Research And Implementation Of Digital Predistortion In Broadband Transmitter

Broadband transmitter has become an inevitable trend along with evolution of the communication system.The bandwidth of baseband signal increases exponentially,which makes the nonlinear distortion of power amplifier in transmitter more serious.Simultaneously,more complex modulation methods are also used in the transmitter,which affects the performance of the transmitter due to high peak-to-average ratio.In this dissertation,the key technologies of digital pre-distortion are studied in detail in the background of broadband transmitter system,including crest factor reduction,predistortion model and down-sampling.The main work of this dissertation summarizes the following points:1.In this dissertation,the relationship between crest factor reduction and digital predistortion is analyzed firstly.Then,according to the inverse relation between the two technologies,the existing cascade structure of the crest factor reduction and digital predistortion are improved by introducing crest factor reduction parameter compensation module.The module can find the right peak-cutting factor to make the best pre-distortion effect.The simulation and test results show that the improved structure finds a suitable peak cutting factor to reduce the peak ratio to 8.25 dB when a 20 MHz LTE signal with a peak ratio of 10.52 dB is used to excite the continuous class amplifier,and the adjacent channel power ratio decreases by about 17 dBc compared with the non-pre-distortion,and reduces by about 2dBc compared with the existing structure.2.Based on the research of two existing neural network models,this dissertation analyzes the model of input signal processing and network construction method,and proposes a new neural network model on the basis of two models.The proposed model can achieve higher model accuracy in a short training time.The test results show that the proposed model decreases by about 1.2dBc and 2.8dBc relative to the FFNN model and the TADNN model in the adjacent channel power ratio.Then the proposed model is analyzed and the improved model is put forward.The improved model achieves high model accuracy under less model coefficients.The test results show that the improved model has better linearization ability than other models.3.In this dissertation,the existing down-sampling technology is studied,and the reconstruction model and alignment algorithm are improved on the basis of the down-sampling technology of the amplifier model reconstruction.The alignment algorithm proposed in this dissertation solves the problem of high complexity,and the simulation results show that with the reduction of sampling factors of 5,10 and 20 respectively,the improved alignment algorithm can save 30% running time under the condition of ensuring accuracy.At the same time,the improved amplifier reconstruction model is proposed by introducing fractional power into the model,and the simulation and test results show that under different down-sampling factors,the improved model not only expands the high modeling precision region,but also improves the modeling accuracy by about 1dB in a larger area.With the increase of the reduction sampling factor,the improved model has a greater advantage in linearization effect,and the adjacent channel power ratio is about 1dBc lower than that of other models.