Channel Identification and Predistorter Design Using Stochastic Gradient Method

We present the identification and compensation method for non-linear systems with memory which arise due to the transmit/receive filters and the non-linearity of a power amplifier. Such a non-linear system can be modeled as a linear system, a memory-less nonlinear system, and a linear system cascaded sequentially. We obtain the filter coefficients of the linear systems and the coefficients of a polynomial approximating the memory-less non-linear system through the stochastic gradient method(SGM). We compensate for a non-linear communication system that can be modeled as above with a predistorter using the same method SGM and the indirect learning architecture. The non-linear compensation method presented here does not need a specific modeling scheme and can be applied adaptively.