Statistical reliability control from an IC user's perspective

Abstract

The primary cause of disc drive failures continues to be component failure of integrated circuits (ICs). A solution to the dilemma is proposed, which is described here as `statistical reliability control' (SRC). After reviewing some limitations and opportunities with the alternatives, this paper focuses on wafer level reliability (WLR) and its potential for being the most statistically significant and cost-effective way to achieve SRC `upstream,' in the IC manufacturing process. This paper briefly discusses WLR's potential as the best vehicle for studying IC failure mechanism kinetics and the relationship of critical process parameters to IC reliability, to help accelerate development of advanced wafer processes and IC packages with built-in reliability. Finally, some comments on Seagate's SRC specification and ideas for implementation and use of this type of spec by users and IC manufacturers are proposed, with the goal of at least reducing many traditional `end-of-line' reliability tests and screens that are sometimes overemphasized, at the expense of directing efforts upstream, where early detection can have the greatest impact on payback and effectiveness.