ITR: Fast Approximate Simulation Methods for Rapid Design and Test of Complex Electrical Systems

Georgia Tech Research Corporation, Atlanta GA

Investigators

Abstract

With the high levels of system-level integration for electrical systems available today, tools for simulation and test of complex mixed-technology systems are urgently needed. Circuit simulators such as Spice are not adequate for simulating large circuits due to large run times. Even behavioral level simulation is too slow for the large numbers of simulation runs needed to do simple things such as design centering and fault simulation. The key thrust of this research is to develop new simulation algorithms that significantly speed up the process of design and test of complex mixed-signal/mixed-technology systems. The algorithms will quickly assess the merits of mixed-technology system design perturbations. The key observation is that the impact of design perturbations on the relative performance of the designs can be assessed accurately by approximate simulation methods, rather than by expensive exact simulation. These techniques are expected to speed up evaluation of design transformations for optimization of complex systems and test optimization by three orders of magnitude. The project has three specific goals. First, it will develop techniques for fast time and frequency domain approximate numerical simulation of mixed-technology systems. Second, it will develop algorithms for fast time and frequency domain approximate spatial simulation of mixed-technology systems by partitioning the system into blocks that can be simulated under "open-loop" conditions. Third, it will develop techniques for fast time and frequency domain approximate numerical and spatial simulation. The project will accomplish this by combining the first two techniques.