GPS驯服中无偏滑动平均滤波算法的研究

Improved Moving Average Filter for GPS Disciplined

分析已有GPS驯服中滤波算法的特点,提出了基于GPS的实时频率误差处理及状态估计的无偏滑动平均滤波算法.该方法继承了普通滑动滤波算法低噪声特点,且用线性回归估计补偿了普通滑动滤波算法的偏差,利用该方法滤除频率测量误差中的频率偏差和多通道GPS接收机秒信号(GPS1PPS)的锯齿误差,并预报晶振状态.MAT-LAB仿真和实际测试结果都证明了无偏滑动滤波算法比普通滑动滤波有效,提高了晶振频率的长期稳定度和准确度,实际系统中恒温晶振OXCO-131的长期频率稳定度的Allan方差提高了约三个数量级,达到3.5E-12/d.

According to the features of filtering algorithm used in GPS, an improved disciplined method using Global Posi- tioning System one pulse pre-second （GPS1PPS） timing signal is proposed in this paper. This unbiased moving average filtering al- gorithm can eliminate sawtooth errors of GPS1PPS signal and produce an unbiased evaluation of local clock state by using linear re- gression estimate.Digital Phase-Locked Loop （DPLL） is used to make GPS1PPS timing signal tracked steady by local clock and combine the short term stability of crystal oscillator and long term stability of GPS1PPS timing signal, which ensure high frequency stability and accuracy of local frequency standard when GPS liming signal is lost. The results of MATLAB simulation and real measurement have proved that this method can improve long stability and accuracy of frequency, and Allan variance of frequency stabili- ty of crystal oscillator OXCO-131 pre-day in a real system has been improved 3 orders of magnitudes and reached 3.5E-12.