Passive Intermodulation Cancellation in 5G Systems Using Artificial Neural Networks

Passive intermodulation (PIM) has been a serious challenge in 5G Frequency Division Duplexing (FDD) carrier aggregated (CA) wireless systems leading to the degradation of system performance. Digital cancellation techniques have been used to dynamically cancel the time-varying PIM resulting from passive nonlinearities. These techniques are usually based on Volterra-like behavioral models which are very complex and hard to implement. In this paper, a Feedforward Neural Network (FFNN)-based PIM cancellation scheme is proposed for PIM cancellation in a CA FDD wireless system. Simulation of the proposed scheme shows that the FFNN cancellation scheme is capable of acheving above 20-dB PIM cancellation ratio over a 30-dB input power range.