Regularization of the RLS algorithm for adaptive digital predistortion

Abstract

Problem statement. Modern digital pre-image input systems, as a rule, involve the use of adaptation algorithms to correct their parameters. Adaptation algorithms are solutions based on LMS and RLS methods. At the same time, the efficiency, in terms of the convergence rate and the accuracy of the solution, of RLS methods is significantly higher. This is due to the presence of a covariance matrix as part of the solution based on the RLS method. However, when using RLS algorithms (an adaptation algorithm using the RLS method), problems may arise related to poor conditionality of the covariance matrix resulting from signal noise, high inertia of the nonlinear properties of the power amplifier, limited accuracy of parameter calculation, etc. To solve this problem, it is advisable to use the regularization procedure of the RLS algorithm, taking into account the specifics of the digital predistortion systems. Objective. Creation and analysis of the regularization procedure of the RLS algorithm for adaptive input systems of digital presets. Results. The paper presents the implementation of the regularization procedure of the RLS algorithm for adaptive input systems of digital presets. Using the regularization procedure of the RLS algorithm avoids the state of poor conditionality of the covariance matrix, thereby ensuring stable operation of the digital pre-order input system. The state of the covariance matrix is monitored at each iteration of the RLS algorithm by evaluating the trace of the matrix. Practical importance. The proposed solution can be used to stabilize the parameters of adaptive systems for entering pre-orders based on RLS algorithms.