硅基毫米波共面波导可缩放神经网络模型  

Silicon-Based Millimeter-Wave Coplanar Waveguide Scalable Model Based on Atificial Neural Network

作　　者：程知群[1] 靳立伟[1] 张齐军[2] 

机构地区：[1]杭州电子科技大学射频电路与系统教育部重点实验室,杭州310018 [2]卡尔顿大学电子工程系

出　　处：《微波学报》2012年第1期1-4,共4页Journal of Microwaves

基　　金：国家973项目(2010CB327400);浙江省自然科学基金重点项目(Z1110937)

摘　　要：随着微电子工艺技术的发展,硅基CMOS器件的截止频率已经达到毫米波频段,使硅基微波单片集成电路实现成为可能。因此,建立硅基毫米波频段共面波导结构模型使准确设计硅基微波单片集成电路成为必要。文章提出了一种基于神经网络技术的共面波导结构(CPW)毫米波可缩放模型,采用3层神经网络结构,根据共面波导的测试结果,用神经网络来学习其物理变量和测试的相应S参数空间映射关系。仿真与测试结果比较表明:基于神经网络方法建立的毫米波共面波导可缩放模型对不同几何参数CPW能够快速和准确地给出对应的CPW的S参数结果。With the develepment of microelectronics technology, the cut-off frequency of the silicon-based CMOS devices has reached the millimeter-wave band. It makes it possible to realize silicon-based microwave monolithic integrated circiuts. Therefore, it becomes necessary to establish the model of silicon-based millimeter-wave coplanar waveguide for accurate design of silicon microwave monolithic integrated circuits. Silicon-based millimeter-wave coplanar waveguide （CPW） scalable model based on neural network technique is proposed in this paper. A three-layers neural network structure is used. Neural network is adopted to learn the mapping between the geometrical variables and S parameter of the coplanar waveguide from measured results of CPW. Comparison of simulation and measurement results shows that CPW scalable models based on neural network can provide accurate and fast prediction of the S parameters of CPW for differential physical sizes as variables.

关 键 词：硅基 毫米波 共面波导 神经网络 可缩放模型 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TN814[自动化与计算机技术—控制科学与工程]