Supporting analog synthesis by abstracting circuit behavior using a modeling methodology

For analog and mixed-signal circuit design a modeling methodology is needed which is well suited to the requirements of a structured synthesis flow. It ensures that the intended specification is met without over-design. Behavioral models are able to provide circuit level knowledge on higher abstraction levels by including non-idealities and parasitics from the circuit realization, which allows realistic comparison of different architectures. At the same time these models are fast enough to explore design space and optimize specification translation with respect to system performance measures. This paper shows how numerical, analytical, and symbolic methods as part of an integrated modeling methodology are used for model generation, model parameterization, and model refinement in order to improve and accelerate analog circuit synthesis.