Font Size: a A A

Research On Digital Predistortion Of Power Amplifiers Based On Neural Network

Posted on:2015-03-31Degree:MasterType:Thesis
Country:ChinaCandidate:H E XunFull Text:PDF
GTID:2308330473950826Subject:Circuits and Systems
Abstract/Summary:PDF Full Text Request
The application of broadband modulation signals demands for a higher requirement for the linearity of the power amplifier. With the rapid development of digital processing devices,the adaptive digital predistortion(DPD) technology has became the mainstream technology in the RF amplifier linearization technology. Firstly,the paper analyzes the nonlinear characteristic and memory effects of the power amplifier, and then completes the simulation of the behavioral model and digital predistortion technology based on Back Propagation(BP) neural network. Compared to memory polynomial, BP neural network has a definite advantage on the modeling of the power amplifier with strong nonlinearity. The simulation result shows that the improvement of ACPR can be up to 21.19 dB with the digital predistortion system based on BP neural network for the 33 dBm power amplifier.In the systems of WCDMA or WiMAX,the high PAPR of signals will bring adverse effect on the performance of DPD. This paper generates 20 MHz WiMAX singal and 5MHz WCDMA signal and uses the cascaded pulse cancellation method to clip the peak signal of the above signals. The PAPR of WCDMA is decreased to 8.2dB from 11.4dB and the PAPR of WiMAX is decreased to 7.2dB from 11.2dB by this method. Compared with peak windowing and noise shaping,this method consumes less logic units and the performance of suppression of PAPR is better.The way of parallel processing of the BP neural network is well suited for the implementation in FPGA. This paper implements a BP neural network predistorter that has memory depth of 3 and the numbers of neurons of the first hidden layer and second hidden layer are 7 and 3 respectively. In addition,this paper completes the time delay estimation and adjustment of the DPD system based on the DSP evaluation board of the ADI company.Finally,the off-line test platform is built to verify the performance of DPD system. Test result showed that the best ACPR improvement is up to 15.78 dB for the WCDMA signal.
Keywords/Search Tags:digital predistortion, BP neural network, CFR, time delay estimation
PDF Full Text Request
Related items