LMS Algorithm for Blind Adaptive Nonlinear Compensation

Abstract

This paper presents low-complexity blind adaptive nonlinear compensation algorithms for bandlimited signals. The new algorithms utilize highpass filtering to extract the out-ofband signal energy caused by nonlinear distortion. A leastmean-square (LMS) algorithm and its normalized version are derived based on minimization of the square of the extracted out-of-band signal without access to the original input signal or prior knowledge of the nonlinearity. In this sense the developed algorithms are "blind" and only require prior knowledge of the signal bandwidth. Unlike the P th-order power series inverse, the proposed nonlinear compensation method is not affected adversely by large input amplitudes. The effectiveness of the online algorithms is illustrated with several simulation examples.

Blind identification and real-time calibration of memory nonlinearity based on RLS algorithm

“…Important classical approaches of digital post calibration include 1) acquiring distortion target by employing the inverse of the original nonlinearities with the help of extra circuits or components [1]; 2) applying Volterra series to model the nonlinear system with the prior knowledge of the nonlinear distortion [2]- [4]; 3) Using a blind identification technique to estimate the parameters of the model for the inverse nonlinearity [5]- [7]. Memoryless nonlinear system has been well modeled by using the power series expansion [6].…”