Input Nyquist sampling suffices to identify and compensate nonlinear systems

Abstract

Nonlinear systems usually cause spectral spreading resulting in an output signal bandwidth greater than that of the input signal. When identifying and compensating nonlinear systems by digital methods, it has been common practice to set the sampling frequency at the Nyquist rate of the output signal. This correspondence presents experimental evidence for the author's proposal that nonlinear systems can be identified and compensated at the lowest possible sampling rate: the Nyquist rate of the input signal.