Research on RF signal envelope prediction based on BP neural network

The envelope tracking power supply greatly improves the efficiency of power amplifier compared with the traditional constant voltage power supply. However, Envelope Tracking (ET) technology requires high hardware conditions and produces high extra delay. In this paper, a data prediction scheme of RF signal envelope based on BP neural network is proposed, which is conducive to generating the reference signal of control switch converter in advance. Base band data generates envelope data of RF signal by Orthogonal Frequency Division Multiplexing (OFDM) modulation technology, which is used to train the prediction model. Finally, the network is trained under different subcarrier numbers and mapping modes to obtain and the feasibility of the model is verified after a better number of nodes is selected. The results show that the method proposed in this paper can achieve accurate envelope data prediction with the maximum Root Mean Squared Error (RMSE) value of 0.1752. Moreover, the predicted envelope fits well with the actual envelope, which verifies the feasibility of the proposed model. By calculating the number of floating point operations on the BP neural network prediction model, and the saving rate of the calculation times can reach 49.40%.