Enhanced identification technique for dynamic nonlinear power amplifiers modeling using twin-nonlinear two-box structures

This paper presents an enhanced identification technique for the behavioral modeling of radiofrequency power amplifiers using twin-nonlinear two-box structures. The proposed technique intentionally uses low nonlinearity order in the memoryless block and allows for performance degradation in the first box of the model. This helps reducing the coefficients dispersion of the second box, and therefore improves the numerical stability of the identification process. The model's accuracy is then recovered by adding the memory polynomial function of the twin-nonlinear two-box model. Experimental validation of the proposed technique shows a model size reduction of approximately 50% without any noticeable performance degradation.