Research On Microwave Nonlinear Circuit Based On Neural Network

With the development of electronic technology, microwave circuits have been widely used in communications, radar, electronic countermeasures, instruments and other fields. How to accurately design and analysis of microwave circuits becomes a very crucial part in these applications. Traditional design and analysis methods of these circuits are based on circuits in weak nonlinear circumstance. But for strong nonlinear circuits, these methods are not suitable. Therefore, it is urgent to put forward a new theory to solve this problem.This paper presents the traditional description of nonlinear microwave circuit theory, as well as Agilent's large signal modeling method. Based on small signal S parameters, we proposed a new theory of nonlinear microwave circuit—nonlinear scattering function. Furthermore, we have studied its physics meaning and characteristic. At last, we have built microwave test system of nonlinear scattering function and have got the data of nonlinear semiconductor device from this system.A black-box model based on neural network could be a good approximation of nonlinear mapping in theory. Therefore, we have built the model of nonlinear semiconductor circuit by using wavelet neural network that based on improved BP algorithm and we have got reasonable results. Finally, we have designed a microwave power amplifier by using extracted data from the test system. It has verified the accuracy of this new theory.