Pruned Volterra-Based Distortion Mitigation of High-Speed GaN Micro-LEDs for Energy-Efficient Optical Communication

High-speed micro light-emitting diodes (LED) are attractive candidates for optical communications (OC) due to their wide bandwidth and energy efficiency. However, their strong nonlinear distortion severely degrades system performance. In this work, we present the first adaptation of the generalized memory polynomial model to micro-LED-based OC and propose a second distortion mitigation approach based on a novel pruned Volterra model. Training is performed using digital postdistortion, to be used as a predistorter for practical deployment. Experimental validation on galium nitride micro-LEDs driven by direct current-biased optical orthogonal frequency division multiplexing shows significant improvements, including a 71% throughput increase from 4.32 Gb/s to 7.35 Gb/s. These results demonstrate the strong potential of these compensation schemes to unlock energy-efficient, high-speed micro-LED optical links.