Comparative Analysis of Single-Box and Two-Box RF Power Amplifiers’ Behavioral Models Sensitivity to Delay Misliagnment

摘要

This paper examines the sensitivity of power amplifiers behavioral models to time-delay misalignment. Memory polynomial and twin-nonlinear two-box models are derived, from measured data, to model the behavior of a GaN based Doherty power amplifier under various time misalignment conditions. The results show that the memory polynomial model is much less sensitive to delay misalignment than twin-nonlinear two-box models. In particular, the memory polynomial model accuracy is not altered by a delay underestimation up to one sample, and is only degraded when the delay is overestimated. Conversely, twin-nonlinear two-box models performances are degraded whether the delay is underestimated or overestimated. Hence, the identification of twin-nonlinear two-box models unavoidably requires accurate delay alignment with a sub-sample resolution while the memory polynomial model can be derived from low complexity coarse delay alignment. Thus, in comparison with the two-box models, the complexity of the memory polynomial model identification can be offset by the use of low-complexity coarse time-delay alignment algorithms.