Site: Oki Electric Industry Co., Ltd.
Semiconductor Technology Laboratories (STL)
Research & Development Group
550-5, Higashiasakawa-cho, Hachioji-shi
Tokyo 193, Japan
Date Visited: November 16, 1994
Report Author: D. Keck
Oki Electric is a very old company -- its motto is "Progressive Since 1881." The company has annual sales of about $4 billion and spends $200 million on R&D. The thrust areas for the company are communications, switching, and computing. For the two years preceding the JTEC visit, Oki Electric was shrinking due to the recession, but management believed that the company had stabilized at then-current staffing levels.
About 10 to 20% of the company's activities are in the research phase of the innovation cycle. Optoelectronics products are sold through two business groups: the Electrical Component Business Group and the Transmission Network System Business Group.
The R&D Group is divided into 5 laboratories: Semiconductor Technology (STL), Micro Systems, Multimedia, Electronic Communications Systems, and the Kansai Laboratory (see Fig. Oki.1). Total personnel in these groups is about 400. The Semiconductor Technology Laboratory has about 100 people, approximately 15 to 20% of whom have PhDs. Our hosts estimated that about 60 scientists in the R&D group, or about 15%, work on optoelectronics.
Within STL are six projects or departments. Areas of investigation for each are listed in Figure Oki.1. Dr. Kamijoh used a graph (Fig. Oki.2) to describe how the various organizations interacted. Plotting on one axis the continuum from research to production and on the perpendicular axis the continuum from devices to systems, he identified STL as doing device research. This group views the Electrical Component Business Group as its primary customer. Kansei Research Lab primarily does systems research for Transmission Network System Business Group. Information is shared between research and development people by direct meetings about once a month and frequent faxes and phone conversations.
Figure Oki.3 shows the major Oki projects in the three stages of innovation the company recognizes: research, development, and production.
Within the Optical Communication area, the company has in production 150 Mbit/s to 2.4 Gbit/s transmission systems, an Er-doped fiber amplifier transmission line, a 2.4 Gbit/s FDM system, and an active double star system. The company's scientists and engineers are developing a passive double star system (current NTT architecture) and 40 Gbit/s TDM and WDM systems. In the research phase they are investigating soliton transmission, 160 Gbit/s TDM, a switched access network, and a high-density WDM ring network.
The switching (photonic) area has a parallel board-to-board interconnection system in production, and an optical ATM switch in research.
In optical computing, STL has in production an LED printer and a 400 dpi LED array. Its research topics involve interconnection: board-to-board, chip-to-chip, and gate-to-gate.
Devices in production include both 1300 nm and 1550 nm Fabry-Pérot and DFB laser diodes, high-power 1480 nm lasers, semiconductor optical modulators, 1300 nm superluminescent LEDs, and polarization-insensitive semiconductor optical amplifiers. In development are 980 nm high-power lasers, large-scale waveguide matrix switches, waveguide wavelength filters, monolithic multiwavelength lasers, and multi-electrode tunable lasers. Research topics include ultrafast optical pulse generation, wavelength conversion devices, multimaterial OEICs, and surface-emitting lasers.
Fig. Oki.1. Oki R&D group.
Fig. Oki.2. Photonics business-development-research map.
Fig. Oki.3. Oki research-development-production.
Dr. Yukihiro Ozeki described transmission systems R&D. He returned from studying in the United States 3 years ago, having earned his PhD from Prof. C. L. Tang at Cornell. He is located at Oki's downtown Tokyo laboratory, which develops systems from various Oki components. He indicated that Oki entered the low-bit-rate systems business in the mid-1970s, selling to NTT. Since he returned to Japan, he felt business was much more competitive and international. Oki's system sales are $60 to $70 million, mostly to NTT, some to Honduras or Southeast Asia; none are to the United States.
Oki's scientists are working on a 10 GB/s system that they believe will be the backbone of the future. Strategically, they divide their work into high-speed (HS) and low-speed (LS) blocks. The topics within the HS block area are optical transmitters and receivers, where the specific topic areas are external modulation, optical amplification, dispersion compensation, WDM, adaptive filtering, and optical nonlinearity suppression. Also within the HS block area are IC/LSI devices consisting of work on GaAs MESFET IC/LSIs (0.3 microns) and dielectric resonator filters. Photonic devices such as MQW lasers, external modulators, APD and PD, and tunable filters are a third section within the HS block area. Also in that area are optical amplifiers, where the researchers are looking at 980 nm pump lasers, isolators, and waveguides. High-speed packaging work completes the HS block area. Less was said about the LS block, where topics were broadly grouped into power consumption, high-density packaging, and flexibility.
The JTEC team's hosts described specifics within the HS block topic area. Oki researchers envision electrically multiplexing sixteen 622 Mbit/s channels to get the 10 GB/s stream. The chips used will rely on some 0.3 microns MESFET technology for the very high-speed parts, but mostly 0.5 microns technology. For the optical transmitter, they are looking to either LiNbO 3 Mach-Zehnder or electro-absorptive (EA) modulators. They favor the EA modulator approach monolithically fabricated with their DFB laser. Their unit will include an aspheric collimating lens, an optical isolator, and a ball lens integrated into the fiber connector assembly. This product is in development now, and they hope it will be ready in two years. Their optical receiver also uses a connectorized ball lens to input to a pin photodiode. This is mounted to a GaAs IC preamp. They expect to use an EDFA preamp before the electro-optic module. Without the EDFA, they have a receiver sensitivity of about -15 dBm. With the EDFA, this is improved to -33 dBm at 10-11 BER. They showed a very complete mapping of all the nonlinear limitations to fiber transmission systems. At the time of the JTEC visit, Oki researchers were building a 10 GB/s system prototype for NTT. They showed pictures of the unit housed in a 7ft high rack.
In the low-speed block area, Oki was selling 10,000 156 Mbit/s systems annually. These are used primarily as optical interconnections for telephone switching systems. The price was estimated at $1,000 to $2,000 for a transmitter/receiver pair. Oki's competitor is Hitachi, but Oki is selling it some systems for private label. Oki researchers are working, however, on a board-to-board 12-channel, 150 Mbit/s unit for the subscriber network, with a target cost of about $20/channel.
Oki has been working on optical amplifiers since 1986. Its researchers began looking at a fiber Raman amplifier and developed a high-power 1480 nm pump laser for it. They then sold these pumps worldwide for EDFAs and for a time believed that they were number one in the business. Now with the shift to 980 nm pumps, they are not number one but have a program to introduce this product. Oki sells an EDFA product that uses beam optics and splitters. It buys the Er fiber and assembles the unit. Its scientists are working on gain-flattened amplifiers and were achieving flatnesses to about 3 dB over a 30 nm band. Oki, along with Hitachi, Fujitsu, and NEC, are producing EDFAs for NTT for 10 GB/s systems. Typically, NTT gives target specifications and then the companies work to meet these.
Dr. Tatsuo Kunii described some of Oki's device work on laser arrays and EA modulators. (He also is a Cornell PhD graduate who studied under Professor Tang.) Approximately 20 people are working on 6 - 7 telecommunications device projects. They see the EA modulator as key to 10 Gbit/s systems. They are using a polyimide buried ridge structure for the modulator. Their work shows a modulation index, a = 0.3, which is required for the Japanese systems. Current performance is maximum extinction ratio > 40 dB, polarization dependency < 0.7 dB, extinction efficiency > 10 dB/V, insertion loss 10 to 12 dB, coupling loss 6 to 7 dB, and propagation loss 2 dB/100 microns.
Dr. Kunii described Oki's multiwavelength DBR laser work. By changing mask width during deposition, laser wavelength can be changed controllably. He showed 4 lasers with 5 nm peak wavelength separation and 6 to 8 mW output. Oki's competitors are NEC and Hitachi. Each is doing the same device research in anticipation of the next NTT request.
Oki's researchers are investigating methods to generate ultrafast pulses. This work is expected to be part of the MITI Femtosecond Project. Oki views the following hierarchy to generate pulses in laser diodes: gain-switching, 10 GHz; external modulation, 30 GHz; active mode-locking, 40 GHz; and passive mode-locking, 100 Ghz. Oki's scientists are investigating passive mode-locking and reported results they believe are the best in the world at the Semiconductor Laser Conference in Hawaii. Usually the mode locker cavity length controls the laser frequency, but if current is injected and modulated within the cavity at a harmonic of the cavity frequency, the mode-locked pulse width can be dramatically reduced. They reported 650 fs pulses by operating at the 40th harmonic of the basic 40 GHz cavity frequency. Output power remains high at 15.7 mW, and depth of modulation is 60%.
Dr. Kamijoh described Oki's work in optical switching using principally LiNbO 3 Y-branch technology. The company's scientists are using curved branches and electrodes for reduced losses together with steep crossings. They have made a 32 x 32 device optimized for size that is approximately 5 to 7 cm long and have worked up a chart of optimized size for various port counts. Acousto-optic devices are also being studied in LiNbO 3. Dr. Kamijoh indicated, however, that he doesn't think only LiNbO 3 will be the winning technology, because with semiconductor devices amplification can be built into the device.
This work is part of the Oki Real World Computing (RWC) effort and is being done in a distributed or company laboratory. A young researcher, Hiroshi Wada (with Oki for 9 years), is working on direct wafer bonding. The researchers envision accomplishing interboard free-space connections using InP SLEDs. While they are looking at epitaxial growth of GaAs on silicon and have made much progress, direct bonding is giving more success. In this area they have done InP on GaAs and have made a laser. They have also done InP on silicon. Basically clean wafers are placed together and heated to 450 to 700 deg. C, giving a sufficiently strong bond for subsequent processing. Their typical wafer size is 1 x 1 cm, and their largest is 2 x 2 cm. This seems to be the state of the art and has also been accomplished at UCSB and Bellcore. Bubbles occur with any larger wafers, which degrade the bond strength.
Oki researchers have fabricated InGaAsP lasers on GaAs. The light output as a function of current is identical to grown materials; however, the drive voltage is 0.3 to 0.4 volts higher, due to band offset between InP and GaAs. Work is proceeding on InP on silicon, but the TCE is giving a problem. If bonding is done above 550 deg. C, the crystal quality is degraded, and if done at lower temperature, the bond strength is degraded. They envision that optical technology will have to connect at least 1,000 pins to be effective.
Oki expects to be involved in the MITI Femtosecond Project. The company sees this work as taking place at two locations: the current OTRC laboratory in Tsukuba and the member company or distributed laboratories. Oki states that the rationale for private companies supporting this arrangement is that they cannot afford the risk for these projects. Company representatives indicated to JTEC panelists that they would envision sending basic research people to the central laboratory and conducting device research in their own laboratory.
Oki representatives indicated that for the Kansai project, "tomorrow" would start about 1996 and end in 2010. They stated that the year 2010 has been selected as the completion date for the Japanese network because predictions suggest that with the aging of the population, insufficient savings will exist to supply the needed capital.