Accurate modeling noise characteristic of microwave field-effect transistor

In this paper, a new method based support vector regression (SVR) is introduced for modeling noise parameters of FETs. Support vector machines (SVMs), which are based on the structural risk minimization (SRM) principle, have powerful generalization ability, and can handle problems with finite samples . By using the proposed method, the effect of the measurement errors can be excellent treated with. As a result, the noise characteristics of FETs can be accurately predicted with a few measurements. The SVR based modeling method about bias and frequency dependences of noise parameters is developed. An AlGaN/GaN HEMT is used as an example to demonstrate this method. Experimental results are given out to validate the proposed methods.