稀疏的归一化功放模型及预失真应用

Sparse normalized modeling of power amplifier and its predistortion application

针对射频功放的非线性特性进行了研究,提出一种新的稀疏化的Volterra级数模型。该模型基于压缩感知算法,将稀疏系统的辨识等效为信号的重构问题,利用正则正交匹配(ROMP)算法对核系数进行稀疏化并选择出活跃的核系数。将提出的模型与记忆多项式(MP)模型、通用记忆多项式(GMP)模型进行比较,较MP模型的建模精度提升10.7 dB,模型系数减少25%;较GMP模型的建模精度提升3.9 dB,模型系数减少57.65%。仿真结果表明,提出的方法实现了良好的预失真线性化性能,极大地降低了模型系数,优于传统的功放行为模型,由此验证对功放的线性化技术发展具有参考价值。

Aiming at the nonlinear characteristic of RF power amplifier, this paper presented a new sparse Volterra series model. The model was based on the compressed sensing algorithm, the identification of the sparse system was equivalent to the signal reconstruction problem, and the kernel coefficient was sparse, used the regular orthogonal matching （ROMP） algorithm to select the active kernel coefficient. Compared with the memory polynomial（MP） model and the general memory polynomial （GMP） model,the proposed model improved the modeling accuracy by 10.7 dB and model coefficient decreased by 25% compared with MP model and it improved the modeling accuracy of 3.9 dB, but the model coefficient decreased by 57.65% compared with the GMP model. The simulation results show that the proposed technique predistorters achieve similar linearization performance while requiting significantly less coefficients than the traditional models. This verification of power amplifier linear technology development has the reference value.