Elements of a Successful Foundry Startup

In Southeast Asia, competition is fierce and rapid time-to-market is crucial. SilTerra was founded in December 1997 as one of the first foundries in Malaysia. The foundry, which produces 0.25 µm CMOS devices on 200 mm wafers, was built in Malaysia in close coordination with LSI Logic's employees. The partnering companies adopted a "copy-exact" methodology to transfer the 0.25 µm technology to a new, state-of-the-art fab.

SilTerra first developed a macro-level strategy that outlined a schedule for developing the company's business infrastructure as well as models for financing and creating a budget; fab construction and startup; deployment of computer integrated manufacturing (CIM) and manufacturing execution systems (MES); and other critical components. Managers and engineers hired in Malaysia and abroad were sent to LSI Logic's campus in Gresham, Ore., for extensive training. After carefully reviewing the plan, the team generated a solid list of major milestones, goals and deliverables. The members then developed a macro-level ramping plan for Fab 1 in Kulim, Malaysia, and began construction in June 1999.

	1. The photolithography area of SilTerra’s 91,000 ft2 cleanroom uses minienvironments and is 100% scanner-based to improve yields and product quality. (Source: SilTerra)

We determined the number of operators and technicians needed for fab ramp-up using output from Visions together with Tefen's benchmark staffing data. The spreadsheet model took into consideration such factors as production levels (wafers/month), number of tools in each bay and average turns/wafer/process flow. The model also considered the unique multi-cultural aspects of Malaysia, such as prayer time and holiday schedules. The result was a quarterly manufacturing technician (MT) projection that showed the required number of MTs for different periods and wafer start levels.

Keeping in mind major milestones, we next developed a detailed micro-level schedule. The main building block was a tool template that incorporated all the activities required for receiving, installing and qualifying each production tool and its support systems. We used LSI's tool template as part of the copy-exact policy.

SilTerra's management set a target period of 40 days for move-in, installation and qualification of each major piece of equipment. Using Tefen's experience with similar projects in the United States, Southeast Asia and Europe, and LSI Logic's strong foundation for the micro-scheduling effort, every tool template was completed and reviewed by Silterra, LSI and Tefen. It was important for engineers and managers to understand the terminology and all the activities shown on the schedule. This helped to facilitate communication and prepared the foundation for seamless integration.

	2. The foundry has the capability of fabricating 30,000 wafers/month (200 mm) at full capacity, and is currently producing 0.25 and 0.22 µm devices. (Source: SilTerra)

Based on the critical path of the micro schedule, we developed a detailed move-in plan. It used information such as crate size, weight, delivery dates and flight schedules supplied by equipment vendors and/or freight forwarders. The most significant outcome of this plan was the move-in sequence. The goal was to move in seven tools during a 24-hour time period. This was to begin as soon as the cleanroom met the specifications for cleanliness, temperature and humidity control. Some contingency plans were developed with alternative move-in paths, equipment and resources.

To support the aggressive micro-schedule, a "do it right the first time" approach was needed. Therefore, equipment that was absolutely necessary for the startup line qualification received highest priority. These first-of-a-kind (FOAK) tools were scheduled to be moved in, installed and qualified before any other tools. IE designed and published phase-in fab and sub-fab layout drawings that identified all equipment and their priorities. Priority tools and their auxiliary components were spaced apart on the layout to facilitate the move-in operations and installation.

Different manufacturing areas were also prioritized to allow cleanroom builders to allocate their resources more efficiently. This included installation of such crucial elements as interior walls, air filters, lights and sensors. The location of each tool was marked on the floor, and tool templates and 3-D mock-up panels were moved in to prepare the ground for equipment installation.

A quick and flawless execution of the tool move-in plan was very important. To minimize the impact of delayed flights, bad weather, transportation equipment breakdown and other unforeseen events, we rented several warehouses close to the fab and equipped them with air conditioning. This allowed the move-in sequence to proceed as planned. While many tools were temporarily stored in these warehouses, major tools such as photolithography scanners were transported directly from the airport to the fab floor.

Details of the move-in plan such as move-in path, moving equipment, de-crating, wipe down and cleaning were clearly defined, and several specifications were developed to capture and formalize these details. We moved in several small tools before the official start to help streamline the process and identify potential problems and "show stoppers."

SilTerra began its move-in activities in September 2000. Workers exceeded their initial goal of moving in seven tools/day. In one month, more than 200 pieces of equipment and their support components, together with hundreds of pumps, panels and other items were moved into their final locations in the fab and sub-fab. These locations had been marked by the IE group based on micro layout.

SilTerra's and Tefen's industrial engineers headed up the move-in planning and execution, and provided around-the-clock coordination of the operation.

One of the major challenges was to ensure that the basic utilities needed to operate the equipment (such as electricity, exhaust and drains) were ready on time. Several construction delays caused the schedule for these utilities to fall behind the tool installation schedule. Initially, the construction was not in line with the micro-schedule, and resources were used for non-critical activities.

Because the manufacturing and scheduling people were in Oregon while construction was underway in Malaysia, communication between these groups became difficult. To resolve these issues and bring construction back on track, a team of manufacturing managers and engineers went to Kulim months ahead of their scheduled return date. During face-to-face meetings, the parties established major project milestones. The construction schedule was revised and brought back in line with the overall milestone schedule.

To minimize the impact of base build construction delays on the overall plan, all efforts turned to providing FOAK tools with the basic utilities. Even within FOAK toolsets, different systems were prioritized based on the date needed and complexity. SilTerra designed contingency plans to accommodate operation-critical systems and those that had a high risk of delay. Considerable time was devoted to ensuring the main fab support systems were operational.

To speed the recovery process, we formed several cross-functional teams, each consisting of construction, hook-up, facility and production engineering representatives. The main responsibility of these teams was to ensure that utilities were available on time, in the right amount and within the specifications set. These measures minimized project delays caused by base build activities.

Installing equipment in a cleanroom environment is a challenging task. Many cleanliness requirements need to be met before installation teams can begin assembling the tool and connecting it to the different types of utilities within the fab. The number of tools to be installed simultaneously within a very short period of time, along with the shortage of skilled labor, made advance hook-up planning a necessity.

SilTerra developed a detailed schedule for "pre-facilitation" of major equipment, making sure all facilities were ready at the point of connection (POC). One of the main advantages of such a schedule was that it allowed the installation and hook-up crew to accurately estimate its workload and balance it against available resources. This plan allowed maximum utilization of scarce resources and was later linked to the overall project micro-schedule for daily update and activity tracking.

Critical materials and those with long lead times were ordered well before they were needed. Prior to the main equipment move-in, pedestals for vibration-sensitive tools such as photolithography and CD measurement equipment were ordered and installed.

During this period, the IEs provided support for ongoing activities including:

• Assigning an IE to each manufacturing module to help with ongoing installation activities.
• Tracking daily activities and updating schedules.
• Raising red flags in case of delays.
• Updating as-build layout drawings.
• Line qualification.

As more equipment became ready for SilTerra's process integration group, the emphasis of daily project review meetings shifted from installation to qualification.

Each manufacturing module was internally qualified before line qualification began. This ensured that equipment in one module would perform well with other tools in that module. The integration group was in charge of this operation. Before equipment move-in, they developed a detailed schedule that incorporated all activities necessary to qualify the line after equipment was installed. It included details of critical activities including short loops, first yielding wafer, process freeze, electrical test, burn-in, reliability and packaging. Focusing on these efforts, SilTerra managed to qualify its production line and produce its first yielding wafer in 89 days, a world record in the semiconductor business. The first revenue wafer was produced two weeks ahead of schedule, and SilTerra is well on its way to ramping up to 30,000 wafers/month by 2002.

SilTerra managed not only to achieve its aggressive milestones, but beat them on several occasions. Manufacturing module qualification was completed one week prior its scheduled date. High yielding wafers were produced on the first qualification lot, and production started 10 days ahead of schedule.

SilTerra's management was committed to building a world-class manufacturing facility in the shortest timeframe in the most effective way possible. Management determined a set of milestones and incentive plans to reward people who contributed the most to the project.

It became evident early on that, with 2500 on-site construction workers, more than 400 engineers and staff and equipment vendors from around the world, communication was going to be a challenge. Without the proper delegation of responsibilities, decisions would be difficult to make and the project would not be successful. Therefore, many tasks were delegated to the managers and engineers companywide. This created a fast decision-making cycle and boosted employee morale. Management commitment and strong leadership are essential factors for the success of any project.

SilTerra promoted a "schedule-oriented" culture, and all activities were measured against the micro-schedule targets. Managers, engineers and contractors were required to know the schedule and major milestones. During the FOAK tool move-in and installation, manufacturing module managers were required to report the status of their modules daily and compare them against set milestones. All contractors drove their efforts toward on-time completion of tasks, and main contractors were required to present their progress reports during the daily operations meeting.

Emphasis on the schedule resulted in proactive problem detection and resolution. Many issues were detected ahead of time by closely monitoring daily activities and comparing them to the schedule.

As the technology partner, LSI provided training to SilTerra engineers and managers. When fab construction began in Malaysia, LSI provided its expertise and helped SilTerra with construction problems. A team of experienced LSI engineers and managers stationed in Kulim helped SilTerra rapidly install and qualify the equipment. These engineers and managers were involved in LSI's Gresham fab startup project, and their experience was invaluable to SilTerra. After production tool installation and line qualification began, the LSI integration group became involved in line qualification activities and issues, and was in constant communication with Kulim.

Before line qualification starts, LSI performed a comprehensive audit of SilTerra's information technology (IT) and CIM. This audit, by LSI's IT experts, resulted in a report on the achievements and problems in the CIM and IT infrastructure.

The broad scope of this project made teamwork a necessity. Various teams of experts were created to handle multiple tasks. For example, move-in and installation of tools in each manufacturing module was handled by a team consisting of equipment engineers, module IEs, module managers, hook-up contractors and facility engineers. Some hook-up meetings were conducted in the sub-fab so that all issues could be covered while looking at the actual item being discussed.

A daily project review meeting was held to discuss progress, issues and problems. Participants included vice presidents, general managers, manufacturing module managers, main contractor (construction and tool hook-up) representatives, and facility and IE group representatives. Meeting daily helped close the communication gap among all parties. After completion of line qualification activities, sessions were converted to daily operations meetings.

SilTerra developed a strong relationship with its vendors, contractors and suppliers. To support the fab, some major equipment suppliers established support offices in the nearby Kulim High Tech Park (KHTP) Business Center.

The factors described above outline a very aggressive, but realistic, fab building project. It was proved that, to succeed in a project of this magnitude, teamwork and management commitment is critical. The success of this project is also proof of the value and contributions brought by the industrial engineering discipline.

We believe that exceptional projects cannot succeed by following conventional methods. In pursuit of success, companies and individuals must develop new techniques and methodologies. They have to plan for all possible situations and accept nothing but success.