A novel method for adjusting Digital predistortion with Modulation Error Ratio in OFDM System

Abstract

Power amplifiers (PAs) are important components in OFDM systems which can efficiently uses the spectrum but very sensitive to constellation error. The nonlinearity of PAs causes in-band distortion and spectral regrowth, thereby increasing the constellation error. To overcome the PA's nonlinearity, Digital predistortion is a highly cost-effective way to linearize PAs. When modeling of digital predistorter, there are many coefficients that determining the predistorter's characteristics which are affecting the effect of predistortion. Different values of the coefficients lead to a large gap in the effect of predistortion, thus constructing the predistorter with appropriate coefficients is complex and necessary. Many research when estimating the coefficients by minimizing mean square error (MMSE) of the signal amplified by the amplifier. But the MSE can't indicate the constellation error rate, furthermore, OFDM rely on the constellation map to transfer message, so that the coefficients corresponding to MMSE method may not the most appropriate. In this paper we propose a novel method to estimate the predistorter coefficients by maximizing the Modulation Error Ration (MER) which can directly indicate the degree of constellation error, and we introduce the quasi-Newton method which can help us find the optimum coefficient to maximize MER. The proposed predistorter can better reduce the constellation error.