Thermally robust self-calibration and entropy loading for micro-LED links: a simulation study

Abstract High-speed visible-light communication (VLC) using gallium-nitride micro–light-emitting diodes (micro-LEDs) is sensitive to thermal variations that can degrade predistortion accuracy and link performance. This numerical study proposes and evaluates thermally robust adaptive compensation and entropy loading (TRACEL), a self-calibrating physical-layer method for direct-current–biased optical orthogonal frequency-division multiplexing (DCO-OFDM). Here, self-calibration means that the coefficients of the inverse are estimated online; the junction temperature is assumed to be known, and no hardware is calibrated. TRACEL employs complementary odd-symmetric probes that preserve average optical current, event-triggered recursive least squares to update a nine-term memory-polynomial inverse, subcarrier-level covariance projection for distortion reservation, and capped water filling with Maxwell–Boltzmann entropy selection under a fourth-moment peak-risk constraint. A 256-point DCO-OFDM link with 96 data subcarriers and 4-GS/s sampling is evaluated using a temperature-dependent micro-LED behavioral model whose bandwidth decreases from 1.62 to 1.21 GHz over 25–85 °C. Under a 25–85–25 °C thermal cycle, TRACEL achieves 6.902 Gbit/s net achievable information rate, 17.59 % error-vector magnitude (EVM), and 0.392 service-outage probability, compared with 6.507 Gbit/s, 21.00 %, and 0.522 for static joint loading (static predistortion with entropy loading). Under rapid thermal steps, TRACEL achieves 6.863 Gbit/s, an 8.19 % gain over static joint loading, while updating only 6.48 % of blocks and remaining within 1.46 % of an oracle. All results come from a behavioral-model simulation and are not hardware measurements.