3D Monte Carlo Random-Walk for IC Parasitic Extraction

The patent replaces the traditional mesh-based mechanical/numerical system with a stochastic random-walk system. Instead of using finite element or boundary element methods that require space partitioning into mesh elements, the invention uses random walks of particles through the dielectric space to compute capacitance. This substitution allows handling of large-scale IC designs with millions of conductors while maintaining accuracy, as the random-walk approach has linear computational complexity rather than super-linear complexity.

The patent replaces the pattern-matching approach with a fundamental stochastic simulation method. Instead of approximating complex IC layouts by matching them to pre-solved patterns in a library, the invention directly simulates the physical random-walk process in the actual dielectric space. This eliminates the need for pattern libraries and approximations, providing both full-chip capacity and accurate extraction of coupling effects including fringing and interline coupling.

The patent implements a self-verifying extraction system where the stochastic random-walk method inherently provides accurate results for all nets simultaneously without requiring manual selection or verification. The random-walk simulation naturally captures all coupling effects for every conductor in the design, eliminating the need for manual identification of critical nets and subsequent verification steps. This self-service approach maintains accuracy while dramatically improving productivity for large-scale designs.