Macro model generation for synthesis of analog and mixed signal circuits [electronic resource] /

by Kanakipati, Sunder R.; Theses and, OhioLINK Electronic

Abstract (Summary)
Improved performance estimation methods hold the key for automated synthesis of analog and mixed signal circuits. Macro models provide one such fast and efficient method to estimate the performance charecteristics of a circuit. Many macro modelling methods such as table look up methods, nueral network based methods, convex geometric programming methods have been used so far. In this thesis, we build macro models using a table look up method and two basic numerical analysis techniques, the multiple regression and the multivariate cubic spline interpolation. Macro modelling can be viewed as a two step process. The first step is, the generation of charecterization data and the next step is, the generation of macro models from charecterization data. In the first step, an initial set of values of design variables for any given analog circuit is chosen. A simulation of the transistor netlist is done with the initial values and various parameters are calculated. These values are tabulated. This process is repeated till a user specified upper bound for each of the variables is reached. This accomplishes the generation of charecterization data. In the next step, the charecterization data is evaluated using multivariate numerical techniques and a macro model for each of the performance parameters is extracted. In an alternative method, the charecterization data is searched for a set of design values using a table look up method and the performance parameters corresponding to this set of design variables are returned. The evaluation time for a macro model is much shorter than the normal simulators such as HSPICE and SPECTRE. Also, the accuracy of macro models is almost the same as the simulators. Hence, these models can be easily incorporated in to any circuit synthesis algorithm or architecture generation algorithm to optimize the cost function based on performance constraints. Also, this thesis presents the application of regression models, spline models and the table look up based model in circuit synthesis and architecture generation for automated synthesis of various analog and mixed signal circuits.
Bibliographical Information:


School:University of Cincinnati

School Location:USA - Ohio

Source Type:Master's Thesis



Date of Publication:

© 2009 All Rights Reserved.