双/单通道传导电磁干扰噪声盲源分离研究

Study on Blind Source Separation of Dual/Single-channel Conducted EMI Noise

传导电磁干扰差共模噪声分离中，硬件分离网络结构复杂、成本较高，传统软件分离受环境噪声影响明显，测量同步性差。针对上述局限，提出一种基于盲源分离理论的传导电磁干扰噪声差共模分离策略。首先，利用快速独立分量分析对L线、N线噪声实施双通道盲源分离，实验证明其比射频电流叠加法信噪比范围更宽，适用性更强；其次，单通道盲源分离采用连续小波变换实施虚拟通道扩展以克服同步测量误差，同时改进快速独立分量分析目标函数以提升收敛性能。实验证明，其分离性能取决于小波函数的选取而非扩展通道数量，性能相比双通道盲源分离有一定降低，但节省经济成本，具有较好应用前景。

Aiming to the complexity and high-cost of hardware-based separation network in conducted electromagnetic interference noise, as well as the poor anti-noise ability and measuring synchronism of traditional software-based separation approach, an EMI difference mode and common mode noise extracted strategy based on blind source separation theory is presented. Firstly, the fast independent component analysis algorithm is used for live and null line noise extraction, with a dual-channel BSS approach. Experiments demonstrate a wider signal-to-noise ratio range than RF current approaches, meaning stronger applicability. Then, the single-channel BSS approach extends extra virtual channel using continuous wavelet transform algorithm. Tests prove the separation performance depends on the choice of wavelet function, rather than the number of extended channels. Single-channel BSS approach saves economic cost and possesses wide prospect, with a little less separation ability.