is done below the MAC layer, which is called linux driver scheme. The adventage of the scheme is that it ignores the jitter from the above protocol stack and does not need additional hardware. Then we use Matlab to simulate the algorihms. First of all, we set the clock and the jitter model, which is induced by MAC firmware, physical layer, and asymmetric wireless link. Then use average filter, Kalman filter and LMS filter to eliminate delay jitter before offset compensation. Based on comparation, we chose Kalman filter as the pretreatment module at last, which has the best tracking performance.This paper proposes two compensation methods, one is called linear incremental compensation, and the other is called PI control continuous adjustment. The simulation results show that, the method combines Kalman filter and PI control can achieve a better synchronization precision. Due to adjusting the clock offset and the skew both, the performance of clock synchronization dones't depend on the synchronization interval. Besides the algorithm inproves the synchronization stability, it is easy to be put into practice, and cost a little.