Site: Fujitsu Quantum Devices, Ltd.
Kokubo Kogyo Danchi
Yamanashi 409-38, Japan
Date Visited: December 14, 1994
Report Author: P. Shumate
Fujitsu Quantum Devices (FQD), which employs about 800 people, is a manufacturing facility established 10 years ago specifically for transferring microwave and optoelectronic device technology from research labs into production. The current product line includes microwave amplifiers, GaAs ICs, power FETs, photodetectors, LEDs, 1.3 microns and 1.48 microns Fabry-Pérot (FP) lasers, and 1.3 microns and 1.55 microns DFB lasers. These support Fujitsu products in wireless systems, SONET and other transmission systems, ATM and conventional electronic switching systems, magneto-optic storage, computers ranging from PCs to supercomputers, and information-processing systems. In these areas, Fujitsu, Ltd., had sales in 1993 of approximately $30.5 billion, of which $3.2 billion (10.5%) was reinvested in R&D. (1994 values for sales and R&D are 9.3% and 14.7% less than 1993, respectively. The larger decline in R&D than in sales was due to a decision to be more selective in R&D programs.)
The first product at Yamanashi was a microwave amplifier (1984), followed by photodetectors, LEDs and power FETs in 1985, lasers in 1987, and linear lasers for cable TV applications in 1990. Other important facilities include the Aizu plant 200 km north of Tokyo, which manufactures FP access (loop) lasers in coaxial packages using fully automatic alignment procedures, and which assembles and tests HEMTs processed in Yamanashi; the Kofu plant, which assembles butterfly and DIP laser packages with both manual and semiautomatic alignments; and the Suzaka plant, which assembles photodetector modules and receivers.
The Yamanashi plant consists of two buildings, Building 1 (88,300 ft2) for microwave and OE devices, and a new facility, Building 2 (183,000 ft2), for GaAs gate arrays and MMICs. Both buildings have extensive clean-room space: Building 2 has 22,000 ft2 in Class-10, for sputtering, and so forth; otherwise, it has <Class 1,000 for assembly and testing. Building 1 uses <1,000 for wafer processing and <10,000 for assembly and testing. Even in the Class-10,000 areas, there is great emphasis on complete outerwear (hats, face masks, bunny suits, socks, and shoes) and air showers usually reserved for cleaner to much cleaner areas. JTEC panelists feel Fujitsu enforces this to remind workers constantly of the importance of cleanliness and quality control. The facilities are of the highest construction quality and unusually well laid out and neatly kept. Extensive automation is employed in Class-10 areas for wafer handling and other functions. There are also extensive areas in less-clean space for processing of all types, burn-in, and testing.
Seventy percent of the 800 staff members are production workers with high school degrees. They hold two-hour meetings monthly to discuss means to improve reliability. The JTEC panel saw a video tape showing great emphasis on physical activities to promote teamwork. Ten percent of the staff are engineers, 15% are in administration, and 5% are in maintenance. There are formal, internal career development programs, such as a one-year program that develops a researcher into a manager following on-the-job training.
As mentioned above, the Yamanashi plant was specifically designed to facilitate technology transfer. People move from one of the company's four research labs, where they develop the concept through prototypes, to FQD where they work with local engineers to develop the full production process. In addition, research is increasingly carried out at the plant itself to facilitate direct transfer. Dr. Takao Fujiwara was the first (in 1986) to move from the Atsugi lab, where he developed short-wavelength transmission lasers. Dr. Syouji Isozumi worked on 1.3 microns laser yield improvement at Atsugi, and Dr. Katsuhito Shima worked on 780 nm lasers. Preproduction lines are established in the lab to evaluate quality and develop the data necessary for transferring devices into production. No automation is developed prior to transfer.
Automation is begun when production quantities rise to greater than 10,000/month. But the market for specific devices can fluctuate significantly from month to month. First, this makes it difficult to look at the need for automation solely in terms of quantities. Second, it makes certain types of automation more important because of the need, for these devices, to depend less on specially trained production people. Therefore, it is important first to automate steps that are very specific to the device. Fujitsu has an engineering department specifically for planning automation, but the company cooperates with outside companies to design the equipment itself. Fujitsu develops the basic idea, and then goes to the equipment supplier.
The company and JTEC panel representatives discussed the "open tender" process to learn more about how device people are involved in a bid. Basically, the contract is negotiated between NTT and the factory-side system people, and then the system people go to the laboratories for any new devices that may be needed. Sometimes the systems-level researchers at the laboratories are also involved. Any work done at a laboratory is then transferred back to the factory. Since NTT does not guarantee production quantities, the Fujitsu Business Group does an independent market analysis. On the other hand, success in winning one of the bids under the open-tender process implies a large market in Japan, and that is a factor. If a system (or device) developed under this process is to be sold to anyone other than NTT, however, NTT must give its permission first.
No visible or other short-wavelength lasers, including 980 nm pumps, were in production at Yamanashi at the time of the JTEC visit. It appeared that short-wavelength pump lasers may have been under consideration. No GaAs-based optoelectronics were being fabricated except 850 nm LEDs for low-speed data links.
Both the 1,300 nm analog laser and the regular DFB for transmission use the same MQW buried heterostructure, but with some parameters changed depending on the application. These two types of devices are produced in alternate periods of time (months). The cost of these devices is dominated by packaging materials, mainly the isolator.
During the plant tour, the JTEC panel saw a display of the components for Fujitsu's commercial 2.488 Gbit/s (OC-48) SDH long-haul trunk transmission system. The display consisted of eight GaAs ICs for all multiplexing, transmitting, and receiving functions, plus the DFB transmitter, the complete APD receiver (-32 dBm sensitivity, gain <10), and an EDFA pumped at 1480 nm, 70 mW. The system operates to 200 km. All these components are made at FQD. Fujitsu researchers are currently working on the next-generation 10 Gbit/s (OC-192) system.
Fujitsu was ISO-9000 certified last year, but many people in the business side say it is not important for sales. One key customer carries out an equally or more-stringent audit apparently similar to ISO-9000, whereas another carries out one as stringent but unrelated to ISO-9000.
Fujitsu Quantum Devices, Ltd., is a very large and capable facility that is tightly coupled with more basic research carried out at its laboratories. Efficient technology transfer is a central focus. The company's device work, fabrication, and packaging facilities are world-class, supplying devices to transmission, switching, wireless, and computing products that are also among the most advanced of their type available.