Quantifying the value of ownership of yield analysis technologies

Abstract

Abstract-A model based on information theory, which allows yield managers to determine optimal portfolio of yield analysis technologies for both the R&D and volume production environments, is presented. The information extraction per experimentation cycle and information extraction per unit time serve as benchmarking metrics for yield learning. They enable yield managers to make objective comparisons of apparently unrelated technologies. Combinations of four yield analysis tools-electrical testing, automatic defect classification, spatial signature analysis and wafer position analysis-are examined in detail to determine the relative value of ownership of different yield analysis technologies.

Yield learning and the sources of profitability in semiconductor manufacturing and process development

“…Experimentation cycles of different practices vary from real time in the case of in situ sensors to months in the case of reliability physics. The amount of information extracted per experimentation cycle and per unit time can serve as a metric for learning rate [33]. 4) Within the problem solving process, rapid localization of a problem to a particular technology provides the most leverage [30], [31].…”

Organizational Learning and Capital Productivity in Semiconductor Manufacturing

“…During CS, ramping to volume production concurrently with fault reduction and die shrinkage becomes affordable once die-sort yield rises rapidly, and a revolution in organizational performance takes place [15], [25], [35]. The financial success of a semiconductor venture to a large degree depends upon whether this revolution occurs on time [7]- [10], [14].…”

Inspections and Assembiels When two negative effects turn positive

“…With a single, final inspection with α = 5%, the yield decreases to 47% and the input grows to 2,127.22 units. Suppose six inspections are added, say after each ten operations, as in; e.g., [8], [9], with α = 5%, for each inspection. Then, the good intention results in a yield decrease to 34.55%.…”

“…"The potentially dire consequences of not detecting an electrical fault early during the process have motivated technology managers in the semi-conductor industry to introduce an inspection step after about every ten process steps." [9] When inspections are introduced, inspection errors appear -missing defective unit and false rejection of conforming units; e.g., [5]. Each falsely rejected unit cannot be used as intended and hence requires compensation just as a defective unit.…”