Dr. Zhenhai Zhu

From Cadence Berkeley Labs, Berkeley, CA

Abstract

Rough surfaces are pervasive in integrated circuits and systems at both on-chip and off-chip levels.  They are caused mainly by the imperfection in manufacturing processes such as electroplating and etching.  And, they can lead to serious performance degradation.  For example, extensive experiments have shown that surface roughness can cause the conduction loss on the printed circuit boards to increase by a factor of three at high frequencies.

The calculation of the equivalent circuit elements of complicated 3D interconnect structures, so called parasitic extraction, has become a mature sub-field in EDA research and industry.  The integral equation-based methods have been proven to be the most powerful approaches.  But so far, the surfaces of those 3D structures have always been assumed to be perfectly smooth.  The presence of surface roughness changes the nature of the problem and has unfortunately raised the numerical difficulty of the parasitic extraction to such a new level that even the state-of-the-art parasitic extraction algorithms are not able to handle practical structures.

Described in this talk:

• Why the rough surface problem is important
• A quick review of integral equation based parasitic extraction solvers
• Why these solvers fail to efficiently solve the rough surface problem
• The key elements in the proposed fast stochastic integral equation solver (FastSies)
• The potential applications of FastSies in high frequency electromagnetic analysis of 3D structures with rough surfaces