A New Microwave Device Modeling Method Based On Time Delay Neuro-Space Mapping

As we all know, we need a mass of microwave device model to advance the microwave circuit computer aided design technique. The high quality model can help people design an accurate circuit, although it is large, complex or harsh to apply. With the development of semiconductor material and technique, a series of new devices come out(i.e., GaN HEMT). The existing model can not represent the electrical characteristics of these devices exactly. Thus, we need new device model.Generally, verify and update the circuit repeatedly is the main idea of the traditional device modeling method, which need a lot of time and experience. Using the traditional device modeling method, we can build a coarse model which expending little time. The Neural Network Space Mapping(Neuro-SM) method accurately and automatically maps the gap between the coarse model and the fine model under the learning ability of the Artificial Neural Network(ANN). Based on the Neuro-SM Method, the efficient high quality model will be established with less training data.In recent years, the microwave devices always work on large power and high frequency conditions. It is no doubt that there have some dynamic effects in the device and hard to be ignored(i.e., capacitor effects). This dissertation combines Time Delay Neural Network(TDNN) and Neuro-SM method for proposing an advanced microwave device modeling method based on Time Delay Neural Network Space Mapping. This method will apply on those devices which have capacitor effects. In the end, we compare the traditional method, Neruo-SM method and TDNN SM method with a fictitious microwave device model. The results illustrate that the proposed method is better than the other two methods in the devices with capacitor effects.