Nonlinear preprocessors, including clipping, blanking, and joint blanking/clipping have been proposed to be employed in orthogonal frequency-division multiplexing systems for mitigating the impulsive noise in power-line communications. In this paper, we propose using a new variant of nonlinear preprocessor, namely, deep clipping, to mitigate the impulsive noise. In contrast to clipping, blanking, and joint blanking/clipping, where one or two thresholds is used to detect if the impulsive noise occurs, and if the impulsive noise occurs how to handle it, deep clipping is characterized by two alternative parameters, namely, a threshold and a depth factor. By optimizing the two parameters, we find that deep clipping achieves better signal-to-noise ratio (SNR) performance than clipping, blanking, and joint blanking/clipping. In addition, we present a mathematical expression for deep clipping and show that the SNR expressions regarding conventional clipping, blanking, and joint blanking/clipping can be easily obtained by using our analytical methods.