| In order to make full use of the spectrum resources,5G communication and low earth orbit communication satellites all use a higher peak to average power ratio modulation strategy to improve channel capacity and communication rate.Faced with higher and higher PAPR modulation signals,the working efficiency of the power supply unit decreases continuously,resulting in a large amount of energy waste.The output voltage of the ET power supply changes according to the change of the envelope of the RF input signal,so that the power amplifier always works in a high efficiency state,and then improves the conversion efficiency of the whole system.However,due to the lack of cooperation with the communication unit,the efficiency optimization of the traditional envelope power architecture in hardware has tended to be limited.In view of this bottleneck,this paper introduces the neural network and communication modulation technology into the generation of ET power control signal,and uses neural networks to obtain the amplitude and variation law of the envelope signal.Compared with the traditional ET technology,the generation of the control reference signal of the switching converter can be accelerated,the time delay of generating the control reference signal is reduced,and the computing resources of the hardware are saved.Firstly,the principle of communication modulation technology is studied,and 64 QAM and OFDM modulation technology are simulated and designed based on MATLAB platform.Secondly,the envelope extraction method of radio frequency signal was studied,and four kinds of envelope extraction methods were analyzed.Hilbert transform was selected as the method of radio frequency signal envelope extraction,and the radio frequency signal envelope extraction simulation of64 QAM and OFDM system was carried out respectively,and the envelope data of two modulation modes were obtained.At the same time,the principle of BP neural network is studied,and a BP neural network model is designed for the prediction of radio frequency signal envelope.Taking64 QAM which is a single carrier modulation as an example,the radio frequency signal envelope is predicted and simulated,and the influence of the parameter of hidden layer on the network error is compared in the simulation.Finally,aiming at the envelope prediction of OFDM radio frequency signal formed by superposition of multiple QAM modulation subcarriers,the selection method of input and output data of neural network was changed,the influence of different number of hidden layer nodes on the prediction effect was studied,and the error of envelope prediction under different number of subcarriers and different mapping methods in OFDM system was analyzed.Different mapping methods under 12 subcarriers were selected to compare the prediction effects,the flow of generating control signal of switching converter from envelope is introduced and the effect diagram of control reference signal is obtained by simulation.BP neural network is used to predict and simulate the envelope of RF power amplifier based on MATLAB platform.The experiment proved that: The BP neural network with single hidden layer and 16 hidden layer neurons under 64 QAM had the best prediction effect,and the RMSE and MAE between the predicted and actual envelope amplitude were 0.2741 and 0.3219.The prediction of RF signal envelope of OFDM system also has the best effect when the number of neurons in the hidden layer is 16.The proposed prediction scheme is valid under different subcarrier numbers and multiple QAM mapping modes.Under 12 carrier channels,multiple mapping modes are used to predict the envelope,and the maximum RMSE value obtained is 0.3097.The maximum MAE value is 0.4995.In conclusion,the error between the envelope predicted by BP neural network and the actual envelope modulated in 64 QAM and OFDM systems is within the acceptable range,so that the ET power supply can supply power to the RF power amplifier stably,which verifies the feasibility of the proposed algorithm for envelope prediction. |
PDF Full Text Request