The road to higher productivity
01/01/1999
The road to higher productivity
If it costs $1.5 billion to build a new semiconductor fab, then getting one free of charge is a super bargain. That is what Texas Instruments set out to do a few years ago - create enough additional good wafers per month across several fabs to equal the output of a brand new fab. In effect, the result would be a "virtual fab," producing the same output from existing fabs as the company would achieve by building a whole new factory. It worked! In fact, TI was able to continue this project for multiple years, achieving a number of virtual fabs over the period.
What secret formula made this remarkable achievement possible? There was no secret formula. Fab managers from TI facilities all over the world came together to discuss their ideas for boosting yields and improving productivity. Data showed that some fabs were already getting much better results than others. It turned out that there were reasons for this. Sometimes vendors helped to boost throughput, or achieve higher uptime for critical tools. Tool operators often found better ways to cycle wafers through the processes in their plants. By exchanging information and ideas, each fab was able to learn from the others, and improve its own operations. The fab managers continued to communicate by e-mail and through facilitators who would travel from fab to fab, passing on valuable lore.
As with so many projects, many small steps added up to major gains, a process the Japanese call "kaizen," meaning continuous improvement. Since TI was the first American semiconductor manufacturer to build a plant in Japan, it was one of the early US proponents of the total quality concepts that had enabled the Japanese to become pre-eminent in world manufacturing.
In these tough times, something like the virtual fab concept is being pushed throughout the semiconductor industry to meet tremendous cost pressures. Chip manufacturers have boosted output by accelerating the shrink of circuitry, thus allowing many more chips on each wafer. This has helped forestall a costly shift to 300-mm wafers, with all the new process tools, robotics, and automated systems needed to move larger, heavier wafers around the fab. To achieve this rapid shrink, however, it has been necessary to concentrate on defect reduction. Smaller features increase the probability that there will be particles or other defects large enough to cause circuit failures on many more chips. So yield management must be an integral part of any attempt to boost productivity.
In addition, as process steps are added to produce more metal layers, cycle times need to be reduced to meet customer demands. This requires new attention to equipment downtime, cleaning cycles, fab management software, inventory management, and all the other aspects that can help optimize cycle times.
This issue of Solid State Technology includes a Special Section on Productivity (starting on p. 33) that looks at these critical areas, offering ideas that could help your own efforts to boost productivity, improve yields, and shorten cycle times.
While concentrating on improving present operations, fabs are also moving toward new materials, such as copper, and low-k and high-k dielectrics. The circuit shrink is pushing the limits of optical lithography, and new processes for cleaning, ashing, and effluent control are needed to cut the consumption of chemicals and DI water. Many of the technologies that will be important in the coming years will be addressed in the Millennium Series, starting in this issue with a look at future flat-panel display technologies (p. 51). Each issue this year will contain one article in the series that will look at some key aspect of semiconductor processing; February`s article will probe methods being developed to extend optical lithography. Looking ahead is becoming an important part of the job for everyone in the fab. The Millennium Series will help.
For readers wishing to contact us, our editorial offices have a new street address: 98 Spit Brook Road, Nashua, NH 03062. Other contact information remains the same.
Robert Haavind, Editor in Chief, [email protected]