Dynamic extended Hammerstein model of RF power amplifiers for digital predistortion

In this paper, an extended Hammerstein architecture for dynamic behavioral modeling of GaN power amplifier (PA) is presented. Static strong nonlinearities, long-term linear memory effects and first-order nonlinear dynamics are considered in the model. The parameters of the proposed model are estimated by least square algorithm. For experimental identification and validation, a 20MHz four-carrier orthogonal frequency division multiplexing (OFDM) signal is used as the excitation of a 2.5GHz GaN PA. Test results show the proposed model has a better fitting accuracy of the PA's behavior when comparing to the conventional Hammerstein model. For digital predistortion (DPD) application, an over 15dB improvement of adjacent channel leakage ratio (ACLR) is achieved with the proposed inverse model works as the digital predistorter.