A yield based replacement for capability indexes

In the competitive world of high performance microprocessors, time to market is of first priority, and extremely steep ramps in yield are critical to financial success. Such rapid increases in yield require rapid and accurate detection of problems in the manufacturing process. Classically, detection of process problems relies on the use of capability metrics, such as C_p and C_pk. Such metrics are accurate if the design limits are absolutely precise but, in early stages of prototype production, this assumption is often not justified. In this paper, we define point and interval estimates for ´yield impact metrics´. Such metrics circumvent the lack of precise design limits by directly measuring the impact of process inaccuracy upon yield. Improvement of processes with high yield impacts results in maximal acceleration of the yield ramp. In addition, we demonstrate application of these metrics to a yield ramp of one of Digital Equipment´s new microprocessors.