Balancing noise and nonlinear distortion by power loading in FBMC OWC

Nonlinear distortion (ND), after intersymbol interference, is the second most significant data rate-limiting factor in LED-based optical wireless communication systems, particularly when using multicarrier modulation formats. Dealing with ND typically requires either power back-off or the application of computationally exhaustive nonlinear equalizers (e.g., Volterra, artificial neural network). In this paper, we demonstrate theoretically and experimentally that in multicarrier modulations, it is possible to mitigate ND by combining subcarrier power loading with either bit or entropy loading (EL). By doing so, ND is suppressed without any additional computational load. We consider the following power loading strategies: uniform power loading, pre-emphasis (application of an inverse channel filter at the transmitter), and the newly, to the best of our knowledge, proposed adaptive nonlinear power loading algorithm (NPL). Formal analysis carried out in a behavioral model of nonlinear LED reveals that uniform loading excels at low input powers and pre-emphasis at the highest. However, for intermediate power levels, neither of the above is optimal, and it is our NPL algorithm that achieves the best performance. We demonstrate that the NPL can be combined with conventional bit and power loading, as well as entropy loading algorithms. To achieve even higher data rates than in orthogonal frequency division multiplexing, we apply offset quadrature amplitude modulated filter bank multicarrier. The experimental results show an improvement of 11% in terms of asymptotic information rates over conventional EL and a gain of 17% compared to the well-known Levin-Campello algorithm.