<ul><li>Digital Predistortion (DPD) is widely used to improve signal quality in wideband RF power amplifiers.</li><li>DeltaDPD is introduced to address energy efficiency challenges faced by DPD, utilizing dynamic temporal sparsity in recurrent neural networks.</li><li>DeltaDPD reduces computational complexity, arithmetic operations, and memory accesses while maintaining good linearization performance.</li><li>In testing with a 3.5 GHz GaN Doherty RF PA, DeltaDPD achieved significant improvements in ACPR, NMSE, and EVM metrics with a 1.8X reduction in estimated inference power.</li></ul>

DeltaDPD: Exploiting Dynamic Temporal Sparsity in Recurrent Neural Networks for Energy-Efficient Wideband Digital Predistortion

Discover more