Power Amplifier Linearization Using Orthogonal Polynomial Identified in Frequency Domain

Digital pre-distortion (DPD) is a common technology to counteract nonlinear characteristics of power amplifiers (PAs), where the memory polynomial (MP) model is widely used. Though authors previously proposed an MP model parameter calculation method in the frequency domain which mitigates numerical instability problem that resides in a conventional time domain calculation method with Least Squares (LS), it was not directly applicable to the identification of nonlinear characteristics from measured input/output data of the PAs. In this paper, we extend the previously proposed calculation method to employ orthogonal polynomials as the linear/nonlinear base functions of the MP model and to perform the calculation step-by-step with identification targets, from which lower order components identified in previous steps are subtracted. By an analysis of computational load and computer simulations, the extended frequency domain MP model parameter calculation method offers reasonable and acceptable performance in DPD with indirect learning architecture and requires less computational load than the conventional LS algorithm in the time domain.