Ring oscillators yield analysis: Improving Monte Carlo models with optimized clustering methods

Abstract

SummaryThe aim of this article is to demonstrate the interest of the optimized clustering method (OCM) when dealing with the yield analysis of ring oscillators, given uncertain assumptions over inputs. Indeed, the complementary metal–oxide semiconductor (CMOS) integrated circuit domain is facing a major challenge since these devices are subjected to noticeable fluctuations (mostly due to transistor constant scaling down at sub‐micron sizes), with utmost expectations regarding their performances (e.g., for digital circuits). Whether the classical approach requires use of Monte Carlo methods coupled with classical circuit solver (e.g., Spice and/or transistor model), the natural complexity of systems and varying inputs needs to provide alternative techniques, such as OCM. The foundations of the method and its efficiency will be presented and illustrated through ring oscillator applications.