2001: The
05/01/2001
Jack Ghiselli, GW Associates Inc.
 |
At this point, the global semiconductor industry has announced some 17 new wafer fabs using 300mm technology by the end of 2002 (according to IC Insights and Solid State Technology). That would suggest that 2002 would be the critical year in the transition to the larger diameter. In truth, however, this current year is really the pivotal one because we will know better if the 2002 goals can be met.
Right now, it is questionable, and I believe the equipment industry is the stumbling block. Even though major semiconductor producers have loudly stated their insistence on 300mm FOUP-carried wafers being delivered directly from the automated material handling systems (AMHS) to all production equipment, few equipment producers seem to be close to compliance to the necessary standards.
The complexities involved in adding this capability are only just becoming apparent. Infineon's highly sophisticated 200mm fab in Dresden has literally kilometers of AMHS conveyors, yet delivers only part way - to a drop-off point in the bay, not directly to equipment. Instead, a human operator moves the carriers between AMHS and the equipment.
Direct AMHS-to-tool handoff is certainly a challenge for the equipment industry, but it also is a major headache for the IC makers. Instead of dealing with only one AMHS equipment supplier as they did with 200mm facilities, they must now also deal with 50 or 60 different production equipment types, each with its own "mini-AMHS" interface. It is left to these suppliers to follow the applicable Semi Standards already in place for connecting each individual tool into the automated system.
Based on their experience with earlier wafer fab communications standards of the SECS/GEM generation, many equipment companies assumed that "connectivity" could be added on after the product development cycle ended. Unfortunately, it is no longer that simple; implementation of the standards - which had been hoped for by year's end - is running far behind schedule.
International Sematech and Semi held a 300mm workshop in Austin last October. There, three major suppliers - Applied Materials, Tokyo Electron Limited (TEL) and ASM Lithography - described their work in developing custom software solutions for 300mm connectivity. All had some success, but still had some additional way to go.
Only one supplier speaking at the Sematech conference - Therma-Wave, a California metrology company - had used commercial software to implement the E87 and related E84 and E39 standards in a finished product, the company's OP 5340 thin film measurement system with dual FOUPs. The company said that validation tests were run on a commercial tester, primarily as a means of testing the Semi standards involved. While some tests turned out to be invalid, the system passed 97% of the tests that could be run, reporting that the other 3% were attributable to issues to be resolved in the E87 standard itself. This is the "AMHS-ready" capability 300mm fabs are requesting and Therma-Wave's Rakesh Khanijow told them that the connectivity standards were implemented in only seven weeks utilizing generic connectivity software, which compares favorably to the 1-2 years already invested by some of the larger companies.
Development of a generic software solution to the connectivity problem has been at the heart of the Semi Standards communications protocol development program over the past two decades. The costs and complexities of the conversion to fully automated 300mm production have made a unified industry approach to connectivity even more critical. Certainly the standards development process itself is different this time around. Historically, equipment suppliers have developed their own solutions to industry problems as they developed their products. Then, after the fact, committees would be formed to study the various approaches and agree on an acceptable industry standard.
In the case of 300mm, the industry (as reflected by the members participating in the standards effort) felt that the high costs of product development justified a first-time effort to develop the standards in advance of need. Thus, each equipment company was not duplicating costly engineering efforts to explore the possibilities. That has resulted in a wide body of standards for various aspects of 300mm production, especially in the area of automation, that are being changed and refined as experience in the field warrants it.
Certainly there are some major players in the equipment industry with the resources to develop their own unique connectivity standards. The majority of Semi members, however, are the small- and medium-sized companies, where the cost of an independent effort adds unwanted expense to an already costly development program. A single company developing its automation software in-house must devote hundreds of engineering man-hours to keep abreast of these changes and incorporate them into the system.
This approach, it may be recalled, was also taken by the largest equipment companies in developing the original SECS/GEM communications protocols, with some firms feeling that developing their own protocol would give them a competitive advantage. This only delayed the standardization of software for the automated fab until it became obvious to most that an outside vendor serving numerous "partners" can better absorb these costs and spread them over a wider base.
Cost savings should be an easy trade-off, especially if you compare a seven-week implementation period against 1-2 years. A company setting out to develop its own in-house connectivity program can anticipate something like $1 million in nonrecurring costs to do it themselves. That amount doesn't include the lengthy debugging process or the constant documentation revisions to keep it current with changes in the software, which are coming rapidly as we gain experience on the fab floor.
Perhaps the ultimate motivator, however, as we begin to reach the deadlines set by some of the most important semiconductor producers, is the impact on time-to-market required for an in-house effort. On paper, you might argue that an in-house effort could save as much as $100,000. (I said, "on paper.") But any company independently connecting its individual products will still be required to connect with other manufacturers' fab equipment. The customers are insisting on this capacity.
The need for communications between every piece of fab equipment and the host computer is already established as critical to the success of the fully automated 300mm wafer fab, and the semiconductor industry has said it will not settle for less. So the $100,000 saved through in-house development begins to pale when it is compared with the number of $50 million equipment orders that can be lost by not having fully compliant automated systems available when needed.