Beyond DFM: when manufacturability has to be guaranteed by design

Designers and test engineers are increasingly being challenged to introduce new high performance products within ever tightening tolerance windows. Traditional production test solutions may not be able to deliver the required accuracy to guarantee performance by test without additional and often prohibitive investments in extreme high-end test systems. In order to maintain a reasonable cost of manufacturing, it is not sufficient that these parts be designed for manufacturability-they must be guaranteed by design. To be considered guaranteed by design, one needs to have the utmost confidence in their design for manufacturability system and particularly in the accuracy of their simulations. This paper describes multivariate analysis techniques, leveraging electrical test data on the process level, to accurately model the statistical variation of the process as well as provide the means for evaluating where any given ET sample falls within that distribution for validating simulations against actual measured performance data. A case study drawn from a recent product introduction in a 0.35 /spl mu/m CMOS technology is used for demonstration.