Digital Predistortion of Radio Frequency Power Amplifiers with Flexible Spectrum Operation

To maximize the utilization of the overcrowded frequency spectrum, future wireless systems must support flexible spectrum operations that coordinate the utilization of spectrum resources. This introduces great challenges to the linearity performance of wireless transmitters, as the system may impose multimetric linearity requirements where different frequency regions may have distinct linearity requirements. Additionally, transmitted signals may exhibit different spectral characteristics. DPD is a widely adopted linearization technique for compensating the nonlinear distortions introduced by RF PA. However, existing DPD techniques often struggle under these flexible operation scenarios. This thesis presents several novel DPD techniques targeting the linearization tasks in flexible spectrum operation scenarios. For the multimetric linearity requirements, a suboptimal multimetric model extraction algorithm is proposed to efficiently meet the requirements with low-complexity progression steps. A Pareto-optimal model extraction method is further developed to achieve optimal tradeoffs among multiple linearity metrics. It also achieves linearization performance breakthrough in interested frequency regions and significant reduction of kernel functions for certain multimetric requirements. For linearizing the transmitted signals with different spectral characteristics, a low-complexity DPD technique is proposed, which significantly reduces both the diversity complexity of kernel functions and the number of coefficients. The proposed methods are thoroughly validated through substantial experimental tests conducted on different types of PA, various test cases, and signals with different bandwidths. The results provide solid justification for the effectiveness of the proposed methods and demonstrate their potential in future wireless systems.