A frequency-shift-filtering approach to cyclostationary noise modeling in MIMO NB-PLC

In this paper, we propose a cyclostationary noise model for multi-input multi-output (MIMO) narrowband power line communication (NB-PLC) based on frequency-shift (FRESH) filtering. The MIMO FRESH filters are designed to shape a multi-input white noise spectrum to a muti-output cyclic spectrum extracted from experimental noise measurements of three-phase low-voltage power lines. The noise modeling problem is formulated as a system identification problem that minimizes the time-averaged mean square error (TA-MSE) between a reference measured noise waveform and the model-generated noise waveform. We propose the normalized mean-square-error (NMSE) for both the cyclic auto-correlation and cyclic cross-correlation between the measured and the generated noise waveforms as a performance measure. The noise waveform generated by the proposed model shows a good agreement with that of the measured noise in terms of the cyclic auto/cross-correlation NMSE. Moreover, the proposed model provides a scalable performance-complexity tradeoff.