Search for a command to run...
Abstract A time scale serves as a more stable, uniform, and reliable time reference compared to a single atomic clock. However, conventional time scale algorithms are typically based on a single noise model assumption and focus primarily on the dominant noise characteristics of specific atomic clock types. While such algorithms can improve frequency stability within limited intervals, they often fail to simultaneously meet both short-term and long-term stability requirements. To address this challenge, this paper proposes a time scale fusion method based on Complementary Multi-Scale Wavelet Transform (CMWT). The proposed method begins with wavelet functions exhibiting favorable sparsity and compactness. It integrates feature decomposition via multi-scale wavelet transform, selectively fuses wavelet coefficients based on individual time scale advantages, and constructs complementary structures through normalized wavelet variance weighting. The fused time scale is then obtained through wavelet reconstruction of these complementary structures. Experimental results demonstrate that the fused time scale generated by the proposed method exhibits superior short-term and long-term stability compared to individual time scales involved in the fusion process. Specifically, compared with individual time scales ALGOS-like time scale (TA1) and AT1 time scale (TA2), the fusion result with the highest hourly stability has its stability improved by 57.8% and 20.12% respectively. The fusion result with the lowest hourly stability has its stability improved by 57.42% and 19.4% respectively compared with TA1 and TA2. Compared with TA1 and TA2, the fusion result with the highest 128-hour stability has its stability improved by 29.83% and 39.92% respectively. The fusion result with the lowest 128-hour stability has its stability improved by 22.07% and 32.16% respectively compared with TA1 and TA2, thereby fulfilling the requirement of enhanced both short-term and long-term stability.