Digital predistortion of OFDM signals using lookup tables for amplifier linearization.

摘要

Orthogonal frequency division multiplexing (OFDM) has been widely adopted in newer communications systems due to its high spectral efficiency and resistance to multipath fading and impulse noise. However, because of its large peak-to-average power ratios (PAPRs), OFDM exhibits a serious disadvantage in its sensitivity to nonlinear distortion. In wireless systems, where in order to achieve good power efficiency high power amplifiers (HPAs) operate near the saturation point, OFDM signals lead to spectral regrowth, resulting in unacceptably high levels of adjacent channel interference (ACI). To mitigate these effects, signal predistortion techniques are commonly employed at the input to the HPA. In basic nonlinearity compensation algorithms, the HPAs are considered to be memoryless, which means that the current output depends only on the current input. One such digital predistortion scheme investigated in this project involves multiplying the signal, based on its level, by the value stored in a one-dimensional (1-D) lookup table (LUT), so that the overall gain of the predistorter and HPA cascade is independent of the level of the incoming signal. The values in the LUT are adjusted in an adaptive manner based on the feedback from the output of the amplifier. In the present research, for memoryless HPAs, predistorters using non-uniformly spaced 1-D LUTs and variable step sizes in adaptive implementations are investigated, in order to improve the spectral performance of conventional LUT predistorters.; In systems with wider bandwidths, the memory effects of the HPA cannot be ignored. In this research, improved predistortion based on two-dimensional (2-D) LUTs is pursued, which is capable of linearizing amplifiers with memory effects, where the amplifier gain depends mainly on the short-term average power. Specifically, building on the experience gained from operating 1-D LUT predistorters, in this project the algorithm compensating for HPA nonlinearity with memory switches between different LUTs corresponding to different past power levels of the signal. This approach for compensating for nonlinearities with memory is designed for HPAs with the behavior observed in practical implementations. As documented in this research, the proposed 2-D LUT predistortion method requires the matching of some parameters such as switching thresholds to the amplifier model and the development of new LUT adaptation strategies. For transmitters with HPAs exhibiting nonlinear memory effects, this research demonstrates significant improvements in the suppression of ACI when using predistorters based on 2-D LUTs as opposed to using simpler 1-D LUTs.