Technical Viewpoint

Facing the Challenges of APC's Second Decade

Alan Weber

Advanced process control (APC) has had a major impact on the semiconductor manufacturing industry since the term was first coined by Sematech more than a decade ago. It is the unifying theme for a community of practitioners who congregate by the hundreds at three annual conferences around the globe. And it has spawned a dedicated group of suppliers that serve the APC market segment with a wide range of products and services, from sensors to fabwide control and optimization systems.

Now a mainstream production technology, APC is considered essential for the 300-mm, 130-nm technology node and beyond. As a result, it has attracted much attention and investment funds over the years. However, like many other competitive segments in the cyclical IC industry, APC is not an easy business to be in. Continued progress at the rate we have come to expect will depend on improving our effectiveness in that area.

The need for improvement was the main motivation behind a recent series of discussions with APC veterans involved in both the business and technology arenas. Engaged in activities from fundamental R&D to standards development, their host organizations include equipment and components suppliers, software product and services vendors, and IC manufacturers at the head of the food chain. A summary of the collective challenges facing APC, the opportunities for addressing them, and the changes that may take place in the coming decade are the subjects of this article.

Key Questions

Emerging from the deepest industry downturn in 20 years, the APC community must address several important questions that will shape the APC software market segment for much of the coming decade:

• In what direction is the commercial semiconductor APC software market heading—will it grow, remain flat, or decline? How does the industry value software?

• What business models have or have not worked in selling software to semiconductor manufacturing customers? What changes in end-user software buying behavior would improve business models?

• What types of organizations are planning to develop APC software? How much will be done by the end-users themselves?

• What are the prospects that process and metrology OEMs will supply application software beyond embedded control systems?

• What other factors have limited the growth of the APC market segment? What is going to happen from this point forward?

Commercial Software Market Outlook

Everyone has opinions about the place of commercial manufacturing software in the semiconductor industry (including APC)—and they're all over the map. But the prevailing attitude is fairly positive. There are several reasons for this. First, the ever-increasing complexity of process and equipment technology requires a high degree of specialization to develop viable control solutions, creating an opportunity for process-specific niche suppliers. Second, chip fabrication at current and future technology nodes relies heavily on automation, which demands much more overall systems engineering than has been applied to date. That, in turn, increases the market for system technologies. Moreover, end-users have much more access to process information than they did previously, partly because of the passage and continuing adoption of SEMI's diagnostic data
acquisition/equipment data acquisition (DDA/EDA) standards. Finally, the rosy demand picture for commercial software products is emerging against the dark backdrop of the industry's three-year recession, which is preventing end-users from growing their IT resources at the same rate they did in the pre-2001 period.

Although the market outlook is largely optimistic, industry observers across the board view the overall valuation of manufacturing software as problematic, especially when compared with software of comparable complexity such as design automation products. While manufacturing software is seen as a method for reducing costs (and is purchased by organizations under continual cost pressure), design software is seen as an asset for increasing revenues, where time to market is the principal concern. Moreover, the purchase price of standard commercial software is often capped by a customer's build-versus-buy analysis for a custom solution, which seldom considers the same functional and technology requirements as a commercial product or ongoing support costs. Until these factors are well understood and appreciated, manufacturing commercial software will continue to be a tough business.

APC Business Models

APC business models have evolved as technology and the market have undergone great changes over the past few years. To be successful, any business model must make it easy for customers to purchase what they need at all stages of adoption—from first contact through the feasibility study phase to fabwide deployment. What this means from a pricing perspective is that the cost to a customer and the value delivered must scale at the same rate. Given the expertise needed to deploy and maintain APC products successfully, the mix of service costs, license or usage fees, and maintenance/support rates must be balanced against the long-term viability of the business in question. That lesson was learned the hard way in the manufacturing execution systems (MES) sector.

In the business model that resulted from the downturn, prospective customers expect free product evaluations. Although this practice may be good or bad depending on which side of the purchasing table you sit, it is here to stay in an environment in which the number of independent global customers is shrinking. However, the goal of free evaluations must be to validate how potential solutions fit problems, not to postpone real project expenses under the cover of "free" consulting.

Finally, since no single solution will satisfy all of a customer's needs, technologies and attitudes that support effective coexistence with other APC solutions are necessary. The sooner an entire team is picked and begins to work on its shared objectives, the better.

Who Should Develop APC Technology? Depends on Your Perspective...

Chipmakers. Semiconductor manufacturers generally agree with one another in many APC-related areas. But being engineers at heart, they all tend to drift across the boundary between system requirements and solution design, ultimately limiting the range of potential systems and suppliers available to them and probably adding unnecessary costs. Chipmakers' energies would be better spent developing real end-user (i.e., process engineering) requirements, creating detailed test plans and other artifacts that increase the likelihood of success.

One reason that end-users spend so much time in "solution space" is that most of the leading ones are developing their own run-to-run (R2R) control software and the required system infrastructures, relying on the commercial market for fault detection and classification (FDC) packages. R2R and FDC systems differ from each other in two respects: First, R2R software requires more connections to complementary factory systems and is therefore inherently more custom than FDC; and second, the intellectual property (IP) incorporated in R2R systems is viewed as more process- or operations-specific and, therefore, more closely guarded than FDC systems.

Another key factor in end-users' APC project planning is that many of them have been burned too many times buying software from either tool OEMs, resulting in commitment and capability issues, or small software companies, raising survival issues. Consequently, many end-users rely on their own efforts to create critical APC technology.

Advanced Micro Devices (AMD) is the most visible example of a chipmaker committed to internal APC technology development. The company runs corporate- and fab-level organizations that provide the full range of disciplines to build, deploy, and support APC applications. While these applications run in a common commercial environment, the bulk of the real process control technology resides in the custom software created by AMD itself.

Most semiconductor manufacturers want more-detailed process data from the equipment they use and expect OEMs to make it accessible to them. Their motivation is to better understand fundamental tool and process behavior, which is the first step toward improving performance. Ultimately, it is in the tool suppliers' best interest to provide chipmakers with critical tool information, but IP ownership hurdles remain.

Finally, what do chipmakers expect from software suppliers? They believe that their suppliers must listen better to understand real production software
requirements and should spend more time in the fab to fully appreciate the meaning of "mission critical." Some IC manufacturers have even suggested that every software engineer's "rite of passage" should include a serious assignment on a production test or support team. There is no better way to prevent middle-of-the-night support calls than to put the developers on the front line.

Software Suppliers. The industry's software suppliers are a dedicated, hard-working group whose opinions are every bit as strong as the chipmakers'. First and foremost, they believe that their applications depend on good data to perform properly and that everyone would benefit if the chipmakers would demand better data quality and more openness from the OEMs. Addressing these issues may come slowly, however, since many data-quality problems are caused by the limitations of the equipment's embedded control systems, not the standards that define the external interfaces.

The wave of internal development on the chipmaker side is a major cause for concern among software suppliers, because it is coming at a time when the suppliers hope (and need) to see some return on the investments they made during the long downturn. This conflict points to a deeper structural issue, namely the inharmonicity of semiconductor business cycles (see Figure 1).

Figure 1: Representational diagram showing the inharmonicity of semiconductor business cycles.

The APC product development cycle takes at least two years and requires outlays totaling millions of dollars from the initial requirements stage to first release. In contrast, the sales cycle can span nine months or longer for a complex, fab-level system. In addition, closing an APC deal with an end-user often requires the support of the same scarce technical personnel who must put the finishing touches on the product.

Once a sale is made, the next cycle is the customer deployment process itself, which may last two to three years from the initial pilot phase in a single process area to a full production rollout across an entire fab. Juxtapose all of this on the industry's infamous boom/bust cycle, and no wonder there is so much churning in the APC supplier base.

However, every problem presents new opportunities, and this case is no exception. Realizing that semiconductor manufacturing is truly a "custom" industry, several suppliers are responding to industry challenges with products that enable a fab's internal IT and process engineering staff to buy and/or build APC applications that meet their local specifications while allowing them to operate in the company's overall framework. One such approach is to combine a robust industrial platform and a "plant-centric" application model (see Figure 2). Building a layered system on the elements that vary the least in a factory (equipment, processes, and materials) is a novel way to maximize a system's stability while enabling almost unlimited degrees of freedom in the higher-level business processes. The higher up the pyramid a supplier goes to supply a "complete" solution, the more custom it becomes. Hence, while particular customers or fabs may see very high value in a complete APC approach, most will be unwilling to buy into a specific package that embraces everything from sensors to fab goals.

Figure 2: A "plant-centric" APC application model can meet the specifications of fab elements that are least variable (from the bottom of the pyramid up).

Several software companies offer commercial APC products from generic stand-alone FDC packages to process-specific analysis and control systems to fab-level application and integration frameworks. The diversity of the players in this space is demonstrated by the companies that participate in the yearly AEC/APC symposiums sponsored by International Sematech.

Equipment Suppliers. Equipment and subsystems suppliers have a key part to play in the APC area, but their role is still in flux. Most suppliers' APC efforts focus appropriately on the local aspect of the control problem (i.e., advanced equipment control); their goal is to hit the process targets that are set for a given run.

While some toolmakers are responding to end-user requests for more and better data, they are not likely to fully adopt the new SEMI DDA/EDA standards until customers demand them in their purchase orders. Still others are beginning to provide process-specific analysis and modeling software to enable deeper process understanding and fab-level optimization and control. The latter group have even dabbled in this market directly.

However, over the years the same toolmakers have "trained" their customers to expect free software, an expectation that will be very hard to undo given the high cost of semiconductor equipment and the competitive nature of the APC market. Consequently, the best return on a supplier's APC product investment will be through a software channel.

Tokyo Electron is the best example of an equipment supplier that has shown a long-term, strategic commitment to APC product technology (beyond embedded control systems). However, their products are only marketed with TEL equipment. Moreover, the company has not tackled the general-purpose R2R control market.

The Persistent Integration Problem

Integration was the biggest barrier to the adoption of APC technology in the mid-1990s, and it still accounts for nearly 50% of the effort required to deploy a production solution. The need for integration spawned the original APC Framework program (a more than $10 million, four-year investment scheme) and related standardization efforts. Integration continues to drive much in-house development. The industry is taking another shot at this target through the process control systems task force, a follow-on SEMI body that has focused on determining the major "functional groups" that comprise the APC space and specifying the message sets they use to communicate among themselves and with other fab systems.

However, communications is only part of the problem, and standards will only provide part of the solution. Why? First, because APC is fundamentally a continuous-improvement technology—the more you apply it, the more ideas you have to improve it. It is impossible to standardize things you have not thought of yet. Second, advanced APC algorithms depend on good data, which increasingly come from process tools and many other sources. At this point, it is hard to forecast what form these algorithms will take and, therefore, the nature of their interactions with other systems. As we have seen with the GEM300 standardization efforts over the past five to six years, integration in an automated environment involves much more than just communications.

The APC market is still relatively immature and very dynamic—there will be many valid solutions to complex processing problems. Therefore,
APC-related integration will be a moving target for some time to come.

Conclusion

The investment in new fab construction around the world represents the first chance for the industry to quantify the benefits of APC when applied early in a fab's life cycle. With only one or two exceptions, APC has been installed in fabs only after tools and processes have been qualified and used in early production and after many other complementary systems have been selected and installed. However, shortening a fab's "time to money" by just a couple of months can yield huge returns, as shown in the graph in Figure 3. Much required tool and process characterization can be well supported with only a partial system infrastructure. As the mechanisms and standards for integrating disparate manufacturing systems continually improve, the selection and deployment of an initial APC platform does not have to wait until all other related systems are in place.

Figure 3: Applying APC early in a fab's life cycle can yield significant returns.

We have come a long way since the first AEC/APC symposium in the early 1990s, and the current challenges will be overcome (and replaced by new ones!) as the APC market matures in response to increasing demand. How quickly that will happen will depend largely on the degree of collaboration among industry players. An open discussion of requirements, problems, and progress will result in the most efficient use of the industry's APC talent. That, indeed, should be our common goal.

Acknowledgments

The author would like to thank his partners, Jim Hollister and Paul McGuire, for their help in preparing this article. He would also like to acknowledge the many industry colleagues who have freely shared their insights about APC over the past few months.

Alan Weber is the president of Alan Weber & Associates, a consulting company that specializes in semiconductor advanced process control, E-diagnostics, and other related manufacturing systems technologies. Previously, he was the vice president/general manager of KLA-Tencor's control solutions division, which the company acquired from ObjectSpace. While at ObjectSpace, Weber created the fab solutions division and was responsible for the company's semiconductor manufacturing systems business, including development of the APC Framework and its eventual commercialization and global deployment. Before joining ObjectSpace, Weber spent eight years at Sematech, where he was responsible for advanced manufacturing systems and related standards R&D, including the CIM framework. He received a bachelor's and a master's degree in electrical engineering from Rice University in Houston. (Weber can be reached at 512/494-0700 or alan@alanweberassociates.com.)