Quasi-Optimum Detection of Nonlinear OFDM: Performance Bounds and Receiver Design

Nonlinearities are usually something to be avoided in orthogonal frequency division multiplexing (OFDM) transceivers. However, it has been demonstrated that nonlinear (NL) OFDM can outperform linear OFDM, since NL distortion has useful information on the transmitted signals that can be exploited by the optimum maximum likelihood receiver. The optimum ML receiver for NL OFDM is too complex and it is not easy to obtain its performance, although some theoretical performance bounds are available for scenarios with a large number of subcarriers and signal-to-noise ratio (SNR). However, the optimum performance of NL OFDM in small-to-moderate SNR values is still not well studied. In this paper, we consider the optimum detection of NL OFDM schemes, where we take advantage of the NL distortion to improve performance. We present a detailed study of the contributions to the signal associated with a given subcarrier from the distortion received at the other subcarriers. This allowed us to obtain a bound on the maximum likelihood (ML) performance of NL OFDM that is valid over a wide range of SNR values subcarriers and demonstrates that the performance gains observed in the asymptotic SNR region also take place for lower SNRs. Additionally, we propose a practical, iterative decision-directed receiver that exhibits large performance gains over linear OFDM in both uncoded and coded scenarios.