LED communications with linear complexity compensation of dynamic nonlinear distortion

An in-depth study of the second order Volterra kernel of light emitting diodes (LEDs) is provided, with a special focus on compensation of dynamic nonlinear distortion generated by the LED when used as a communications signal transmitter. It is shown that from the factorization of the kernel in the frequency domain follows a simplified time-domain model for LED nonlinearity, consisting of three operations: two linear filtrations and squaring. Next, using the pth inverse theory, the corresponding block structures of nonlinear pre- and postdistorters are derived. As it turns out, they exhibit linear computational complexity. It is shown that the model coefficients defining the kernel may be estimated up to the Nyquist frequency. Numerical results indicate that pre- and postdistortion achieve the same performance with respect to receiver signal to noise power ratio (SNR) and have a considerable advantage over linear equalizers. However, in a practical scenario, the predistortion increases the peak to average power ratio of the transmitted signal (unless it is not high already), possibly leading to performance inferior to postdistortion. Finally, it is shown that the simplified equalizer based on the LED model has roughly identical performance to the regular, quadratic Volterra equalizer.