Predistortion for RF Power Amplifiers Under Varying Load Conditions with Shallow Learning Neural Networks and Charge Trapping Compensation for 5G and 6G Mobile Networks

Future mobile networks require highly linear radio frontends for different frequency ranges from cm to mm waves. To achieve the increasing requirements for data through-put and energy efficiency, radio frontends need semiconductor technologies which are capable for higher frequencies, cheap in production and very energy efficient. Therefore, Gallium Nitride (GaN) is more widely used in actual and future radios. Unfortunately, this semiconductor technology still suffers from nonlinearities, especially charge trapping effects which get worse for varying load conditions. Modulation schemes of more than 1024 QAM, which are desired for future 5G and especially 6G networks require an overall error vector magnitude (EVM) of less than 1 %. A novel digital predistortion system with shallow learning artificial neural networks which use sensor information to determine the intrinsic charge trapping state is proposed within this work.