| Nowadays the transmission information technology of modern communication systems is developing rapidly in the direction of multi-carrier, multi-level, broadband and high PAPR (Peak Average Power Ratio), The new broadband digital transmission technology (such as OFDM, MC-CDMA and WCDMA, etc.) and modulation methods of high spectral efficiency (such as QPSK and M-QAM, etc.) is obtaining more and more widely used. However, these efficient digital modulations transmission technology of multi-carrier and multi-channel based on non-constant envelope have high linearity requirements for RF power amplifiers, the signal amplified by non-linear RF power amplifiers will bring out a large number of inter-modulation distortion, result in expansion of the spectrum to output signal of power amplifier, bring adjacent-channel interference, but also result in signal distortion in the band, increase the BER(Bit Error Ratio). At a large number of linear technologies, the base-band pre-distortion technique does not involve the difficulty of the RF signal processing, just does compensate treatment to the base-band signal in the low-frequency part, this approach facilitate the use of modern digital signal processing technology. It has a strong ability of adaptation and high precision, in favor of the implementation of hardware, such as DSP and FPGA, it also agrees with the idea of Software Radio, so the development of base-band pre-distortion technology will have important significance.This paper is base on the basic principles of pre-distortion, discuss and research base-band adaptive pre-distortion technology from two aspects of memoryless and memory according to the characteristics of RF power amplifier model. The article first discusses the nonlinear characteristics and parameter of power amplifier, as well as several commonly used linear technology, and focus on analyzing the principle of pre-distortion technology and system architecture, proved the similarity of pre-distortion and post-distortion, provide a theoretical foundation of the adaptive structure for the follow-up. For the memoryless pre-distortion, we firstly verify LUT pre-distortion technology in the form of polar with the LMS adaptive algorithm used, work out LMS algorithm of complex coefficients, and apply in the look-up table technology, compare the differences between the two through the simulation and structure, we found that the structure of complex coefficients is relatively simple, just one-dimensional look-up table storing the index data, but its convergence rate is relatively slow. The pre-distortion system after adjust the contents of LUT can be regarded as a more simply and convenience program of pre-distortion. In addition, based on the Traveling Wave Tube Amplifiers Saleh model, used of non-directly study structure, respectively, with polar coordinates and complex coefficient polynomial, we do adaptive pre-distortion, and work out complex coefficient vector iteration formula of adaptive LMS, RLS algorithm which is commonly used, by comparing these two, find that polynomial with complex coefficients of the pre-distortion method, although the iterative formula of plural is relatively cumbersome, but it decreases the conversion of R / P and P / R, has greatly reduced the calculation amount, and the result of pre-distortion is effective.Furthermore, colligating existence defects of LUT and polynomial pre-distortion, we bring out a new pre-distortion method which based on the amplifier with AM / AM and AM / PM characteristics that are the combination of look-up table and polynomial structure. As the ceiling of normalized saturated output, in two steps: first, by a group of well pre-designed data, obtain the phase shift corresponding data and stock in look-up table ROM; second set up a simple fifth-order polynomial to AM correction of nonlinear amplitude distortion. Only after a few data to adaptive adjustment by LMS algorithm, we can quickly get a good pre-distortion effect and its computation reduce nearly a half relative to the structure of polynomial in polar coordinates.On the other hand, if consider power amplifier of the memory effect, the impact of historical data can not be ignored. We discussed several commonly used memory amplifiers model, in accordance with the Wiener amplifier model which relevant references had mentioned, do pre-distortion processing by memoryless and memory polynomial using RLS convergence algorithm, and we found that the signal constellation diagram of the former exist proliferation, while the latter is more desirable effect of the constellation diagram; in the calculation of volumes, the former contain a total of four iterations of the polynomial coefficients without historical data, while the latter's memory depth is 2, non-linear coefficient is 3, a total of nine iterations of the polynomial coefficients. On this basis introduce another commonly used filtering algorithm - Kalman filter, by simulating used this approach, observe power spectral density map, we find this approach improve channel interference to be better compared to the RLS.In view of the defect of large size calculation of abovementioned memory polynomial pre-distortion, using a large number of ROM memory cell, the LUT + FIR cascade method for pre-distortion is brought forward, both two adopt the structure of direct studying, index with amplitude of current input signal after quantization, respectively, does update processing to the contents of the LUT and FIR filter coefficient, the pre-distortion effects have been ideal. Second, we attempt to replace LUT with polynomial to correct non-linear, that is, poly + FIR cascade model pre-distortion, its convergence is faster. Finally, from the theory, we derived that for linear systems of the power amplifier, it is necessary to make the total loss of the system be optimal, we should not blindly rely on pre-distortion device to offset the non-linearity, we should selecting the appropriate point, that is appropriate input power backoff and the reduction of the peak back, to allow the system to be maximize efficiency.About another important aspect of pre-distortion linearization system, that is, the estimation of delay factor for comparing two sets of signals, at the last of the article, introduce two compensation methods for delay factor briefly, that is correlation coefficient of loop delay compensation and time delay compensation of Fourier transform theoretically, give out two formulas for calculating the delay factor. This paper summarizes and analyzes a variety of structures and pre-distortion algorithm, and simulate for analysis to compare the similarities and differences between them, the complexity of calculating amount of pre-distortion system, and provide a theoretical basis for practical use, as well as studying structure, the basic for algorithm selection. It has an important significance. |
PDF Full Text Request