Fueling the productivity engine
02/01/1998
Fueling the productivity engine
Moore`s Law states that the number of transistors on a chip doubles every 18 months while the cost of the chip remains constant. This steady decrease in cost/function is directly responsible for the increasing market penetration of semiconductors and the emerging information age. Yet, since the cost of equipment, facilities, and engineers climbs steadily, Moore`s Law requires continuous productivity improvements. Traditionally, these gains have come from improved yield and increased wafer size.
Yields are already approaching 80-90%, however, and the costs of further improvement may exceed the benefits. The transition to 300-mm wafers is proceeding, but recent estimates put the cost of a 300-mm fab at a staggering $2.6 to $2.8 billion.
Meanwhile, only about 30% of equipment time is spent actually processing product wafers. The majority of equipment time is devoted to setup, maintenance, and idle waiting for upstream tools. Complex interactions among process steps make yield problems, when they do occur, more difficult to diagnose. Higher value chips and larger wafers increase the costs of scrap, even as re-entrant process flows and less-experienced operators increase the risk of misprocessing. Future productivity improvements must come from these sources.
In this issue, Solid State Technology begins a three-part series devoted to fab level productivity issues. The first article ("Improving wafer fab productivity with efficient floor layouts," p. 62) looks at fab design and layout. How can fab designers support manufacturing goals and avoid building limitations into a new facility? How can fab managers optimize the inevitable tradeoffs among cycle time, flexibility, process quality, and capacity? The answers will vary from one facility to another, but this article presents a uniform decision-making framework.
The next article, to appear in the March issue, deals with manufacturing execution systems. A well-designed, well-integrated manufacturing execution system can improve the performance of the entire fab by smoothing work flows, scheduling downtime, and eliminating mistakes. Conversely, poorly implemented software can be worse than no automation at all, bringing production crashing to a halt. This article examines what software can and cannot do to improve factory effectiveness.
The third article, appearing in the April issue, shows how to optimize staffing levels. If a fab has too many operators or technicians, they get in each other`s way and quality deteriorates. If a fab has too few personnel, equipment problems are not handled quickly enough and escalate into disruptions. Inflexible "activities/hour" staffing models make no distinction between constraint and nonconstraint equipment, and fail to accommodate the random nature of fab events. This article presents a structured approach to determining staffing levels.
Of course, productivity concerns don`t conform to the boundaries of this series. We are always interested in novel approaches to this and other semiconductor manufacturing issues. If you would like to share your experiences, please drop me a line at the e-mail address above, or at 10 Tara Blvd., Fifth Floor, Nashua, NH 03062.