Compensation for nonlinear high power amplifier using data predistorter with variable step-size LMS algorithm

Abstract

When an M-ary Quadrature Amplitude Modulation (QAM) is employed to increase data rate in satellite or microwave communication system, the system performance is limited by the nonlinearity of the High Power Amplifier (HPA) caused by nonlinear AM/AM and AM/PM effect due to the multi-level envelope of 16-QAM waveform. The nonlinearity of the HPA causes Adjacent Channel Interference (ACI) and Bit Error Rate (BER) degradation. However, it can be reduced by backing-off the signal power level, resulting in the reduction of the effective signal power in transmitter, which ends up reducing data rate. Therefore it is necessary to compensate the nonlinearity of the HPA and an adaptive predistortion technique that can compensate for the nonlinear distortion of a HPA in 16 QAM system is proposed. The proposed predistorter is based on a fixed type polynomial predistorter and an adaptive type predistorter. The fixed type polynomial prdistorter is used for roughly compensating the nonlinearity of the HPA nonlinearity and the adaptive type predistorter is used for compensating residual nonlinearity of the HPA after doing the fixed type polynomial predistorter. The fixed type polynomial predistorter is modeled by inverse transformation of the HPA model such as AM/AM and AM/PM characteristics, respectively. Also inverse function of the HPA using even order polynomial is simplified by least square curve fitting method and the phase mapper is truncated by using Taylor series for complexity saving. The adaptive type predistorter is based on Variable Step-size Least Square Mean (VSSLMS) algorithm method that is modified version of simple LMS algorithm. In general, LMS algorithm is very widely used for its simple implementation but it need to modify LMS algorithm due to slow convergence speed. Then VSSLMS algorithm is used in the adaptive type predistorter, which its step-size is variable by reflecting error signal power and we improved the convergence speed of adaptive type predist...