Empirical model generations of microwave devices exploiting linear regression models

In this paper, we propose two empirical model generation techniques of microwave devices for a large range of geometric and material parameters opportunely chosen. The empirical models are based on an initial inaccurate empirical model and linear-regression models. The linear-regression models compensate the error between the initial inaccurate empirical model and an electromagnetic simulator or measurement data. These techniques provide accurate empirical models computationally very efficient with respect to any electromagnetic technique. Moreover, these simple models could be easily integrated in a toolbox of any commercially available computed-aided design (CAD) tool for RF/microwave circuits. The two techniques have been successfully tested for two microwave tunable phase shifters.