Business @ any speed
02/01/2000
Jack Ghiselli
Brooke Banbury-Masland
New software development tools are having a profound effect on today's semiconductor factory automation and communications projects, particularly object-oriented tools like those used in Microsoft's Windows DNA (distributed internet application) architecture. Techniques such as DCOM (distributed component object model) allow distributed computers to exchange data in new and useful ways. These new tools can provide great benefits, but we must be careful to fit them into our factories wisely.
Semiconductor factory automation systems incorporate a wide range of different computer hardware and software "platforms." Since semiconductor companies design and build microprocessors, it is no surprise that they have strong and differing opinions about the "best" hardware and software, and that different factories use different systems. They are not likely to change anytime soon, and in this regard, our industry differs from, say, the banking industry. Equipment suppliers have similarly strong and differing opinions about the "best" computer and operating system platform on which to implement their embedded equipment control software. For complex controllers, many choose specialized or even custom-written, real-time operating systems. Some of the platforms are compatible with DCOM, but many are not. Since it is not practical to scrap all legacy software in the factory, we are challenged as to how to wisely introduce systems using the latest software development techniques, leveraging their power, while at the same time leveraging our investment in existing hardware and software.
Another complication is the huge difference in speed of change. Software development technology changes rapidly, whereas factory equipment changes much more slowly. Once a semiconductor factory installs and qualifies a unit of equipment, the factory typically "freezes" it and prohibits any changes to software during the equipment's productive life, which might range from five to eight years. Then, the equipment might be sold to a second-tier manufacturer that might run it for another five to eight years.
Compare that to the speed of change in desktop software development systems, which during the last five years have evolved through MS-DOS, Windows 3.1, Windows 95, several versions of Windows NT, and Windows 2000. MS-DOS is considered obsolete, except by the people who use it successfully to control their semiconductor tools.
There is even a lively business in "re-manufacturing" older equipment. In a 10-year-old diffusion furnace, the control computer and software would almost certainly be considered obsolete, but the mechanical components, quartz tubes, RF heating coils, etc., are still useable. A re-manufacturing company might replace the computer and software, using the latest techniques, yet keep the old mechanical equipment.
Reliability is also a major driver in an industry that operates continually with low tolerance for downtime. Here again, there is strong disagreement about the "best" systems. Companies ranging from Saturn, Pratt & Whitney, and Merrill Lynch to NASDAQ are now using Windows NT to run their operations. On the other hand, NASA and the air-traffic control system are examples where older operating systems or custom-built applications are used because of historical reliability performance. The point is not to determine who's right, but rather to recognize that users disagree, and to build software systems that can work together regardless of differences in platforms.
Much automation software is targeted toward the wafer fab. However, test, assembly, and packaging (TAP) operations have their own set of problems, as described in "Paths to assembly automation: Different data for different objectives" by Gerard Michaud and Richard Burda of Kulicke & Soffa Industries in SST's September 1999 issue. The most cost-effective automation solutions for TAP might not involve wholesale factory-wide integration of automated material-handling systems, but rather incremental implementation of automated data handling. As the industry tries to juggle the diverse needs of the frontend and backend, it is no wonder that a one-size-fits-all approach is not going to work.
The dual challenge of using the latest software tools while interfacing them to the legacy systems found in the factory — in both the shop-floor and embedded equipment arenas — can be met by implementing "islands" of software. We implement an island using the latest and most powerful software techniques, but recognize that the island must remain useful even as those techniques are superseded by newer ones. An island might be one particular type of equipment, or it might be one segment of a shop-floor MES system. The 300mm transition has taught us the value of widely used standards. Open industry standards, like Semi's SECS/GEM and CIM Framework standards, and perhaps new standards, allow the interconnection of new "islands" with old "islands," and each island can evolve at its optimum rate without requiring all other islands to change at the same speed.
Providing solutions based upon standards — without requiring users to change faster than their tolerance and comfort level — is the only way to enable integration and meet the communications needs of an industry whose business model demands making sound business decisions. The semiconductor industry has become a major driver of the world economy with progress fueled, in part, by being able to change and adapt — or to paraphrase the title of a well-known book* — being able to do business at any speed.
*Business @ the Speed of Thought Using a Digital Nervous System, Bill Gates with Collins Hemingway, Warner Books, 1999.
 |
Jack Ghiselli is president of GW Associates Inc., 1183 Bordeaux Drive, Suite 27, Sunnyvale, CA 94089; ph 408/745-1844, email [email protected].
 |
Brooke Banbury-Masland is managing consultant of the Manufacturing and Engineering Practice Group at Microsoft, 1 Microsoft Way, Redmond, WA 98052; ph 425/936-5693, email [email protected].
Insight welcome!
Do you have an opinion on problems in the industry or on where you think the industry should be headed?
Send us a proposal for an Industry Insights column, or, if you prefer, an actual article of 1000 to 1200 words.
Send to:
Julie Callahan, Solid State Technology
98 Spit Brook Road
Nashua, NH 03062-5737
[email protected]