Adaptive Filtering Algorithm in Clock Calibration System

Clock bias is a vital element in clock calibration system. Aiming at effectively eliminating noises contained in clock bias and improving accuracy of time synchronization, we propose an adaptive algorithm. In this adaptive algorithm, ensemble empirical mode decomposition (EEMD) is employed to decompose original data to different intrinsic mode functions (IMFs). For accurately abandoning IMFs which contain plentiful noises, principal components analysis (PCA) is proposed. Eigenvalues of all IMFs are ascertained on the bias of PCA. Then, bias between adjacent eigenvalues is used to evaluate level of noises contained in each IMF. Finally, residual IMFs are used to reconstruct the signal. Clean signal with adding noises is used to evaluate this improved algorithm. Consequence indicates that this improved algorithm is effective. We design a clock calibration experiment, where the time signal is transferred through a  wireless channel. The adaptive algorithm is used to dispose bias, which is acquired in clock calibration experiment. Consequence also demonstrates that this algorithm can effectively and adaptively eliminate noises contained in actual data.