Technical Insight
GaAs industry embraces new technology trends
Despite temperatures of over 100 ºF and several no-shows due to the effects of SARS, the attendees at this year's Mantech conference enjoyed a high-quality program covering GaAs and InP manufacturing technology, as well as wide-bandgap devices and the threat posed by silicon.
This year s Mantech conference, held on May 19-22 in Scottsdale, AZ, was opened by Dave Aldrich, CEO of Skyworks Solutions, who described current trends in the wireless market. A key trend is integration towards modules and subsystems, which is driven by shorter product lifecycles and a growing abundance of outsourced handset makers. These companies lack RF design expertise and want to use the lowest possible number of components.
As wireless companies fight for survival, the smaller ones can t afford to pay for R&D in order to integrate their products, and can only compete by lowering prices. This will lead to further consolidation, said Aldrich, who identified five key areas for survival: broad design skills; process and packaging flexibility (figure 1); product breadth and depth; system-level expertise; and manufacturing excellence.
Certain areas of manufacturing need to be internal, since the merchant market can t provide the level of value required. For Skyworks, these include InGaP HBT and GaAs PHEMT processes, and module manufacturing. In other areas, such as low-temperature co-fired ceramic (LTCC) and SiGe, Skyworks maintains strategic alliances with other companies, and also has more straightforward foundry agreements for areas such as CMOS.
Electronic component and system requirements are far more stringent for military applications than in the commercial field. These requirements include the need to operate in harsh environments, to process high-rate sensor data in real time and to operate outside commercial bands, and many others. John Zolper of DARPA described how the US Department of Defense is funding a number of programs to develop highly advanced electronic components. These include the wide-bandgap semiconductor initiative (Compound Semiconductor November 2002), the TFAST program to develop high-speed InP DHBTs (see "InP battles to beat the slump in optical communications"), and the antimonide-based compound semiconductors (ABCS) program. This class of materials is expected to enable extremely low-noise millimeter-wave receivers and sub-one V mixed-signal logic, as a result of their extremely high mobility and saturated electron velocity.
Zolper also discussed the technology for efficient agile mixed-signal microsystems (TEAM) program, which aims to exploit the integration and processing capabilities of aggressively scaled SiGe HBTs. The target is to develop complex mixed-signal circuits such as direct digital synthesizers and analog-to-digital converters, using SiGe HBTs that have already demonstrated ft values of 350 GHz.
Silicon germaniumAlso in the plenary session, Mark Wilson of Motorola compared the cost of using GaAs HBT and SiGe:C BiCMOS technologies in a manufacturing environment (see also Compound Semiconductor June 2003). SiGe can be produced more cheaply than GaAs for three main reasons: substrate cost; the lower cost of running processes in very high-volume production centers; and the use of 200 and 300 mm wafers. However, GaAs is very well established in the power amplifier (PA) portion of wireless handsets, and SiGe:C is actually under threat from 90 nm CMOS, which could drive SiGe out of handsets with the possible exception of the PA driver.
Another paper from Motorola described the emergence of SiGe:C HBT technology. The SiGe base layer is B-doped, and carbon is also incorporated into this layer to prevent the out-diffusion of boron during subsequent thermal cycles (figure 2). The paper reviewed Motorola s 0.35 µm and 0.18 µm SiGe:C BiCMOS platforms, which enabled first-generation ICs in the 2 GHz range, and second-generation devices with ft and fmax values of 123 GHz for the 0.18 µm version. A high-breakdown device was also developed, having a DC breakdown voltage (BVCEO) of 5.5 V.
High-voltage PAsWhile low-voltage operation is a key requirement in wireless handsets, higher voltages are readily available in wireless base stations, satellite ground stations and radar systems. There are many potential benefits of using higher voltages for PAs in such systems - for example, power density is usually increased at a higher voltage, so the device size and hence the cost can be reduced.
D Miller of M/A-COM in Roanoke, VA, compared various high-voltage transistor technologies, ranging from GaAs MESFETs and PHEMTs to silicon LDMOS and wide-bandgap technologies (SiC MESFETs and GaN HEMTs). Table 1 shows some of the characteristics of discrete power devices, including those manufactured with newly developed high-voltage (HV) GaAs technologies. The SiC and GaN technologies have the highest power densities, while Si LDMOS, the most mature technology, has the lowest. However, the much higher cost of SiC and GaN results in a higher cost per watt than LDMOS. Even so, the price of GaN might be competitive with LDMOS if factors such as higher-temperature operation result in a lower overall system cost.
It should be noted that the cost figures in table 1 include some favorable predictions for cost-related improvements in manufacturing for the wide-bandgap technologies in the coming years. Also, the improvements in power density achieved with the HV GaAs processes could allow them to compete with low-cost LDMOS technology.
Compound semiconductors in AsiaThe audience enjoyed several talks about the compound semiconductor industry in various Asian countries, including China, Taiwan and Korea. Unable to travel, Yung Liu of ITRI in Taiwan made his presentation by telephone (at 4.00 a.m. Taiwanese time) while the session chair changed slides. The growth of the industry in Taiwan has been well documented, and the company is one of the top three compound semiconductor producers after the US and Japan, according to Liu. In addition to commercial interests, there are a number of ongoing collaborative programs, for example the Next Generation Lighting R&D Consortium. A number of partners, including Epistar, Forepi, Tyntek and OptoTech, are working towards a target of 50 lm/W for commercial white LEDs and 100 lm/W for R&D devices.
This year, ITRI s Optoelectronic Systems (OES) laboratory launched a nanophotonics program as part of a $700 million national initiative on nanotechnology. The project will feature work on quantum dots and photonic crystals manufactured in III-V materials, and includes a QD laser collaboration with the Ioffe Institute, Russia, that has already yielded 1.3 µm InAs QD lasers with a single-mode output of 400 mW. Also in Taiwan, the $260 million, three-year national Si-Soft initiative is developing system-on-chip technology for wireless, processor and optoelectronic applications.
A joint paper from Tokyo Institute of Technology and NTT Photonics Labs described InP-related activities in Japan, which have been hit by the downturn in the communications industry. Almost all commercial activity is in optical devices, although Japan leads the world in InP microelectronics, having recorded for example the highest ft values for both HEMTs (Fujitsu, 562 GHz) and HBTs (NTT, 341 GHz).
Japan has some of the major suppliers of InP substrates, including Japan Energy, Sumitomo Electric and Showa Denko. While both Sumitomo and NTT-AT supply InP-based epiwafers, many laser manufacturers have internal epitaxy capabilities because growth is an integral part of the fabrication process for DFB lasers. Sometimes the Japanese business approach can hinder commercialization of new technologies; long-wavelength VCSELs were developed by several companies but not released as products because they are potentially competitive with edge-emitting lasers.
As wireless companies fight for survival, the smaller ones can t afford to pay for R&D in order to integrate their products, and can only compete by lowering prices. This will lead to further consolidation, said Aldrich, who identified five key areas for survival: broad design skills; process and packaging flexibility (figure 1); product breadth and depth; system-level expertise; and manufacturing excellence.
Certain areas of manufacturing need to be internal, since the merchant market can t provide the level of value required. For Skyworks, these include InGaP HBT and GaAs PHEMT processes, and module manufacturing. In other areas, such as low-temperature co-fired ceramic (LTCC) and SiGe, Skyworks maintains strategic alliances with other companies, and also has more straightforward foundry agreements for areas such as CMOS.
Electronic component and system requirements are far more stringent for military applications than in the commercial field. These requirements include the need to operate in harsh environments, to process high-rate sensor data in real time and to operate outside commercial bands, and many others. John Zolper of DARPA described how the US Department of Defense is funding a number of programs to develop highly advanced electronic components. These include the wide-bandgap semiconductor initiative (Compound Semiconductor November 2002), the TFAST program to develop high-speed InP DHBTs (see "InP battles to beat the slump in optical communications"), and the antimonide-based compound semiconductors (ABCS) program. This class of materials is expected to enable extremely low-noise millimeter-wave receivers and sub-one V mixed-signal logic, as a result of their extremely high mobility and saturated electron velocity.
Zolper also discussed the technology for efficient agile mixed-signal microsystems (TEAM) program, which aims to exploit the integration and processing capabilities of aggressively scaled SiGe HBTs. The target is to develop complex mixed-signal circuits such as direct digital synthesizers and analog-to-digital converters, using SiGe HBTs that have already demonstrated ft values of 350 GHz.
Silicon germaniumAlso in the plenary session, Mark Wilson of Motorola compared the cost of using GaAs HBT and SiGe:C BiCMOS technologies in a manufacturing environment (see also Compound Semiconductor June 2003). SiGe can be produced more cheaply than GaAs for three main reasons: substrate cost; the lower cost of running processes in very high-volume production centers; and the use of 200 and 300 mm wafers. However, GaAs is very well established in the power amplifier (PA) portion of wireless handsets, and SiGe:C is actually under threat from 90 nm CMOS, which could drive SiGe out of handsets with the possible exception of the PA driver.
Another paper from Motorola described the emergence of SiGe:C HBT technology. The SiGe base layer is B-doped, and carbon is also incorporated into this layer to prevent the out-diffusion of boron during subsequent thermal cycles (figure 2). The paper reviewed Motorola s 0.35 µm and 0.18 µm SiGe:C BiCMOS platforms, which enabled first-generation ICs in the 2 GHz range, and second-generation devices with ft and fmax values of 123 GHz for the 0.18 µm version. A high-breakdown device was also developed, having a DC breakdown voltage (BVCEO) of 5.5 V.
High-voltage PAsWhile low-voltage operation is a key requirement in wireless handsets, higher voltages are readily available in wireless base stations, satellite ground stations and radar systems. There are many potential benefits of using higher voltages for PAs in such systems - for example, power density is usually increased at a higher voltage, so the device size and hence the cost can be reduced.
D Miller of M/A-COM in Roanoke, VA, compared various high-voltage transistor technologies, ranging from GaAs MESFETs and PHEMTs to silicon LDMOS and wide-bandgap technologies (SiC MESFETs and GaN HEMTs). Table 1 shows some of the characteristics of discrete power devices, including those manufactured with newly developed high-voltage (HV) GaAs technologies. The SiC and GaN technologies have the highest power densities, while Si LDMOS, the most mature technology, has the lowest. However, the much higher cost of SiC and GaN results in a higher cost per watt than LDMOS. Even so, the price of GaN might be competitive with LDMOS if factors such as higher-temperature operation result in a lower overall system cost.
It should be noted that the cost figures in table 1 include some favorable predictions for cost-related improvements in manufacturing for the wide-bandgap technologies in the coming years. Also, the improvements in power density achieved with the HV GaAs processes could allow them to compete with low-cost LDMOS technology.
Compound semiconductors in AsiaThe audience enjoyed several talks about the compound semiconductor industry in various Asian countries, including China, Taiwan and Korea. Unable to travel, Yung Liu of ITRI in Taiwan made his presentation by telephone (at 4.00 a.m. Taiwanese time) while the session chair changed slides. The growth of the industry in Taiwan has been well documented, and the company is one of the top three compound semiconductor producers after the US and Japan, according to Liu. In addition to commercial interests, there are a number of ongoing collaborative programs, for example the Next Generation Lighting R&D Consortium. A number of partners, including Epistar, Forepi, Tyntek and OptoTech, are working towards a target of 50 lm/W for commercial white LEDs and 100 lm/W for R&D devices.
This year, ITRI s Optoelectronic Systems (OES) laboratory launched a nanophotonics program as part of a $700 million national initiative on nanotechnology. The project will feature work on quantum dots and photonic crystals manufactured in III-V materials, and includes a QD laser collaboration with the Ioffe Institute, Russia, that has already yielded 1.3 µm InAs QD lasers with a single-mode output of 400 mW. Also in Taiwan, the $260 million, three-year national Si-Soft initiative is developing system-on-chip technology for wireless, processor and optoelectronic applications.
A joint paper from Tokyo Institute of Technology and NTT Photonics Labs described InP-related activities in Japan, which have been hit by the downturn in the communications industry. Almost all commercial activity is in optical devices, although Japan leads the world in InP microelectronics, having recorded for example the highest ft values for both HEMTs (Fujitsu, 562 GHz) and HBTs (NTT, 341 GHz).
Japan has some of the major suppliers of InP substrates, including Japan Energy, Sumitomo Electric and Showa Denko. While both Sumitomo and NTT-AT supply InP-based epiwafers, many laser manufacturers have internal epitaxy capabilities because growth is an integral part of the fabrication process for DFB lasers. Sometimes the Japanese business approach can hinder commercialization of new technologies; long-wavelength VCSELs were developed by several companies but not released as products because they are potentially competitive with edge-emitting lasers.
