Model-based design and implementation of an adaptive digital predistortion filter

Abstract

Dataflow models of computation are widely used for modeling signal processing systems. These models have inherent concurrency and the task (actor) execution depends only on the availability of the input data (tokens). This property of dataflow models can be exploited for dynamic power management by automatically switching off the actors with no available input tokens. This idea is applied in this paper for efficient modeling and implementation of an adaptive Digital Predistortion (DPD) filter. The DPD filter is required to operate with different profiles under varying operation scenarios, hence requiring a methodology to manage power dynamically. The paper presents a dataflow model for Adaptive Digital Predistortion based on the Core Functional Dataflow (CFDF) model of computation using the Light Weight Dataflow (LWDF) programming methodology. The paper also provides a methodology for dynamic power management under the dataflow paradigm. To the authors' best knowledge, this work is the first to integrate dataflow-based power management systematically in the context of adaptive DPD implementation.