On the optimum multicarrier performance with memoryless nonlinearities

One of the major problems associated with multicarrier signals is their high envelope fluctuations that lead to implementation difficulties. Therefore, multicarrier signals are usually submitted to nonlinear operations such as clipping to simplify the power amplification, DAC (Digital-to-Analog Conversion) and other digital signal processing operations. However, these nonlinear operations lead to nonlinear noise that usually is associated with performance degradation. Recently, it was shown that the nonlinear distortion associated to bandpass memoryless nonlinear devices not only does not necessarily means performance degradation, but can even lead to performance gains. In this paper we study the impact of non-bandpass memoryless nonlinearities on the performance of multicarrier signals. We consider both nonlinear devices operating on the real-valued multicarrier signals as well as Cartesian nonlinear devices (also denoted as I-Q nonlinearities) operating separately on the real and imaginary parts of the complex envelope of multicarrier signals. We present a theoretical analysis of the potential performance improvements associated with the optimum detection of multicarrier signals, by deriving accurate theoretical expressions for both the average asymptotic gains in ideal AWGN (Additive White Gaussian Noise) channels and the asymptotic gain distribution for frequency-selective channels with multipath Rayleigh fading.