Generalized twin-nonlinear two-box digital predistorter for GaN based LTE Doherty power amplifiers with strong memory effects

In this paper, a generalized twin-nonlinear two-box predistorter is proposed for the linearization of highly nonlinear Doherty power amplifiers exhibiting strong memory effects. The proposed predistorter is made of the cascade of a memoryless look-up table followed by a generalized memory polynomial function and thus can be seen as a two-box implementation of the generalized memory polynomial model. The generalized twin-nonlinear two-box predistorter is experimentally benchmarked against the generalized memory polynomial model. The linearization performances of both models when applied on a GaN based Doherty power amplifier driven by a 20MHz LTE signal, demonstrate the superiority of the proposed predistorter which achieves better linearity performance while requiring a lower number of coefficients. Indeed, an extra 5dB is obtained in the ACLR while the number of predistorter coefficients is reduced by more than 60%.