DDR Volterra series behavioral model with fading memory and dynamics for high power infrastructure amplifiers

This paper presents an innovative technique of adapting the Dynamic Deviation Reduction (DDR) Volterra series behavioral model to describe very accurately high power RF amplifiers (PAs) targeting cellular infrastructure applications. A coefficient pruning concept is introduced with the goal of reducing model complexity and the number of series coefficients, while still maintaining excellent model fidelity. This is accomplished by reformulating the DDR Volterra with respect to the discrete time domain to allow independently specifying memory depth for each kernel. DDR coefficients are then identified by a pruning method of forcing memory depth to fade with increasing kernel order while constraining dynamics. The validity of the pruning concept is demonstrated for two high power Doherty infrastructure PAs exhibiting significant memory and non-linearity. Exceptional model fidelity is achieved while the total number of series coefficients is dramatically reduced compared to more traditional Volterra descriptions.