| RF Power Amplifier(RFPA) is one of the most important and expensive component in wireless communication system, and also the main nonlinear source. However, it will produce memory effect during modern broadband transmission, which seriously impact on the performance of communication system. When the input power is less then PA will work on the linearity region with low efficiency. When PA works on saturation region, the power efficiency will be improved but the generated nonlinearity will bring interference between adjacent channels which result in signal distortion in band.In order to obtain high efficiency and excellent linearity, on the one hand, we can change PA's structure to increase its efficiency; on the other hand, we can accomplish PA's linearity using linearity techniques. ET (Envelope-tracking technology), LINC and Doherty techniques are the most common ones for now. But the digital predistortion, as the most promising linearity technique, which puts high emphasis on understanding PA's nonlinearity and memory effect to exactly simulate digital predistortion system with predistortion behavior models. This paper aims to do research on the nonlinearity, behavior models and different techniques for improving the PA efficiency of wideband PA for the digital predistortion system simulation.The PA nonlinearity part main includes the cause of nonlinearity characteristic, nonlinearity characteristic parameters and distortion production. The behavior models of PA part will discuss the analysis and comparison between the classic behavior models and simplified behavior models. Moreover, the digital predistortion system part introduces different predistortion methods for different efficiency improvement techniques, and do structure and simulation verification on digital predistortion system in MATLAB environment.This project does research on digital predistortion system based on a lot of references. The main work has been done as follows:(1) Analyze the nonlinearity and its measurement criteria of PA, and then discuss the features of several common memoryless models and parameter identification.(2) Deeply analyze the causes and effects brought by PA's memory effect, after that, we introduce some memory models, among them, Volterra serial model is the best classical behavior model which shows nonlinearity exactly. But the calculation is so complex. The other behavior models are the simplified form of Volterra serial model.(3) Present adaptive algorithms including LMS and RLS. Beyond that, LMS-Newton algorithm, which is evolved from LMS, it can improve the convergence speed when the input signals have higher correlation.(4) In the different solutions for RFPA, Doherty PA is the best fit for improving the PA efficiency in wireless communication system with non-envelop modulating signal. This paper presents a new predistortion method based on Doherty technique, through analyzing the algorithm transfer threshold in this method which mixed LMS-RLS algorithm, it eventually put forward a improved algorithm transfer threshold.(5) ET PA is to improve efficiency using changing the voltage of small signal, there exist different nonlinearity characteristic at different power in PA. This paper propose a new predistortion method of ET PA after analyzing sub-section predistortion principle and effect on even factorial of behavior models. This method comes up with adding even factorial component in traditional memory polynomial model to better improve the nonlinearity in system.(6) The two new predistortion methods proposed by this paper are theory testified its feasibility by simulation. |
PDF Full Text Request