Designing and Building 300 mm Production Fabs

The process of fab design and building is one of the most important decision processes a company can encounter in its lifecycle. Adding to the complexity, most companies do not build fabs each year. Because the predictability of 300 mm tools cannot compare with that of 200 mm tools, the current wafer size transition poses serious risks to the design because of the requirement for full automation. However, a good design process can address these challenges.

This article will discuss the best way to start the fab design process, create the right toolset from a paper flow and incorporate different market scenarios at the toolset determination stage. It will present a coordinated timeline for modeling capacity requirements, finalizing tool quantity and selection, designing the right physical layout and weighing manufacturing vs. management goals.

The core objectives of the design process include:

• Accurately forecasting performance demands.
• Modeling and understanding the resources needed to meet fab performance.
• Generating the best fab design within budget.
• Getting from initial decision to on-line capacity as quickly as possible.
• Installing and qualifying all tools on time, within budget.

	1. The fab’s conceptual design is based on predominate process flows, which should capture 80% of the processes. Process flows dictate the tool list, size of the functional areas and relationships.

Of course, since your process flows are varying in levels of confidence, the toolset confidence can only match that. But even in very uncertain flows, you can identify problematic areas and find tools to address them. A thorough analysis of the steps allows you to delay the decision of actual tool selection for those steps until late in the process, when the flow is more certain.

Lastly, without knowing market demand two years in advance, we cannot make the enormous investment of a couple of billion dollars in a 300 mm fab that is not flexible. To quote one of Intel's top managers, "I want to be able to switch from product A to product B on a flip of a dime." That kind of agility comes from process flow understanding.

Fab design is not just layout and tool locations. To truly understand fab operations, we must model all resources including:

• Tool capacity for each tool type.
• Product and material flow analysis to create a flow model.
• Automation requirements, transportation times, different buffer needs and systems performance.
• Staffing, including operators, technicians, engineers and all support staff.

	2. The optimal work in process allows for the best cycle time and throughput composite, so WIP=CT×TP.

We suggest the composite of the two as a measure for fab performance. Drawing from Little's Law, there is an optimal work in process (WIP) level for each toolset that allows for the best cycle time and throughput performance. For the case shown in Figure 2, 40,000 wafers is the optimal WIP level, and

WIP=CT×TP ® TP=WIP/CT

The quality of the design can be measured by the ability to sustain manufacturing for that level. To help the actual design process, we suggest you define the organizational criteria, business criteria and design alternative selection criteria as:

Operational Criteria:

• Product/material flow
• Productivity/staffing impact
• Dispatch method (pull, push, etc.)
• Layout flexibility and ramp impact on existing production
• Maintenance practices
• Ergonomic issues
• Automation
• Material handling
• Modularity to future expansion
• Installation/qualification complexity

Business Criteria:

• Capacity
• Time to on-line capacity
• Cost/return on investment
• Quality and yield

Design Alternative Selection Criteria:

• Total cost of operations (TCO) vs. performance
• Budget — project budget vs. performance

	3. By modeling the quantity of product moving from one functional area to another, the designer can start designing the fab shell, even without a final equipment list or exact tool locations.

We can then begin designing the fab shell based on the block diagram. Since you can design the building to accommodate the functional areas, you do not need to know the final location for each tool when you begin building. We still have enough time to finalize tool location since some tool lead times are longer than a year. By starting construction in parallel with the detailed design stage, you can save two to six months.

Regardless of the final automation system you choose, to start building construction you need to know transportation volumes and times. The general type of automation system, e.g. overhead monorail, is also factored into shell design, as well as buffer and WIP stocker locations.

Although the detailed design of the material handling system is not required for the early construction stages, you want to finalize system selection as soon as possible and custom tailor it to the needed performance based on the detailed resource model. Since 300 mm does not allow manual transportation, the wafer handling system sets the pace of the fab.

Calculating the WIP, buffers for bottleneck tools and operational goals is the first step to specifying the handling system. The main challenge is to create and test a dispatching algorithm that will ensure delivery of the right lot to the right place at the right time.

Once you process the high-level design and begin the actual building, your next steps are the detailed layout and A&E (architecture and engineering) design. We pinpoint each tool's location and its supporting infrastructure. For every tool, the interface with the wafer handling system and WIP storage need to be defined accurately.

After the detailed layout is finalized, you begin the execution process by taking control of project management, construction and tool installation. The tools go through several phases during the ramp, including process capability, tool qualification, process qualification, short-loop process control runs and full-loop runs. Toolsets are checked for redundancy (no one-of-a-kind tools) and are then ready for ramping to capacity in graduated steps.

Note that, although you do not need full automation and wafer transportation capabilities during early ramping, it is wise to run the system from day one and create a process to add new stops for late arrival tools through software switches.

Proper management of the tool installation and qualification phases can ensure that the gains achieved by aggressive design schedules are not wasted during execution. We suggest the "critical chain" project management philosophy, where you identify all activities and their actual duration, and reduce fudge factors across the schedule. The time saved is then divided by three and equally allocated among the activity owner, a central project buffer and the project duration.

It is important to keep in mind that, once a building is ready and the tools are hooked up, you still may not have a best-of-breed fab. There remains a slew of business transactions that need to be defined. To be proactive, you should map all business processes ahead of time and create standard operating procedures (SOPs) to cover them. Typical business processes include shift management, maintenance practices, tool and recipe SOP, WIP management and information systems.

Next, you need to create the right systems to support the business processes. Many organizations do not follow a defined process of selecting a system. You should identify the relevant systems needed, analyze them and then look for vendors who can meet most needs. While it is best to specify off-the-shelf systems, there will be some modules that require custom development. A fab's most common information systems:

• Manufacturing execution systems
• Computerized maintenance management systems
• Enterprise resource planning systems
• Facilities control systems
• Networked visual management systems for cleanroom personnel and management

As hard as it is to imagine now, a boom in new product applications will generate semiconductor demand that will drive current 300 mm pilot and R&D lines toward high-volume manufacturing. The 300 mm lines will need to reach operational levels in less than two years — at least a year less than the typical 200 mm fabs.

The struggle for every new fab design is finding the equilibrium between best facility performance or minimizing operating costs. The fab must be designed for optimal throughput capacity and cycle time. Going through the right fab design and project management process reduces the overall time to capacity and thus allows for a better performing fab at a lower overall operating cost.

If you are going to spend $2B-$3B on a new 300 mm fab, remember that it must deliver a significant competitive edge, the kind only a brilliant fab can. If you are not absolutely sure of the volume demand that will allow long-term cost benefits associated with 300 mm, it is better to invest in reducing your new product development time. As we learn in every industry downturn, there is nothing sadder than a brand new fab operating at 40% capacity.

Author Information

Oded Tal is chief executive officer of MAX International Engineering Group , a global consulting firm that provides operational solutions to the semiconductor industry, specializing in facility design and operational improvement. He is a senior member of the IIE and SEMI, and a member of the MANTECH technical committee.

Phone: 1-201-295-2953

E-mail: marketing@maxieg.com