The Waveform Relaxation Method for Tjme Domain Analysis of Large Scale Integrated Circuits: Theory and Applications

The Waveform Relaxation (WR) method is a new decomposition method for solving a system of mixed implicit algebraic-differential equations over a given time interval. This method essentially uses an iterative relaxation scheme such as the Gauss-Seidel relaxation or the Gauss-Jacobi relaxation in whch the elements of the relaxation are waveforms of unknown variables. The decomposed system obtained through the relaxation scheme is also a system of mixed hp l i cit algebraic-differential equations but is much easier to solve than the original rystem. The application of this method in the area of time domain simulation of integrated circuits is emphasized. Although the WR method has a theoretical basis, it can be given a simple physical interpretation when applied to the analysis of integrated circuits. In particular, the convergence condtions of the method can be given either in terms of the numerical properties of the circuit equations or in terms of the properties of the circuit components. This method is shown to be a viable alternative to the conventional techniques for simulating large scale integrated circuits since sufficient conditions for its convergence are quite mild and are always satisfied by a large class of practical circuits. The performance of the WR method when applied to a particular class of circuits, Le., MOS dyital integrated circuits, are studied and evaluated via a prototype simulator called RELAX. The repetivity and directionality of dtgital subcircults as well as the digital nature of the signals are exploited in the simulator t o increase the speed of computation and to utilize the storage eaciently. Preliminary comparisons between RELAX and the standard circuit simulator SPICE have shown that RELAX is fast and reliable for simulating MOS digital integrated circuits.