| As the complexity and performance criteria of analog and mixed-signal systems are steadily increased, behavioral modeling methods are commonly used by design engineers for rapid higher-level simulations. An innovative bottom-up approach for behavioral modeling of analog circuits was presented in the early 1990s. This modeling approach afforded a step-by-step procedure to generate accurate and simulation efficient behavioral models of the original circuits. While this method has the limit of modeling the nonlinear dynamic behavior properly, it was only applied to circuits with low or medium frequencies.; Nonlinear dynamics identification method, which identifies nodes in the large-signal time domain accounting for the nonlinear dynamic behavior of the circuit, is a key contribution of this research. By including nonlinear dynamic effects the model accuracy has an obvious improvement based on simulation results of some circuits. In this research, the algorithms are also improved and extended to model circuits at higher frequencies (with internal nodes) based on the previous approach. This is another contribution of this dissertation.; In the end, this method has been fully automated and a modeling tool, Ascend, has been implemented. Compared to the original circuits, the models generated have provided satisfactory speedup while maintaining high degrees of accuracy. |
