Site: NEC Tsukuba Research Laboratories
34 Miyukigaoka, Tsukuba
Ibaraki 305-8501 Japan
Date Visited: 1 June 1999
WTEC Attendees: T. Itoh (report author), J. Winters, M. Iskander, H. Morishita
Hosts: Dr. Kazuhiko Honjo, Manager, Ultra High Speed
Device Research Laboratory
Dr. Toshio Uji, Assistant General Manager, Optoelectronics and High Frequency Device Research Laboratories
Dr. Masaaki Kuzuhara, Senior Manager, Kansai Electronics Research Laboratories
Dr. Kenichiro Suzuki, Principal Researcher, LSI Basic Research Laboratory, Silicon Systems Research Laboratories
NEC is one of the world leaders in electronics and communication and has been subscribing to the concept of C&C (Computers and Communication). This visit is for its Tsukuba Research Laboratory. NEC also provided information on its Kansai Electronics Research Laboratories, which was represented by Dr. Kuzuhara who made a special one-day round-trip from Otsu to Tsukuba for his presentation to the WTEC panel.
NEC's capabilities concerning semiconductor devices have enjoyed a high reputation throughout the world, particularly in the areas of compound semiconductor devices for optoelectronics and high frequency devices and components.
After welcome greetings from Dr. Honjo and the presentation by T. Itoh on behalf of the visiting WTEC study team, Dr. Uji, as a senior executive, made a brief presentation about the corporate structure, R&D operations, and organizational configurations of NEC. The R&D Group is responsible for the Technology for the Day-After-Tomorrow, while the R&D operation of each business group looks after technology for today and tomorrow. The R&D Group employs about 1,600 and works on multimedia systems/software, C&C, functional devices, semiconductor devices, and materials/fundamentals at several locations in Japan and abroad. R&D expenditure is 1% of NEC annual sales for the R&D group and 10% of NEC annual sales for the entire company. Presentations were heard from the Optoelectronics and High Frequency Device Research Laboratory (Dr. Honjo), Kansai Electronics Research Laboratories (Dr. Kuzuhara), and later from Silicon Systems Research Laboratories (Dr. Suzuki on Microswitch).
The Optoelectronics and High Frequency Device Research Laboratories consist of the Ultra High Speed Device Research Laboratory (Dr. Honjo) and three opto-electronics research laboratories. However, when combined with the effort at Kansai Electronics Research Laboratory, the ratio of the effort on high frequency devices is on the par with optoelectronics research. The Ultra High Speed Device Research Laboratory is engaged in research on wireless devices, compound semiconductor digital/analog LSIs and electron device physics. Kansai Electronics Research Laboratories consists of the Ultra High Speed Compound Semiconductor Devices Group (crystal growth/process technology, device physics/simulation, low noise microwave and millimeter wave devices, and high power microwave and millimeter wave devices and millimeter wave MMIC) and the Semiconductor Photonic Devices Group. Tsukuba tends to emphasize design and new materials, while Kansai emphasizes process technology and design. Another unique feature is that Kansai Laboratories are co-located with a business unit for semiconductor manufacturing.
The use of CMOS grade Si is appealing for low cost wireless technology. NEC has developed Si nMOSFET for high power amplifiers in GSM. Under class AB operation, the maximum PAE of a Si MOSFET amplifier at 900 MHz is as high as 62% with output power of 27 dBm. MOSFET technology is an old previous generation device with 0.6 ?m gate. Since the transmission line loss is high due to the substrate, the amplifier design makes use of a conditionally stable device so that the amplifier is still unconditionally stable (loss matching technique).
A new 0.18 ?m gate nMOSFET for Ku band operation was developed with fT of 50 GHz and fmax of 45 GHz. In order to reduce the transmission line loss, two types of modified microstrip lines have been developed. Both use polyimide layers as a low loss dielectric layer insert. In Type A, a microstrip is simply placed on top of the polyimide layer that is in turn placed on the Si substrate through SiO2 and SiON isolation layers. This is a slow-wave type structure while Type B is a thin-film microstrip type with an Al layer insert in the SiON and SiO2 layers between Si and polyimide. Since the long term stability of polyimide is uncertain, researchers are considering BCB materials as well. A reasonable loss of 1 dB/cm was obtained. Both the single-gate type and the cascade type amplifier were built with a gain of about 10 dB and a NF of 4 ~ 5 dB at the Ku band.
Future mobile wireless communications for multimedia will require high speed (> 20 GHz) and low power PLL. To this end, a high speed and low power dual-modulus prescaler IC is needed. This was accomplished with a new 0.1 ?m Double-Deck-Shaped gate (a two stage mushroom gate) GaAs E/D-HEMT with a coplanar based MMIC with 1.6 ( 0.8 mm size with 50 ohm on-chip termination.
High fmax ( 200 GHz) HBT Technology
In order to reduce the base resistance, selective regrowth was used so that base resistance was reduced to one fourth. At the same time pseudomorphic InGaAs graded base was used. In this way, regrowth-performed GaAs-based HBTs achieve low RbCbc leading to an fmax comparable to those of high-performance InP-based HBT that have higher fT. With these device technologies, a 26 GHz HBT power amplifier module (two device power combined) with 3.63 W and PAE of 21%, a 1W 35 GHz HBT power amplifier with PAE of 29%, a 60 GHz dynamic frequency divider, and a low phase-noise 38 GHz HBT MMIC VCO, were fabricated.
Other Topics Discussed
NEC does not currently work on A/D and D/A for software radio.
SiGe HBT has not been developed in the research lab, but is produced in other divisions. The device tends to have a high fT but a low to medium fmax due to high base resistance.
Researchers are working on a GaN device as the next generation workhorse. They obtained an fmax of 90 GHz using sapphire and SiC substrate.
Dr. Honjo believes that for wireless applications up to 100 GHz, InP cannot be considered as a next generation device. This is because GaAs can cover up to 100 GHz while GaN can provide higher power.
The first HEMPT design is for Japanese PDC (TDMA) systems while the later design takes advantage of the 5 GHz bandwidth allowed by the Japanese government for 60 GHz systems. This is an attractive way to develop mobile multimedia wireless applications.
For the new HEMT (both the depletion type and the enhancement type), RON is reduced and PAE increased by low contact resistance and doped recess with the n+-GaAs layer. The enhancement mode device is attractive, as it requires a single power supply. For PDC application of the E-mode device, the Ron = 1.6 ohm.mm and PAE was 67.6% with Pout of 1.15 W when Vd was 3.5 V. For W-CDMA applications of D-mode HEMT, PAE was 54% with Pout of 570 mW at ACPR of -40 dBc at Vd = 3.5 V. PAE can be improved with a low Pout of 20 mW (13 dBm) when an optimized Vd control by a DC-DC converter is used (similar to the work by UCSD). PAE is increased to 21% from 8 % under such a low power output. This is due to the fact that the reduced drain voltage brings the device into saturation.
Millimeter Wave CPW MMIC
The low dispersion and high isolation nature of the CPW is used for millimeter wave (60 GHz) MMICs for potential wireless multimedia applications in conjunction with a flip-chip mount. In order to avoid gate peeling and to increase reliability of the HEMT, SiO2 filling on both sides of the T-gate supports the gate. The air gap for flip chip was determined in such a way that an EM simulator makes the effect of the ceramic substrate (motherboard) negligible.
Other Issues Discussed
Dr. Kuzuhara emphasized the growing importance of antennas, particularly planar antennas integrated on the ceramic substrate for low cost millimeter wave applications. Point-to-multi-point applications require adaptive antennas.
Dr. Suzuki presented a paper reporting on a project of low insertion high isolation silicon RF microswitches on a glass substrate for satellite communication applications. Glass substrate is needed as this is a phased array with a large aperture. Therefore, silicon substrate is not large enough. Each antenna feed unit is 5 mm ( 5 mm, and eight element switches are used to achieve a phase increment of 22.5º. The switch element is made of single crystal Si with a double-hump configuration. One of the humps enables contact, and the other permits electrostatic force generation. The humps are mechanically connected on a single arm but electrically isolated by dielectric material in between. The voltage is 50 V, which must be reduced to less than 20 V. The insertion loss measured was 0.2 dB at 30 GHz.
NEC presented very high level technical accomplishments and the WTEC panel's hosts were very open in discussions. NEC has traditionally exhibited a high level of success, particularly in semiconductor devices and applications. Therefore, there is great confidence, and researchers do not often fall prey to the "me-too" attitude. The company appears ready for not only low-cost next-generation wireless, but also more future looking millimeter wave applications beyond 60 GHz. Although its strength lies in semiconductor technologies, NEC is well aware of other important issues such as on-chip antennas, self-packaging, etc. In addition, there is a long history of developing products based on research. This is expected again in this case. High performance devices alleviate difficulty in circuit design.