Digital Predistortion of Power Amplifiers Based on Spline Approximations of the Amplifier Characteristics

Digital block-based predistortion (PD) is an efficient scheme to linearize power amplifiers (PA). The performance of this method highly depends on the estimation accuracy of the amplifier characteristics. In this paper, the PA characteristics are approximated by piecewise continuous polynomials in the form of splines. The approximation is based on a block of observed data at the input and output of the PA. This approach leads to a linear parametric approximation problem with moderate implementation complexity, and it also gives the best possible fit to a given set of observations. With cubic splines it is possible to get a fairly accurate estimate even with a short block of data, and hence a fast convergence of the PD algorithm. The performance of the method has been evaluated for an example selected from an application in mobile satellite communications with burst transmission. We use a model of a 2 Watts class A PA that is based on measurements, and the calculations are performed for 16-QAM modulation. The mean square approximation error, the power spectrum and the error vector magnitude (EVM) of the received signal have been analyzed by simulations. For 6 dB back-off, the adjacent channel power ratio (ACPR) level is reduced by approximately 15 dB relative to a transmitter without predistortion, after an adaptation based on only 25 transmitted symbols.