Fast simulation of VLSI interconnects

Abstract

This paper introduces an efficient and accurate interconnect simulation technique. A new formulation for typical VLSI interconnect structures is proposed which, in addition to providing a compact set of modelling equations, also offers a potential for exploiting sparsity at the simulation level. Simulations show that our approach can achieve 50× improvement in computation time and memory over INDUCTWISE (which in turn has been shown to be 400× faster than SPICE) while preserving simulation accuracy.

Adaptive admittance-based conductor meshing for interconnect analysis

“…Second, we consider N conductors in parallel, each with a crosssection of 5 ¢ 1µm and length 1000µm, with an inter-conductor spacing of 1µm. As SPICE simulation is impractical, a fast simulator (from [2]) is used for the simulations. The center conductor is taken as the victim, and the aggressor inputs are ramp signals with a rise-time of 10ps.…”

Generalized method of the time-domain circuit simulation based on LIM with MNA formulation

“…Actually, the algorithm achieved 50 times faster than INDUCTWISE [7]. Roughly speaking, the condition (15) means…”

“…where b 2 is assumed to be zero in (7). Since the inverse inductance matrix is sparse, the linear equation (13) can be solved efficiently, which is capable of the fast interconnect analysis.…”

“…Currently, INDUCTWISE [6] and the RLP-algorithm [7] are known as the best choices of very large scale RLC interconnect analysis due to the computational efficiency with the acceptable accuracy. In this section, we discuss the advantage of the proposed method comparing to these methods.…”

SASIMI: sparsity-aware simulation of interconnect-dominated circuits with non-linear devices

“…However, the inverse of A is dense, which in turn implies that each simulation step involves dense matrix-vector multiplications. In contrast, the authors of [7] perform linear circuit simulation using a new formulation, called RLP, to model circuits by resistance (R), inductance (L), and the inverse of capacitance (P). Although the resulting A matrix in the RLP formulation is dense, its inverse is sparse, which enables fast sparse matrix inversion and sparse matrix-vector multiplication in each simulation step.…”