Can overall factory effectiveness prolong Moores law
03/01/1998
Can overall factory effectiveness prolong Moore`s Law?
Douglas Scott, Robert Pisa, Promis Systems Corp., Nashua, New Hampshire
Gordon Moore of Intel first articulated the law that has governed the success of the semiconductor industry since its inception: chip performance doubles every 18 months at no additional cost to customers. SEMATECH expects shrinking feature sizes to provide significant productivity gains in the future, but larger wafer sizes and improved yields will make a smaller contribution than they have in the past. Gains in overall equipment effectiveness (OEE), while important and on-going, are insufficient, because no machine is isolated. Materials and processes must be successfully choreographed among hundreds of tools to achieve Overall Factory Effectiveness (OFE). The ultimate objective is a highly efficient integrated system, not brilliant individual tools.
The factors contributing to the semiconductor industry`s unprecedented rates of productivity improvement are well known (Table 1), but new approaches will be required to maintain the historical reduction of 30%/chip function/year.
Semiconductor manufacturers must (Fig. 1):
achieve better asset utilization and reliability;
improve yields and eliminate misprocessing;
simplify integration between plants and the enterprise;
accelerate time to full production for new plants, products and processes;
provide better decision support capabilities for plant managers;
reduce costs, inventories and cycle times; and
respond dynamically to unexpected events and requests.
 |
With today`s OEEs typically in the 30-40% range, there are clearly great opportunities for improvement. These include less setup time, fewer pilot wafer runs, less idle time, and more reliable equipment (less unscheduled downtime).
 |
Figure 1. Demands on fabs.
While OEE is about achieving excellence in individual equipment, OFE is about the relationships among different machines and processes. This paper explains the meaning of OFE and practical steps that can be taken to achieve it.
In their study of semiconductor development [1], Iansiti and West projected that the maximum number of interconnects on a logic chip will soar from about 60 in 1992 to 2000 in 2007, the number of gates will grow from 900,000 to 100 million, and metal layers will increase from about three to more than six.
The number of manufacturing steps doubled between 1980 and 1990. It tripled between 1990 and 1995 and is expected to triple again before 2001. This tripling of the number of steps will create extreme pressure to maintain yields. The increased complexity also gives rise to huge volumes of data, which are simply not amenable to manual analysis. Richer and more in-depth information systems and plant-wide integration are therefore essential to improving OFE.
Iansiti and West reached a striking conclusion: "Our data suggest that differences in the technology integration process are more important than disparities in project management methods, leadership qualities, and organizational structure in explaining variations in performance.
"Improvements in performance by US, Japanese, and Korean semiconductor companies did not stem from such sources as increased research or scientific breakthroughs. Discovering new technologies was not enough. Successful companies were those most adept at choosing technologies that would work together in an increasingly complex production system. That combination of novelty and complexity makes a company`s excellence in technology integration critical" [1].
OFE - a holistic approach
The manufacturing process is a complex web of interactions among process tools, metrology tools, materials, people, departments, companies, and processes. These inter-dependent activities cannot be set apart from each other, but too often they are viewed in isolation, and there is a lack of coordination in deploying available factory resources (people, information, materials, tools) to manage work efficiently.
M. Porter wrote that "differences in operational effectiveness were at the heart of the Japanese challenge to western companies in the 1980s" [2]. He emphasized that the fit among different activities drives both competitive advantage and sustainability. "Rather than seeing the company as a whole, managers have turned to `core` competencies, `critical` resources, and `key` success factors. In fact, fit is a far more central component of competitive advantage than most realize. The whole matters more than any individual part. Competitive advantage grows out of the entire system of activities."
OFE requires the successful integration of many activities and many information systems. A successful manufacturing plant requires applications for tasks such as: demand forecasting, order processing, purchasing, data warehousing, cost accounting, distribution and logistics, equipment control, material transport, and automated identification. These applications span the planning, execution and control layers, and many, if not all, of them must be integrated to maximize productivity.
What is OFE?
One of the strongest indicators of semiconductor industry profitability is the average fab capacity utilization. Stanley Myers, president of SEMI, states that "most IC makers want to continually increase equipment utilization and reduce processing delay resulting from queuing in order to increase return on equipment and other factory resource investments" [3]. The rate at which the factory acquires knowledge during process development and ongoing problem-solving must improve to allow faster increases in yield and throughput.
Don Martin of IBM Microelectronics noted that short cycles are an important attribute of a semiconductor manufacturing line, because yield learning is related to the number of WIP turns. Using maximum equipment utilization as the sole target can inadvertently lower performance. It is very easy to create incentives that lower the financial performance of the facility [4].
 |
OFE exists in a factory environment that allows users to:
synchronize the production schedule with planned downtime, setup time and qualification time through tighter connectivity to enterprise planning systems, and finite capacity scheduling;
optimize the sequence of orders/work/jobs;
ensure a balanced line, smooth flow of work, and full loads by integrating micro-scheduling (local lot dispatching) with overall plant scheduling;
increase fab automation with intelligent recipe management, wafer level process control, and automated data collection;
eliminate mistakes by better equipment communication, linking equipment with the factory system for rigorous verification of each step before run start;
minimize engineering holds and reduce/eliminate test runs through advanced process control (APC) - inter-step, intra-step, and in situ;
use automated equipment control (AEC) for machine modeling, for monitoring machine conditions and set points, and to predict impending system failure;
increase responsiveness to special customer orders by providing accurate and timely line status information for orders;
improve engineering productivity through easier process, recipe and PPID specification and revision control;
provide better decision support and exception handling capabilities; and
integrate shop floor systems with corporate business systems, to provide accurate information for projecting completions or revenue flow.
All of the above require better information systems.
What information systems can do
Information systems are as vital to semiconductor plants as lithography and etching. They have a huge impact on line efficiency, and information management solutions play a prominent role in helping manufacturers improve productivity and profitability.
Computer integrated manufacturing (CIM) is the mechanism for improving OFE. We need a full suite of pre-integrated applications that can help us manage virtual factories. Frank Robertson, VP and GM of I300I, recently wrote: "Effective use of costly manufacturing assets also demands automation and CIM systems" [5].
Since the world is littered with failed CIM projects, how do we get it right? CIM is a continuously evolving process, best undertaken through a close alliance between the manufacturer and a solution partner who can provide manufacturing enterprise solutions, implementation and integration services, complementary manufacturing software packages already pre-integrated, and partner products and systems integration services.
Manufacturing enterprise solutions
Manufacturing enterprise solutions (MES-II) are the next evolution of the more familiar manufacturing execution systems (MES). Though no fully integrated MES-II system is commercially available, industry consensus is beginning to define the capabilities of such a system. To meet the decision support needs of plant managers, MES-II applications must be integrable among themselves; they must be integrable with planning and business applications; and they must be integrable with applications running in other (remote or contractor) facilities. MES-II improves OFE because it integrates planning (ERP or MRP II) applications, with tactical execution and control applications.
Imagine software that gives a graphical, high-level status board of any aspect of plant operations; that alerts users, at a glance, to problems with material, equipment, and process stability; that warns of developing bottlenecks, fluctuating yields, or drifting cycle times before they become a problem, all based on up-to-the-minute, real-time data (Table 2).
Specification. In May 1997, Advanced Materials Research Corp. reported on the need for configuration control of process data, revision control, workflow for approval, and product structure to ensure related items are updated as the design or the process is modified. The distribution, not the creation, of controlled information drives enterprise strategies.
MES-II provides factory and process models that accurately represent real-life plants, processes, and resources, and can be shared by all plant and enterprise systems. Engineering costs are reduced through streamlined ECO procedures for on-line review, approval, and release to the manufacturing floor. This speeds the engineering change process, and ensures that correct revisions are applied in production.
Automated shop floor control. Increasing automation and the advent of 300-mm processing are bringing significant change to shop floor control systems. Automated shop floor control (ASFC) applications execute process instructions defined by the specification products. They provide guidance, data collection, control, automation, tracking, and traceability of production across virtual factories, with customer configurable business rules. ASFC products use common framework components to collapse what have traditionally been the execution and control layers into a single layer with redundancy capability to support 7 ? 24 paperless operations. The result of this merger, ASFC for the 300-mm era, includes:
resource management (e.g., equipment, tools, durables, containers);
common flexible GUI (allowing different interfaces for different work areas if required);
batch and sub-lot tracking, sample (run-ahead) tracking, and qualification lot tracking;
equipment interfaces for recipe download, equipment status, events, and data upload;
recipe differentiation at the individual tool level;
recipe verification and operator certification;
APC - inter-step, intra-step, and in situ;
advanced equipment control (AEC);
flexible and fully automated data exchange among equipment, batch, lot and sub-lot, environmental, and supplies objects;
control of process and metrology equipment, and of automated material stockers;
automated material handling (inter-bay and intra-bay) for reticles and wafers;
intelligent step and sub-step tracking, PPID management and electronic data collection for linked, cluster, and clustered tools;
automated ID systems for material, equipment, and personnel;
automated handling of lot additions, removals, splits and merges, both pre-defined and ad hoc;
quality management (e.g., real-time SPC and corrective action plans);
micro-scheduling (real-time dispatching);
cassette and reticle tracking and management;
support for multiple different parts on a wafer, or multiple different panels on a substrate;
wafer and chip level or substrate and panel level process control; and
correlation of data from multiple plant applications.
Butler et al. stated that CIM systems will evolve to an architecture for complete APC [6]. Feedback and feed-forward control will move away from qual wafer-based to production wafer-based. The data will come from in-line metrology tools, final electrical test, and from sensors on the tools. Because the CIM system stores data from all the various unit processes, as well as final test, it has the data required to calibrate and fine-tune the equipment and cell controllers [6].
Reporting. By transforming data into focused information, MES-II spotlights exceptional situations. It provides decision support, with as-it-happens displays for real-time monitoring of performance. Data visualization techniques present information in 3-D graphical format, enabling rapid comprehension and insight into large amounts of complex data.
MES-II monitors and analyzes production information against quality standards for early detection, avoidance, and correction of deviations, thereby lowering manufacturing costs. It automatically alerts users, groups, and locations to conditions requiring action, such as out-of-control charts, engineering changes, and equipment malfunctions. Material is traceable all the way back to the parent vendor lots to make it easy to track down product that may be adversely affected by a bad lot of material from a supplier. MES-II also provides customizable production reports, flexible data analysis, and accurate costing of WIP and finished goods.
Rigorous traceability may have other spin-off benefits. Theft, hijacking, and armed robbery are a global problem in the industry. Estimated losses in 1996 were $8 billion, and losses are projected to increase to $20 billion by 2000. During 1995 and 1996, 400 companies were victims, and according to law enforcement authorities, an armed robbery or burglary of an electronics firm occurs every five days in Silicon Valley. Unique identification of material, and traceability from source to customer helps reduce such losses.
Supply chain management. In Jan. 1997, AMR reported that "planners and plant managers spend much of their time expediting hot jobs, not optimizing plant operations." Real-time data from MES-II can provide significant benefits in planning and decision support. More precise modeling of plant resources and processes, along with better planning and multiplant master production scheduling, can significantly improve the use of assets, a critical advantage in a very capital-intensive industry.
Compelling benefits are derived by integrating planning, scheduling, and real-time execution control. MES-II collects data on resource utilization, queues, and production rates and enables process optimization to meet customer demand.
Billion-dollar facilities must be scheduled effectively for maximum profitability. Capacity constraints must be enforced during master scheduling runs, and during short-term finite scheduling and dispatching. Alternate flows and contract manufacturing must be considered during scheduling, not after.
MES-II downloads recipes and work schedules, and uploads production results. By capturing real-time information about setups, run times, throughput, and yields, an MES measures constraints and identifies bottlenecks, thereby offering a means to improve manufacturing capacity management.
Interoperability. To achieve OFE, MES-II integrates planning, execution and control applications, and remote or contractor facilities. For example, as soon as a master production schedule is released to the shop floor, something happens to invalidate it. Feedback about shop floor conditions has to be relayed back to the scheduler continuously, so it can produce a revised schedule based on reality.
John Garbayo of Motorola observed that "usually a tool`s processing features are touted as the deciding factors in equipment selection, but any tool with a weak interface inhibits the automation process." Through proper integration, recipes can be stored, managed, and the correct versions routed automatically to the tools. Moreover, by communicating with the host, an engineer can determine whether a recipe is dated correctly, has the right parameters, and is properly loaded. An automated mechanism can also verify that each run of each tool used the correct recipe [7].
In addition, interoperability allows users to:
integrate maintenance and production management applications to coordinate production and maintenance schedules, thereby improving equipment utilization;
pass recipes, PPIDs and/or control parameters to automated process equipment, which performs real-time monitoring and coordination at the equipment level;
provide feed-forward/feed-back control of process parameters through interaction among different process and measurement tools; and
integrate yield analysis with shop floor applications.
With SEMI`s new standard for lead-frame marking, traceability can be provided from source materials used in the wafer fab to packaged and shipped devices. This dramatically reduces problems associated with hidden defects. For an automobile chip, for example, it may limit a recall to 1000 vehicles instead of 100,000.
 |
What information systems cannot do
The above discussion of MES-II shows many ways that information systems can improve OFE. Other factors affecting OFE cannot be controlled by information systems, or require both information system advances and changes to equipment and/or software: smaller geometries; orientation to single wafer vs. batch processing; larger wafers; reducing test wafer loads; more reliable equipment; lower particle counts (improved defect density); and more integrated process tools.
How do we measure OFE?
Given the complexity of chipmaking, OFE is the result of many interdependent activities; so how do we know if OFE is improving? Different plants and companies have different goals, so there can be no single indicator. But a composite metric can be derived by asking what the key goals are, defining measurable criteria for success, applying weighting factors as desired to meet corporate goals, and computing the overall result (Table 3).
OFE, or the metrics for individual goals, can be used to compare the performance of plant A with plant B, or to trace changes in a plant`s performance over time. Drill down capability allows examination of the factors contributing to each of these measured results.
Conclusion
Nan Stone, editor of the Harvard Business Review, recently wrote: "In a knowledge-based economy, the prerequisites for success - innovative ideas, individual initiative, responsiveness to change, personal responsibility - cannot be commanded. They can only be volunteered" [8].
Equipment manufacturers are looking for ways beyond traditional methods to improve the process capability of the tools they manufacture, and are beginning to embed some of these information system techniques directly into the process tool.
Although many of the traditional sources of productivity improvement are reaching the point of diminishing returns, new approaches can increase OFE, and improve our chances of prolonging the extraordinary productivity gains achieved by our industry.n
References
1. M. Iansiti, J. West, "The Evolving Challenge of Semiconductor Development," Harvard Business Review, May-June 1997.
2. M. Porter, "Operational Effectiveness is Not Strategy," Harvard Business Review, Nov.- Dec. 1996.
3. S. Myers, Solid State Technology, p. 105, Oct. 1997.
4. D. Martin, IBM Microelectronics, Advanced Semiconductor Mfg Conf., 1996.
5. Frank Robertson, I300I, Solid State Technology, p. 101, Oct. 1997.
6. S. Butler, et al., SEMATECH, "Sensor Based Process and Tool Control," Future Fab International, No. 2, Vol. 1.
7. J. Garbayo, "The seven deepest equipment integration pitfalls," Solid State Technology, p. 117, July 1997.
8. N. Stone, Harvard Business Review, July-Aug. 1997.
DOUGLAS SCOTT is VP marketing, advanced technology, at Promis Systems Corp. He founded Promis Systems and served as PROMIS product manager and president for the company`s first 10 years. Promis Systems Corp., 170 University Ave., Suite 1200, Toronto, Canada M5H 3B3; ph 416/977-0599, fax 416/977-2016, e-mail [email protected].
ROBERT PISA spent 15 years in photomask engineering, wafer fab process engineering, process engineering management, and operations management with Analog Devices Semiconductor Div. and Polaroid Corp. He joined Promis Systems two years ago and is now a senior product manager. Promis Systems Corp., 6 Trafalgar Sq., Nashua, NH 03063; ph 603/886-1230, fax 603/886-4799.