Characterization of nonlinear channels in OFDM systems using signals with spectral notches

In this paper, we consider the problem of characterizing nonlinear channels in real baseband OFDM communication systems. A novel method is proposed to estimate the frequency-domain Volterra kernels of nonlinear channels using OFDM signals with spectral notches. In this method, the OFDM signal is arranged in a way that each full-spectral OFDM symbol frame is followed by copies of the same OFDM symbol frame but each has a spectral notch at a distinct frequency. We derive the mathematical expression of the difference signal between the outputs of a full-spectral OFDM symbol frame and its corresponding OFDM symbol frame with a spectral notch. We find that, through the use of OFDM symbol frames with spectral notches, the task of Volterra kernel estimation can be decomposed into independent subtasks with smaller numbers of kernel coefficients. Therefore, the estimation of the Volterra kernels in the subtasks can be accomplished with more accuracy and less computational complexity using the same amount of available input and output data. We show that, the proposed method has a computational complexity of O(M⁴), while the conventional method would require a computational complexity of O(M⁶). This indicates that a great saving in computation can be achieved by the proposed method without sacrificing the performance. The outperformance of the proposed method over the conventional method is justified by computer simulations. Real baseband OFDM systems, such as powerline and ADSL communications, can benefit from the proposed method.