An efficient solver for statistical capacitance extraction considering random process variations

This paper presents an efficient solver for statistical capacitance extraction considering random process variations, which is written in C and named PSCap. The improved continuous surface variation (ICSV) model [1] is applied to accurately describe the variational geometry of VLSI interconnects. Meanwhile, the weighted PFA, HPC [5], and parallel computing techniques are employed to improve its efficiency. Numerical results show that for the typical 65nm-technology structures, PSCap is 30-40% faster than the statistical capacitance extraction program statcap [1], which was developed in MATLAB, on an 8-core machine. Another contribution of PSCap is that it is able to handle multilayer structures, which statcap fails to handle, producing accurate results with less than 5% error.