自适应数字预失真系统的Householder块精确逆QR分解递归最小二乘法

在非常数包络调制系统中，数字预失真技术已成功应用于线性化功率放大器. 文中提出一种适用于非线性功率放大器宽带数字预失真的增强记忆多项式模型. 与传统的记忆多项式和广义记忆多项式模型相比，该模型通过包含Volterra模型的交叉联合记忆项，在保持低复杂性的同时给出了优越的线性化性能. 预失真系统的参数估计使用Householder块精确逆QR分解递归最小二乘算法，通过计算机仿真研究了Householder块精确逆QRD-RLS算法的性能. 对强非线性AB类功率放大器进行了实验测量，并收集相应的复包络IQ基带数据，根据这些数据建立行为模型并验证预失真器性能. 结果表明，对瞬时带宽为5 MHz的WCDMA输入信号，在5 MHz频
率偏移处将ACPR改善了25 dB.

Householder Block Exact Inverse QR-Decomposition Recursive Least Squares Method in Adaptive Digital Predistortion System

Digital predistortion has been proposed to linearize the power amplifier (PA) for nonconstant envelope modulation schemes. In this paper, we propose an enhanced memory polynomial model for wideband digital predistortion of nonlinear power amplifiers. By including the cross-combination memory terms of
Volterra model, the proposed model is much less complex while maintains excellent linearity performance as compared to the conventional memory polynomial and generalized memory polynomial model. The adaptive parameter estimation of the predistortion system is realized using a householder block exact QR-decompositionbased recursive least squares (QRD-RLS) algorithm. Performance of the householder block exact inverse QRDRLS
algorithm is studied via computer simulations. Measurements are carried out for a highly nonlinear class-AB PA. The captured IQ baseband signal of the PA is used for behavioral model generation and predistorter performance validation. The results show that correction of an adjacent channel power ratio (ACPR) up to 25 dB is achieved at 5 MHz frequency offset for a WCDMA input having 5 MHz instantaneous bandwidth.