Some results in the application of polyspectra to certain nonlinear communication systems

Polyspectral techniques are applied to model certain types of nonlinear systems frequently encountered in satellite communications. The power amplifier in such systems is usually operated close to saturation for best efficiency. Current practice is to model the amplifier as a memoryless nonlinearity. Such a model is inadequate if the transponder bandwidth is large. The RF sandwich model is considered. The identification of this model in the case of real signals has been done before. However, in communication systems, the signal is bandpass in nature, and AM/AM as well as AM/PM effects are observed. It is shown that a complex representation of the signal may be used to handle this case and extend the results reported by Brillinger (1977). In addition, by a further characterization of the sandwich model, the authors obtain its Weiner kernels and an extension of Bussgang's theorem.<>