A Canonical Formulation of Volterra Series Regressors in Digital Predistortion

Digital predistortion (DPD) is a widely adopted technique to linearize power amplifiers (PAs) in modern wireless transmitters, mitigating nonlinear distortion while maintaining efficiency. Volterra-series-based behavioral models continue to be fundamental in DPD design, offering representation of nonlinear and memory effects. However, existing model structures lack a unified formalism that captures their common structural elements, which hinders systematic analysis and extension. This work proposes a canonical and parametrizable formulation of Volterra regressors that unifies widely adopted models within a single framework. The proposed approach enables the characterization of basis functions via a flexible product of delayed inputs, their conjugates, and envelopes. Experimental validation on a commercial high-efficiency PA using a 5G-NR signal over an instantaneous bandwidth of 200 MHz demonstrates improved model interpretability, sparsity, and enhanced linearization performance.