Reducing nonlinear OFDM signal distortions using neural networks in the time domain

The temporal OFDM signal has a high PAPR (Peak to Average Power Ratio), often referenced as "the peak factor problem". This means that the signal has some peaks with a power much higher than the average power and as such is sensitive to the nonlinear characteristics of the HPA (high power amplifier). Neural networks are used to compensate the HPA nonlinear distortion effects in OFDM. This paper put the stress on the robustness of the presented neural networks in the time domain with multipaths channels simulations and with a variable number of carriers. The SSPA (solid state power amplifier) amplifier model and the additive Gaussian channel is used.