MS, University of Cincinnati, 2004, Engineering : Computer Engineering

Systems of Linear Time Invariant (LTI) equations occur often in the simulation of mixed-mode models. Solving these equations is one of the most time consuming operations. Speedup of simulation of mixed-mode models can be achieved if these LTI equation systems are solved in a shorter time. This thesis explores the applicability of analog computers to the solution of LTI equation systems. An analytical proof that an existing approach to solve LTI systems using analog circuits is a good solution strategy for LTI systems with large spectral radius, where iterative numerical algorithms are not applicable, is provided. A comparison between numerical methods and analog computer solutions is presented. The effect of practical issues like fabrication defects, noise and drift on the accuracy of the solutions to systems of LTI equations, computed by an analog circuit, is explored. The need for establishing stopping criteria is identified and two approaches of implementing circuits that satisfy stopping criteria are analyzed. Various issues involved in the use of analog circuits for the solution of LTI systems in an environment involving interaction between software and analog hardware are discussed. An algorithm to automatically generate an analog circuit, that can be efficiently configured on a reconfigurable analog platform, to solve a given system of linear algebraic equations, is presented and analyzed.

Committee: Dr. Harold Carter (Advisor) Subjects: