Study On Digital Predistortion Modeling Of Power Amplifiers

Since the era of mobile Internet access, broadband mobile communication equipments have become an indispensable part in everyone’s life. With these equipments, users can use social software to keep them in touch with the world around them, use news client to get latest information and news, enjoy the current good time through video and audio applications.These users’ needs above largely stimulate the increase of total amount of transmitted information, there is a desperate need to develop and utilize modulation technology, such as OFDM, which can make better use of the frequency resource, and then more information will be carried in limited spectrum resource. Such modulation technology of high spectrum efficiency produces non-constant envelop modulated signal, when this signal passes through the commonly used power amplifier(PA) in communication devices, the output signal is amplified and also nonlinearly distorted because of the nonlinearity of PA, which causes inter symbol interference(ISI) and spectrum spread, and seriously affect the process of signal transmission.For solving the above problem, linearization techniques have to be employed to improve the linearity of PA and eliminate the negative effects. Predistortion(PD) is an important branch in amplifier linearization technologies, with the development of digital signal processing(DSP), this technique is favored by the industry figures and obtained more and more attention gradually. This thesis mainly study the digital PD, the main work of the thesis is as follows:(1) The cause of power amplifier nonlinear distortion is analyzed from two aspects of harmonic distortion and Intermodulation distortion.(2) PA’s common behavioral models are introduced, listing memoryless PA’s and memory PA’s behavioral models respectively, the introduction includes application situations of each model.(3) The corresponding PD technique is applied with respect to each PA’s behavioral model, and the identification method of parameter in predistortion technique is analyzed. For example, in look-up table(LUT) PD of memoryless PD techniques, polar LUT and complex gain PD are given iterative formula to get LUT entries, in the meantime an another non-iterative approach to initialize and update the LUT entries is elaborated in detail; in memory PD techniques, the solving process of least square(LS) method is provided to identify the parameters in memory polynomial.(4) Two types of adaptive learning architecture are introduced, and the principles of two kinds of adaptive algorithms are expounded. In this thesis we propose a predistorter structure diagram, then we make simulation research on both memoryless polynomial and memory polynomial PD, by the use of power spectrum, adjacent channel power ratio(ACPR) and model accuracy indices: normalized mean square error(NMSE) and error vector magnitude(EVM), we finally analyze the simulation results and demonstrate the effectiveness of PD.