一种基于贝叶斯推理的微波器件阻抗调谐算法  

Bayesian Inference Based Algorithm for RF Device Impedance Tuning

作　　者：王飞 苏江涛 蔡佳林 赵思源 刘军[1] WANG Fei;SU Jiang-tao;CAI Jia-lin;ZHAO Si-yuan;LIU Jun(School of Electronic Information,Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学电子信息学院,杭州310018

出　　处：《微波学报》2021年第2期60-65,共6页Journal of Microwaves

基　　金：国家自然科学基金(61827806,61871161);电子测试技术重点实验室开放基金(6142001190103)。

摘　　要：阻抗调谐匹配是设计和优化微波毫米波收发电路不可或缺的关键技术之一。为解决有源负载牵引稳定度差、调谐速度慢的问题,提出了一种利用基于贝叶斯推理的行为级模型进行阻抗调谐的算法。该技术可有效减少负载阻抗合成过程所需的迭代次数,提高阻抗合成精度,从而有效提升有源负载牵引技术的实用性和稳定性。为验证该算法的有效性,搭建了有源实时负载牵引测试系统,对氮化镓晶体管在5 GHz频段进行了负载牵引测试,实验结果证明,该方法具有良好精度,且与传统数值技术相比迭代次数减少了约30%,为射频前端电路的快速设计和优化提供了新思路。Impedance tuning matching is one of the key technologies for designing and optimizing microwave and millimeter wave transceiver circuits. In order to solve the problem of unstable and slow tuning speed of active load-pull method, a behavioral model based on Bayesian inference is proposed to synthesize the impedance. This technology can effectively reduce the number of iterations required for load impedance synthesis and can improve the impedance convergence accuracy of open-loop active load-pull, thereby effectively improving the practicability and stability of the active load-pull technology. In order to verify the effectiveness of this algorithm, a real-time load-pull test system was actually built to perform load-pull tests on GaN transistors at 5 GHz. Experimental results prove that this method has good accuracy, and compared with traditional numerical techniques, the number of iterations of this method is reduced by about 30%, which provides a new idea for the rapid design and optimization of RF front-end circuits.

关 键 词：负载牵引 贝叶斯理论 行为级模型 射频电路 阻抗调谐 

分 类 号：TN12[电子电信—物理电子学]