Support vector regression-based dynamic behavioral modeling for RF power amplifiers

A new dynamic behavioral modelling technique, based on support vector regression (SVR), is presented in this paper. As an advanced machine learning technique, the SVR method provides an effective option for behavioral modeling of radio frequency (RF) power amplifiers (PAs) including the nonlinear and memory effects. In this paper, the basic theory of the proposed SVR modelling method is provided, along with details on model implementation of RF PAs. Experimental test example for a LDMOS PA is presented, revealing that the new modelling methodology provides a very efficient and extremely accurate prediction.