Abstract The digital compensation process for nonlinearly inertial characteristic of the transmit path of dual-band communication system with concurrent data transmission in the presence of unbalance of modulator quadratures has been considered. Analytical relationships enabling us to identify adaptively or change the parameters of dual-band polynomial model of corrector with due regard for the modulator quadrature distortions caused by the LMS, RLS and conjugate gradient algorithms are derived. A model of corrector based on the neural network representing a multilayer perceptron was built for the system. Experimental comparative analysis of the linearization efficiency of link with 25 W power amplifier using correctors with polynomial and neural network architecture is carried out. A comparative analysis of convergence speed, computational complexity and the linearization efficiency of adaptive algorithms (LMS, RLS and conjugate gradient algorithm) based on the polynomial architecture has been performed. The signals using 16QAM modulation with bandwidth of 4 MHz and frequency tune-out of 16 MHz are used as test signals. The results of experimental analysis show the highest computational efficiency without any loss of linearization quality of polynomial compensator that is identified by the conjugate gradient algorithm.

中文翻译：

---

正交失真下双频数据传输系统的自适应数字校正器

摘要 研究了存在调制器正交不平衡的同时传输数据的双频通信系统传输路径非线性惯性特性的数字补偿过程。分析关系使我们能够在适当考虑由 LMS、RLS 和共轭梯度算法引起的调制器正交失真的情况下，自适应地识别或更改校正器的双频带多项式模型的参数。为该系统建立了基于代表多层感知器的神经网络的校正器模型。使用具有多项式和神经网络架构的校正器，对具有 25 W 功率放大器的链路的线性化效率进行了实验比较分析。收敛速度对比分析，已执行基于多项式架构的自适应算法（LMS、RLS 和共轭梯度算法）的计算复杂度和线性化效率。使用带宽为 4 MHz 和频率调谐为 16 MHz 的 16QAM 调制的信号用作测试信号。实验分析结果表明，共轭梯度算法识别的多项式补偿器的线性化质量没有任何损失，计算效率最高。