快速多層介質電容萃取

Abstract

The ever-increasing complexity of nano-scale ULSI design has made parasitic capacitance extraction more and more challenging. Efficient and accurate capacitance calculation is needed urgently for delay evaluation especially for deep sub-micron designs. With the wide-spread application of multiple-layer interconnect structure, the BEM based discretization of dielectric-dielectric interfaces not only significantly increases the execution time and memory consumption, but also gives rise to instability and inaccuracy during the surface charge density calculation. In this thesis, we propose a fast multiple-dielectric capacitance extraction algorithm, MimCap, by integrating the truncated image method (TIM) into a hierarchical sparse capacitance extraction approach. The truncated image method approximates the dielectric-dielectric interface by a finite series of image charges and hence avoids the undesirable dielectric-dielectric interface process. Extensive experimental results demonstrate that MimCap is not only fast but also accurate. By preserving only the first 1~3 orders of image charges, MimCap exhibits over 100X speedup compared with FastCap while maintaining percentage error around 2%.