Nonlinear Characteristics And Correction Technology Of Power Amplifiers

As one of the most important modules in the transmitter, the power amplifier (PA) is widely used in the wireless communication systems, such as the high frequency (HF) PA, ultra high frequency (UHF) PA and microwave PA. Its inherent nonlinearities induce serious nonlinear distortion. The distortion not only causes adjacent and in-band channel interference which cannot be removed by filters but also degrades the signal to noise ratio (SNR). Furthermore, with the increasing demands for large capacity and fast rate of the wireless communication systems, the non-constant envelope modulation signal has become the most widely used wireless communication modulation. And the modulation degree becomes higher and higher, such as 64QAM,256QAM and 1024QAM, the linearity of PAs is required more and more stringently by the high degree of modulation. Therefore, the PAs need to be linearized urgently.Firstly, the nonlinear characteristics of PAs are analyzed, and then the nonlinear correction methods and typical behavioral models for the PAs are discussed and analyzed in this paper. Secondly, in chapter â…¢, the nonlinearities of the HF PA are studied and the look-up table (LUT), augmented Hammerstein (AH) and memory polynomial (MP) behavioral models are employed to model the HF PA respectively. The comparisons of modeling performance among the three models and the linearization capability of digital predistorters which are built based on the three models are also discussed in this chapter. Comparison results show that the MP predistorter can improve the third order inter modulation distortions of the HF PA over 12dB, which is better than the AH and LUT predistorters. In addition, the nonlinearities and their correction are analyzed and experimentally verified in chapter III by a microwave backhaul system transmitter. The digital predistorter used to linearize the transmitter is an LUT predistorter which is established on the exponentially weighted moving average (EWMA) algorithm. The experimental results prove that a simple LUT predistorter can make microwave transmitter inter-modulation distortion improve over 15dB. Finally, in order to reduce the computational complexity of the existing models as well as to build accurate behavioral models for the PAs of the 4G and Beyond 4G mobile communication systems, this paper proposed two new behavioral models for wideband PAs in chapter IV and V. One of the two models is Generalized Two Box Cascaded model (GTBC) and the other one is Generalized Augmented Hammerstein (GAH) which is an extension of the existing AH model. The characteristics of the proposed models all include the lagging memory effects and leading memory effects in the dynamic nonlinear module which is used to simulate the memory effects of the PA. After the lagging and leading terms being added, the modeling accuracy of the models is significantly improved, and linearization capability of the digital predistorter based on the two models is enhanced gravely.In order to verify the GTBC and GAH models in this paper, the normalized mean square error (NMSE), Floating Operations (FLOPs) as well as the memory effect intensity (MEI) are employed to compare the modeling accuracy, computational complexity and the simulation capability for memory effects separately. The validation results show that the modeling accuracy and linearization capability of the GTBC model are higher than the traditional LUT, AH and EH models. The modeling accuracy and linearization capability of the GTBC model are close to the GMP model. However, the computational complexity of the GTBC model is only one-third of the GMP model. The modeling accuracy and nonlinear compensation capability of the GAH model are much better than the traditional LUT, AH, MP and FMP models. The modeling accuracy and linearization capability of the GAH model are weaker than GMP model, but its lower computational complexity has huge advantage.In summary, nonlinearities of PAs operating on different frequency bands can be corrected with the digital pre-distortion technology efficiently. However, due to the fact that the nonlinear characteristics of different PAs are distinct, the nonlinear characteristics of the PA must be carefully studied and analyzed before the behavioral model is selected.