RF Semiconductors Take A Variety Of Forms

At the roots of almost all of today's electronic devices lie semiconductors. These integrated circuits (ICs) stem from a variety of semiconductor technologies, which have evolved to satisfy requirements like lower power, less noise, more broadband coverage, or simply the need to squeeze higher integration into smaller, cheaper packages. In reaction to these trends, high-frequency engineers have found new ways to leverage gallium arsenide (GaAs), gallium nitride (GaN), silicon germanium (SiGe), indium phosphide (InP), and other semiconductor process technologies. As these technologies have evolved, however, they have not eliminated the more standard technologies like complementary-metaloxide- semiconductor (CMOS) processes. Instead, process enhancements and advances in optolithography are enabling smaller feature sizes and making CMOS and other technologies capable of tackling new and more demanding applications. The result is a myriad of semiconductors that can serve all aspects of the high-frequency market ranging from the most demanding military and satellite applications to access points for IEEE 802.11x wireless local-area networks (WLANs).

As can be said for most of the engineering world, research into military applications still drives many breakthrough semiconductor developments. At the 2007 International Electron Devices Meeting in Washington, DC this past December, Northrop Grumman (www.northropgrumman.com) announced that it has produced and demonstrated an InP-based high-electron mobility transistor (HEMT) with a maximum frequency of operation of more than 1000 GHz or greater than 1 THz. This device will support the military's need for higher frequency and bandwidth in future communications, radar, and intelligence applications. Tests conducted by NASA's Jet Propulsion Laboratory in Pasadena, CA validated the ultra-fast transistor by measuring a three-stage millimeter-wave-integrated-circuit (MMIC) amplifier at 340 GHz with greater than 15-dB gain.

In recent years, the NASA (www.nasa.gov) Glenn Sensors and Electronics Branch has been developing silicon carbide (SiC) as a material for advanced semiconductor electronicdevice applications. According to NASA, SiC devices have repeatedly demonstrated proper operation at temperatures as high as +650C while conventional silicon-based electronics are limited to +350C. Research in NASA's SiC laboratory is focused on developing the base crystal growth and device-fabrication technologies that are needed to produce a family of SiC electronic devices and circuits.

This past September, the NASA researchers announced that they succeeded in extending the life of this high-temperature chip. In the past, ICs could not withstand more than a few hours of high temperatures before degrading or failing. Yet this chip exceeded 1700 hours of continuous operation at +500C. The SiC differential-amplifier IC may provide benefits to anything requiring long-lasting electronic circuits in very hot environments. In fact, silicon carbide's ability to function in high-temperature, high-power, and high-radiation conditions is expected to enable large performance enhancements to a wide variety of systems and applications.

In the commercial arena, research and development continues to push the evolution of process technologies like SiGe, GaAs, and GaN. To meet wireless-infrastructure needs, for instance, MAXIM Integrated Products (www.maxim-ic.com) recently spawned a SiGe high-linearity, up-/downconverting, 815-to-1000- MHz mixer. The MAX2029 was designed specifically for two-and-a-half-generation (2.5G)/third-generation (3G) wirelessinfrastructure applications (Fig. 1). When configured as an upconverter, this single IC provides a +39-dBm input thirdorder intercept, -71 dBc local oscillation (LO) 2 intermediate-frequency (IF) suppression, and a conversion loss of only 6 dB. As a downconverter, the MAX2029 delivers an input third-order intercept point of +36.5 dBm, a +27-dBm output power at 1-dB compression, 6.5 dB conversion loss, and a 6.7-dB noise figure.

The MAX2029 houses a double-balanced mixer core with an LO amplifier, two baluns, an LO switch, and dozens of discrete components. To ease the filtering requirements of close-in harmonics, the mixer boasts 2RF-2LO performance of -72 dBc (with a -10-dBm RF tone). The MAX2029 offers a DC-to-250-MHz IF range with a 570-to-900-MHz, low-side LO injection range. It supports frequency hopping with an integrated single-pole, double-throw (SPDT) LO switch that has switching speeds of less than 50 ns (typical) and 53 dB of LO1-to-LO2 isolation. The on-board, 0-dBm-drive, LO buffer amplifier provides 3-dB drive-variance control. In doing so, it ensures stable gain, noise figure, and input-third-orderintercept performance over temperature, supply, and input power. Gain-variation performance is 0.2 dB over the -40 to +85C industrial temperature range. The input-third-order-intercept-point spread is +0.4 dB/-0.6 dB over temperature.

In the handset end of the telecommunications market, GaAs has evolved into a very-low-cost solution. Triquint Semiconductor (www.triquint.com), for example, just unveiled its latest generation of module solutions for ultra-low-cost GSM and CDMA handsets in emerging markets. The product introduction includes a dual-band GSM Tx module for use in the Americas, dubbed the TQM6M4028U, and a power-amplifier (PA)-duplexer module for CDMA cellular-band applications, which is known as the TQM613027. At 6 x 6 mm, the TQM6M4028U achieves a very small form factor while addressing the GSM850 and PCS1900 bands. Although data sheets are not finalized for either product, the TQM6M4028U is designed for 50-O systems at all ports. For its part, the 7-x-4-mm TQM613027 should offer up to +25.5 dBm of RF output power while drawing only 40 mA typical current consumption at maximum output power in low-power mode (+13.5 dBm).

GaAs also has broad applicability in other wireless markets, such as millimeter- wave point-to-point radio, local multipoint distribution services (LMDSs), satcom, and very-small-aperture-terminal (VSAT) applications. To serve these markets, Mimix Broadband (www.mimixbroadband.com) has introduced a 38-GHz, surface-mount-technology (SMT)-packaged GaAs chip set. These MMIC receiver and transmitter devices cover the 35-to-45-GHz and 36-to-42- GHz frequency bands, respectively. The XR1008-QB image-reject receiver has a noise figure of 3.5 dB and 10 dB conversion gain with a +5-dBm input third-order intercept. It integrates an image-reject mixer, LO doubler/buffer amplifier, and low-noise amplifier (LNA). In contrast, the XU1006-QB transmitter boasts a +17-dBm output third-order intercept and 5 dB conversion gain across the band. It integrates a balanced image-reject mixer, LO doubler/buffer amplifier, and output RF amplifier. That RF amplifier comprises a series of gain and attenuation stages to achieve linear gain control with a fixed IF input level.

Of course, semiconductors are at the heart of today's plethora of wirelessnetworking standards as well. An example is a single CMOS chip dubbed the WSR601, which targets wireless Universal Serial Bus (USB) host and device applications. This chip, which hails from Wisair (www.wisair.com), makes inroads into mobile devices by providing power-save modes and an average power consumption of 385 mW at 100-Mb/s throughput. The WSR601 is based on the WiMedia and Certified Wireless USB standards. It integrates the Ultra Wideband (UWB) physical layer (PHY) including RF, media access controller (MAC), and Wireless-USB subsystems. By leveraging its PHY capabilities, the single chip vows to deliver full-room coverage with connectivity ranges of 8 m at 480 Mb/s and over 20 m at 200 Mb/s. A two-wire scheme allows the WSR601 to coexist in close proximity to other radios, such as Bluetooth and IEEE 802.11x.

Now that the IEEE 802.11x standards are enjoying widespread success in the consumer market, engineers are pushing them to support more features. The SE2547A and SE2548A are complete IEEE 802.11a/b/g/n WLAN RF frontend modules from SiGe Semiconductor (www.sige.com). Each device includes the power amplifiers, filtering, power detector, diversity switch, diplexers, and associated matching circuitry required for dual-band WiFi systems in a module measuring just 25 mm2. The two modules can be used alone or in stacked configurations to support multiple-stream applications. With two devices integrated into the footprint previously occupied by one, small consumer applications can provide greater throughput and performance in support of emerging wireless multimedia applications like media distribution throughout the home. The SE2548A is equipped with a single antenna port, which allows it to be used in a two-stream-by-two-antenna configuration. The SE2547A provides two antenna ports, thereby permitting twostream- by-three-antenna or two-streamby- four-antenna configurations. Systems based on the SE2547A or SE2548A vow to achieve impressive linearity at transmit power levels of +18 dBm at 2.5 GHz and +17 dBm at 5 GHz.

Clearly, these wireless-inspired semiconductor developments are all working to pave the way for next-generation consumer needs and requirements. Although that same thing can be said for the satellite arena, this area deserves special attention for the performance that must be attained under extremely harsh environmental conditions. For legacy civil, commercial, and military satellite programs, obsolescence in particular is a major and costly problem. Through an agreement with Actel Corp. (www.actel.com), BAE Systems (www.baesystems.com) is hoping to overcome this issue with a field-programmable-gate-array (FPGA)based semiconductor approach. This radiation-hardened FPGA semiconductor promises to enable producers of satellite payloads and instruments to avoid time-consuming and costly redesigns while allowing satellite programs to remain on schedule. BAE Systems will develop, manufacture, and sell the RH1020B FPGA under a license agreement with Actel.

In recent years, Toshiba America Electronic Components or TAEC (www.toshiba.com/taec) has garnered much attention for its research and development into GaN for satellite applications. Advances in Ku-band microwave amplifiers focus on replacing the electron tubes that are conventionally used at this bandwidth with semiconductors particularly GaN devices. This past October, TAEC announced a Ku-band GaN power field-effect transistor (FET) that achieves 65.4 W of output power at 14.5 GHz (Fig. 2). It operates with 30-V drain voltage and features 8.2 dB linear gain.

To achieve such a high level of output power in the Ku-band, TAEC optimized the composition and thickness of the aluminum- gallium-nitride (AlGaN) and GaN layers that are formed on the HEMT structure's highly heat-conductive siliconcarbide (SiC) substrate. To assure high performance, it also applied a shorter gate length of below 0.3 m. The shape of each electrode and element configuration was optimized to enhance heat dissipation. By replacing source-wire bonding with viahole technology, the company reduced parasitic capacitance while improving higher-frequency performance. TAEC actually succeeded in forming viaholes in SiC substrate, which is a highly demanding process. Lastly, the company applied an overcoat process around each gate electrode, which contributes to suppressing gate leakage to one-thirtieth that of Toshiba's conventional approaches.

Going forward, Toshiba intends to continue its GaN technology development for the 18-to-30-GHz frequencies (Ka-band) and beyond. The company also has extended its semiconductor research and development collaboration with IBM Corp. (www.ibm.com) to include 32-nm complementary-metal-oxide-semiconductor (CMOS) process technology. Since December 2005, IBM and Toshiba have partnered on fundamental advanced research related to semiconductor process technologies at the 32-nm technology generation and beyond. Under the new agreement, Toshiba joins a six-company IBM Alliance for 32-nm bulk CMOS process-technology development. IBM and Toshiba plan to accelerate the development of next-generation technology to achieve high-performance, energy-efficient chips at the 32-nm process level.

Such partnershipsalong with the research and development of many individual companieswill continuously ensure that the next process level is successfully reached. At the same time, researchers will steadily apply newer advancements to existing process technologies. For high-frequency engineers, these trends translate into a variety of semiconductor-process options that are always evolving and improving to meet their needs.