Theoretical and Experimental Studies of a Probabilistic-Based Memoryless PA Linearization Technique

This paper studies the performance of a memoryless power amplifier (PA) linearization technique based on a probabilistic approach. This technique employs a nonparametric method to derive a predistorter function, which does not need any parametric modeling and explicit parameter estimation. It only needs to calculate a probabilistic cumulative distribution function (CDF) and a quantile function (an inverse function of the CDF). Histogram and order statistic methods are proposed to perform the calculation. A rigorous analytic formula is derived for the inter-modulation product power (IMPP) of the PA output signal when a finite number of samples as well as a finite number of bins are used to calculate the CDF and the quantile function. The analytic results show that, with the probabilistic-based technique, the IMPP approaches zero as the number of samples approaches infinity and the bin width approaches zero. Computer simulations are utilized to verify the theoretical analysis and to compare the performance of the probabilistic-based linearization technique with those of other memoryless PA linearization techniques, while a prototype experiment is carried out to demonstrate its performance in a practical application. Results show that the technique can accurately determine the predistortion function that effectively compensates for the nonlinearity in the PA, and that it achieves a much better linearization performance compared to other existing methods, especially in the presence of a loop delay in the feedback circuit.