| Compared with the highly automation of CAD tools for digital circuits, analog circuits are often designed manually by experience using hand calculations and iterative numerical simulations, which leads to a longer design time for analog designs. This dissertation analysizes, develops and verifies a number of methods and procedures for the behavioral modeling and simulation of tree-like RLCK linear circuits, weakly nonlinear circuits and hardly nonlinear circuits using an improved and optimized symbolic data structure, determinant decision diagrams(DDD), motivated by its wide application in linear circuits.; All our descriptions for analog circuits are built on their linear-centric models. The transfer function based model with DDD representation is used to obtain the frequency responses of linear circuits, at the same time the s-domain symbolic expressions make it convenient for further order reductions by direct truncation and Hurwitz approximation, which is very suitable for the analysis of large linear circuits like tree-like interconnection problems. Also dynamic ordering algorithms are applied to improve the effectiveness of DDD representation for the semi-symbolic analysis of linear circuits. For nonlinear analog circuits, we convert them into the combination of a series of linear volterra circuits or a parameterized linear description, depending on if their nonlinearity is weak or strong. Specifically, the parameterized linear description is based on piecewise-linear modeling(PWL) and can reduce the dimension of the corresponding MNA matrix drastically when compared with the existing ideal diode model.; Based on these linear-centric descriptions of analog circuits, we propose a series of procedures for their simulation in time domain and frequency domain. For linear circuits, pole-zero relationship is combined with DDD to derive the transfer functions of coupled circuit structures; For nonlinear circuits, we apply both symbolic and numerical methods to make the simulation more efficient, in which procedure DDD is also exploited for deriving symbolic expressions to avoid repeated numerical calculations. The simulation results are compared with Spice to verify the effectiveness and efficiency of these procedures. |
