基于数据缺失感知动态Allan方差的时钟源稳定性评估方法

Clock Source Stability Evaluation Based on Missing-Data-Aware Dynamic Allan Variance

摘要:

针对工业物联网、通信网络和分布式测量系统等多节点系统应用环境中出现的频率扰动、相位漂移及数据缺失导致本地时钟源时变稳定性难以可靠评估的问题，对缺失数据条件下的动态Allan方差计算进行了研究；采用双环形振荡器拍频观测结构，将被测时钟相对参考时钟的频率偏差和相位漂移转换为低频计数序列；提出有效Allan三元组筛选与窗口有效性判据相结合的缺失感知计算方法，以真实观测样本构成的三元组计算局部统计量；采用FPGA实测序列和构造序列开展对比实验。结果表明：该方法减少了缺失块邻域的局部虚假波动，相较块均值重标定方法，突发噪声和周期纹波对比度均有所提高；该方法可为缺失数据条件下时钟源同步状态监测、质量评估和异常数据分析提供高可靠性的参考。

Abstract:

To address the difficulty in reliably evaluating the time-varying stability of local clock sources caused by frequency disturbances, phase drift, and missing data in multi-node application environments, such as the Industrial Internet of Things, communication networks, and distributed measurement systems, this paper studies the computation of dynamic Allan variance under missing-data conditions; a beat-frequency observation structure based on two ring oscillators is adopted to convert the frequency deviation and phase drift of the clock under test relative to a reference clock into a low-frequency count sequence; a missing-data-aware computation method combining valid Allan triplet screening with a window validity criterion is proposed to calculate local statistics using triplets composed of actual observations; comparative experiments were conducted using sequences measured on a field-programmable gate array and synthetic sequences, and the results showed that the method reduced local spurious fluctuations around missing-data blocks and improved the contrasts of both burst noise and periodic ripple compared with the block-mean rescaling method; the method can provide a highly reliable reference for monitoring clock-source synchronization states, assessing clock-source quality, and analyzing anomalous data under missing-data conditions.