New methods for time-domain modelling of RF/microwave passive structures and active devices

This thesis focuses on improving existing FDTD methods and time-domain modeling of active devices.; In this thesis, four methods are proposed to improve the efficiency of FDTD. First, a new compact two dimensional (2D) FDTD method is proposed to analyze the waveguide discontinuities along the direction of wave propagation, which is uniform in one of the transverse directions; it reduces the original three-dimensional (3D) problems to two-dimensional (2D) problems. Second, a compact one-dimensional (1D) FDTD formula is proposed for uniformly filled waveguide; it has the same numerical dispersion as the original 3D FDTD formula and can be used as an efficient incident wave generator and perfect absorbing boundary condition for the single mode. Then, a ID modal Perfectly Matched Layers (PML) is developed by applying the proposed ID FDTD to the traditional PML formula. Finally, a new 2D FDTD subgridding method is proposed that is not only very simple with controllable low reflections but also has been proven to be stable.; When the FDTD method is used to simulate RF/microwave circuits with active devices, the governing voltage-current equations of a device can be incorporated into the FDTD marching equations. However, the parameters of most electronic components are often given or measured in the frequency-domain and are band-limited. For instance, S-parameters of a field-effect transistor are usually given or obtained only in the frequency-domain and over a limited frequency range or band of interest.; Obtaining a causal time-domain model from the band-limited frequency-domain parameters is a challenge. In this thesis, three methods are proposed to solve this problem. The first two methods are iterative methods based on Fast Fourier Transform (FFT). One applies the FFT in combination with the error feedback principle and the second applies the Hilbert transform in conjunction with FFT. The third method uses the rational function fitting technique. The extracted time-domain parameters using the three methods not only are causal but also contain almost the same frequency-domain information as the original parameters over the given limited frequency range.