For several decades, the output from semiconductor manufacturers has been high volume products with process optimisation being continued throughout the lifetime of the product to ensure a satisfactory yield. However, product lifetimes are continually shrinking to keep pace with market demands. Furthermore there is an increase in 'foundry' business where product volumes are low; consequently it is no longer feasible to optimise the process during the product lifetime resulting in an increase in parametric yield loss. This paper describes the use a combination of two statistical tools namely Design of Experiments (DoE) and Response Surface Modelling (RSM) which permit the identification and modelling of those process parameters whose variation which will impact most on the performance of a circuit. The efficiency of this approach, compared to using a Monte Carlo analysis, is demonstrated with respect to a Mutual Exclusion Element (MUTEX) which is used extensively in synchronisers where process variations can have considerable impact on circuit performance. To obtain the same modelling accuracy, the Monte Carlo approach would require large number of simulations compared to nine using the DoE scheme with the low computational overhead. This method can be used by semiconductor manufacturers and design house alike to bridge the gap between manufacture and design.
With deep submicron technologies, the importance of interconnect parasitics on delay and noise has been an ever increasing trend. Consequently the variation in interconnect parameters have a larger impact on final timing and functional yield of the product.We present a comprehensive analysis to quantify the impact of parametric variations on the reliability of global interconnect links in the presence of crosstalk. The impact of parametric variations on wire delay and crosstalk noise is studied for a global interconnect structure in 90nm UMC technology, followed by a novel technique to estimate the bit error rate (BER) of such links. This methodology is employed to explore the design space of interconnect channels in order to mitigate the impact of variability.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.