Digital Predistorter with Real-Valued Feedback Employing Forward Model Estimation

Digital predistorters (DPD) are used in modern communication systems to linearise nonlinear power amplifiers (PA) and maximise power efficiency. For their function, a feedback signal from the PA output is required. A conventional DPD uses a quadrature mixer and two analogue-to-digital converters (ADC) which consume additional power and increase system complexity. In this paper we have proposed an innovative technique which allows to use a nonquadrature RF mixer with one ADC in the feedback path. The DPD adaptation is noniterative and based on favoured indirect learning architecture. Firstly, the forward PA model is estimated and subsequently it is used to train DPD coefficients. We have verified and compared the proposed method with other DPD architectures in simulations. The results show that the proposed architecture can achieve the same results as a DPD with complex feedback samples and the other real-valued feedback architectures.