With a growing market for cellular and personal communications services (PCS) and third-generation (3G) mobile systems looming on the horizon, increased attention is being given to RF power amplifiers. To make more-efficient, higher-output, and smaller-size power amplifiers, designers are using various types of power transistors, including lateral-diffused (LD) MOSFETs, gallium-arsenide (GaAs) metal-semiconductor FETs (MESFETs), GaAs/InGaP heterojunction bipolar transistors (HBTs), gallium-nitride (GaN) high-electron-mobility transistors (HEMTs), and silicon-carbide (SiC) FETs.
LD MOSFETs are beginning to replace traditional silicon bipolar transistors in infrastructure applications, while GaAs HBTs have progressed significantly to make strong inroads into the wireless handset niche. Here, they're beginning to replace the popular MESFETs as viable alternatives. Other structures, like HBTs based on GaAs/indium gallium phosphide (InGaP) and GaN HEMTs, are starting to emerge as well. Plus, more exotic material like SiC has resulted in FETs that promise to deliver high-density RF power-amplifier solutions with better thermal conductivity. In short, several power-transistor topologies are now vying for RF power-amplifier sockets in the 800-MHz to 2.5-GHz wireless communications marketplace.
Readily available voltages of 26 V and higher in basestation transmitters are making life simpler for LDMOS power amplifiers. Bias-current and threshold-voltage drift issues that previously haunted this transistor structure for quite a while have been overcome.
As developers continue improving gain, efficiency, linearity, and reliability specifications, they're adapting to low-cost plastic packages to deliver maximum power for every dollar spent.
For example, Motorola Semiconductor has expanded the use of plastic casing over a wide range of output power levels. Although low-power LDMOS transistors used as drivers were the first to adopt plastic housing, this year Motorola extended that capability to high-power output stages. In addition to ensuring high-power capability in low-cost packages, Mo-torola guarantees low intermodulation distortion (IMD). This answers the demands of newer digital modulation schemes deployed in current and next-generation cellular systems.
In line with that strategy, Motorola recently took the wraps off of a high-power plastic part. The device has a 45-W peak envelope power (PEP) rating and 18.5 dB of gain at 945 MHz. Suitable for applications of up to 1 GHz, this device achieves −31-dBc third-order IMD (IMD3) and a power-added efficiency (PAE) of 41%.
To use plastic packaging effectively requires aggressive thermal management. Aside from using better packaging materials with higher glass-transition temperatures, Motorola relies on clever temperature sensing and compensation techniques to efficiently manage tens of watts of power dissipation. Proprietary sensing and compensation techniques have been implemented on the power chip to keep the die cool.
Inside the package, Motorola in-cludes LC matching networks to raise the input and output impedance of the transistor. But the company is only interested in making it easier for the designer to increase this impedance to 50 Ω or higher. "It's up to the designer to extend that high impedance to the load impedance. That will require less complex circuitry," notes Leonard Pelletier, applications support engineer at Motorola.
Because of user demand for guaranteed specifications, Motorola has streamlined its test and manufacturing methodologies. Meanwhile, as it begins to roll device versions based on its fifth-generation HV5 process, the company is making enhancements in terms of gain flatness, drift compensation, and linearization. High-power LDMOS devices based on the the process are expected to be released sometime next year. According to Motorola, these new devices will compete directly with GaAs FET modules in the high-power arena (Table 1).
Unlike Motorola, UltraRF is exploiting the low-cost and high-density benefits of low-temperature co-fired ceramics (LTCCs). According to UltraRF, a wholly owned subsidiary of Cree Inc., LTCC is comparable in cost to plastic packaging. Yet it provides the inherent benefits of ceramic material, and a smaller size for 3D structures. (Cree acquired UltraRF late last year from Spectrian, a producer of high-power RF amplifier modules for the wireless communications industry.)
But for those applications where size isn't critical, UltraRF offers an alternative alumina-based single-layer, thick-film solution. Although power amplifiers built on alumina substrates are larger than LTCCs, they're less expensive and faster to develop, claims the maker. For this scheme, UltraRF developed a proprietary method of mounting a large ceramic piece on copper to get the best thermal performance out of the power die (Fig. 1). This allows the power module to operate in the −55°C to 155°C temperature range. Reworking techniques have also been improved to accomplish higher production yields and keep production costs low.
Additionally, UltraRF is pushing the efficiency and linearity performance of high-power LDMOS transistors. "Linearity is a key requirement in multi-channel applications," says John Quinn, vice president of marketing and new product development. "For example, the modulation envelope in wideband-CDMA or W-CDMA is a complex signal. You can't afford to distort this signal. Good linearity is extremely important to keep distortion to a minimum," he adds.
A strong proponent of RF LDMOS, Quinn believes that cost, performance, and ease of use are key attractions for wireless basestation applications. "Compared with GaAs FETs, cost for generating power is much lower with LDMOS," he asserts. Meanwhile, UltraRF also is driving the RF envelope of LDMOS to 3.5 GHz, where it intends to compete with GaAs FETs.
Another major supporter of inexpensive RF power packages is the Franco-Italian chip maker STMicroelectronics Inc. The supplier has combined its patented self-aligned cobalt-silicide (CoSi2) process with novel cost-effective surface-mount packages and readied several thermally stable LDMOS discretes with enhanced reliability and ruggedness (Fig. 2).
Unlike conventional DMOS, LDMOS does away with the insulator to isolate the drain from the ground. According to the manufacturer, this feature both enhances the LDMOS FET's electrical and thermal performance as well as lends itself to cost-effective surface-mount plastic packaging.
To achieve high current capability and lower reactive contribution up to 1 GHz, STMicroelectronics has modified the JEDEC-approved PowerSO-10 package. In the re-vised version, two solid leads replace the PowerSO-10's lateral pins. Additionally, the lead frames are scaled down and depressed to cut wire-bond lengths and corresponding reactive effects.
Moreover, the package's high thermal conductivity allows up to 150°C junction temperatures and 150 W of power dissipation. As a result, devices in surface-mount plastic packages can deliver up to 60 W of continuous output power with the correct level of IMD under Class-AB operation. "There's no degradation in performance, and cost is dramatically reduced compared to ceramic packages," explains Ed Gdowik, RF product marketing manager at STMicroelectronics.
Tweaking the CoSi2 process and reducing the gate width, the company also is pushing the performance bar of its RF LDMOS ICs to address the needs of W-CDMA and other emerging cellular bands. STMicroelectronics will sample 2.3-GHz versions by year's end. Plans exist for using plastic packages, but initial introductions will be in ceramics.
Other notable LDMOS suppliers include Hitachi Semiconductor, Philips Semiconductors, RF Micro Devices, Stanford Microdevices, and Xemod Inc. Hitachi, for instance, will extend the capability of its LDMOS transistors to 2 GHz by migrating to 0.5-µm MOS. "The goal is to shrink the geometry while retaining the same breakdown voltage," notes Ted Sato, principal engineer at Hitachi.
"Besides high output power, good efficiency, and linearity, there are some other attributes that make LDMOS attractive," says Paul Patterson, director of the wireless, lightware, and power business unit at Hitachi Semiconductor (America) Inc. "These include ruggedness, good thermal conductivity, stability over a wide temperature range, as well as the elimination of special temperature-compensation circuitry. Compared with GaAs wafers, LDMOS is produced on 6- to 8-in. wafers to keep the cost low."
Hitachi is producing high-power LDMOS UHF amplifier hybrid modules with built-in impedance-matching circuits. It's in the process of making silicon monolithic microwave ICs (MMICs) with integrated passives and LDMOS devices on a single substrate. Prototype power amplifiers based on this technology will be released later this year.
With the right combination of device structure and gate-oxide processing, Xemod has optimized the performance of its LDMOS transistors to provide higher gain per stage, larger safe-operating areas, and increased linearity. Xemod has crafted a full set of QuikPAC building-block power stages for the design and production of 2.5-GHz RF power amplifiers. The modules can be easily implemented into a mi-crostrip assembly, along with other passive and active components, including input and output matching circuitry, to create a complete impedance-matched power amplifier that generates hundreds of watts of output power.
Xemod developed a patented error-correction technique for QuickPAC. It provides self-adjusting feed-for-ward/predistortion IMD correction with a wide dynamic range and wide instantaneous bandwidth to accommodate complex digital signals. The method requires fewer and shorter delay lines to cut single-path losses.
After extending the cutoff frequency of its second-generation LDMOS power transistors to 3.5 GHz, Philips Semiconductors is now concentrating on slashing the drift characteristics of its devices by minimizing hot-carrier injection and shielding the FET's gate and drain sections. In production with its second-generation parts, Philips has successfully demonstrated less than 10% drift over 20 years without burn-in. The company's next goal is to further cut drift to less than 3% over 20 years without burn-in, in 3G devices.
Like many others, Philips has adopted gold-top metallization for attaining higher mean-time-to-failure (MTF) and better reliability for its LDMOS parts. Additionally, Stanford Microdevices leverages simulation and modeling to wring more continuous power out of LDMOS devices at 2.4 GHz.
Meanwhile, exploiting RF properties of new compound semiconductor materials, other transistor structures have also begun to emerge. A potential candidate with prospects to compete against LDMOS is the GaN HEMT.
Nitronex Corp. has combined RF benefits of GaN with a proprietary methodology for fabricating HEMTs on larger low-cost silicon substrates. The company has successfully demonstrated GaN HEMTs on 4-in. silicon wafers using proprietary transition-layer and patent-pending epitaxial growth and deposition technology, labeled Pendeo. Until now, such efforts were limited to 2-in. wafers using silicon-on-insulator (SOI) and sapphire substrates, which are comparatively expensive materials.
"GaN can't be grown by itself as a semiconductor crystal in an economically viable manner because of physical limitations," says T. Warren Weeks, Nitronex's director of materials engineering. "We addressed those issues to provide a commercially via-ble solution for growing GaN on large-area substrates," he adds.
By comparison, the current density and electron mobility for GaN transistors is much higher (Table 2). "This translates to higher device impedance to begin with," notes Rick Borges, director of device engineering at Nitronex. "Therefore, matching to 50 Ω is easier and less lossy. Also, the GaN HEMTs offer high-voltage operation with high breakdown voltage. In addition, the technology can be extended to higher frequencies, all the way up to 30 GHz," Borges explains.
Nitronex is readying GaN HEMTs from 8 W up to 35 W of peak power at the 2.0-, 2.5- , and 3.0-GHz cellular bands. These devices operate from a 28-V drain supply with −3 to −4 V on the gate as a pinch-off voltage. To get the heat out, Nitronex also is integrating a heatsink with the device for better thermal conductivity. The company has set up a pilot production line at its facility in Raleigh, N.C., and expects to deliver prototypes by the end of this quarter.
Similar efforts are in prog-ress at Cree too, where scientists have developed a 50-Ω AlGaN/GaN HEMT using 100- and 150-µm wide and 0.5- to 0.6-µm long gates. A power density of 9 W/mm was achieved in a single-stage Class-AB amplifier. For optimum thermal performance, the amplifier was flip-chip-attached to an aluminum-nitride (AlN) substrate, on which metal-insulator-metal (MIM) capacitors, metal resistors, and air-bridge interconnects were fabricated to complete the amplifier circuit.
Cree continues to advance the performance of SiC FETs for high-power RF applications. SiC has very high thermal conductivity and power density.
Cree has pioneered a Class-B SiC FET capable of generating about 14 W of output power at 1.95 GHz, with 11-dB power gain and 60% PAE. Tailored for CDMA basestations, this linear device offers excellent adjacent-channel power-ratio (ACPR) in high-power RF amplifier design. Although presently the output of the SiC FET is lower than silicon LDMOS or GaN HEMTs, Cree is optimistic about squeezing more juice out of a smaller die in the future.
In the cellular handset arena, where voltages are lower and the output power levels are moderate, the scenario is different. In this space, GaAs HBTs are gaining popularity among power-amplifier designers, especially for CDMA and W-CDMA handset applications, where linearity specifications are stringent. Here, single-supply HBTs are displacing dual-supply GaAs MESFETs. Recent improvement in the emitter structure of the HBT is making it attractive from reliability and manufacturability standpoints as well.
InGaP Replacing AlGaAs
Several developers are replacing the conventional aluminum-gallium-arsenide (AlGaAs) epitaxial layer with InGaP in the emitter section to further simplify the fabrication process and improve temperature stability of the transistor. For instance, M/A-COM, a unit of Tyco Electronics, is tapping these enhancements to fabricate a new generation of GaAs/InGaP HBTs for its RF power amplifiers aimed at the GSM and DCS bands. M/A-COM has built three-stage power amplifiers that operate from a single 3.0- to 5.5-V supply and deliver 35.5 dBm at 50% efficiency. This is all in a plastic microleadframe package that includes 50-Ω input/output matching. To lower the cost of its amplifiers, M/A-COM will soon start producing its new HBTs on 6-in. wafers.
Other vendors exploiting the benefits of InGaP in HBT fabrication include Anadigics, Agilent Technologies Inc., Alpha Industries, EiC Corp, Stanford Microdevices, and Toshiba. Additionally, Motorola is in the process of finalizing its InGaP epitaxial layer and plans to introduce such HBTs later this year. "The secret recipe is in the epitaxial layer," says Mike Civiello, director of marketing for Motorola Semiconductor's Wireless Subscriber Systems Group. "Traditional AlGaAs/GaAs HBTs have long-term reliability issues."
Additionally, Motorola is readying an enhancement-mode pseudomorphic HEMT (E-mode pHEMT) transistor to expand its product portfolio. Offering substantial improvement over conventional MESFETs, "E-mode pHEMTs suppress leakage and eliminate the use of a negative-voltage generator (NVG)," Civiello adds. These are two driving factors for the E-mode pHEMTs, he continues. Leakage current, in fact, is an issue with many other manufacturers in this battle.
The E-mode pHEMT is based on a positive gate bias technique that makes drain current negligible to eliminate the negative supply voltage (Fig. 3). It provides efficiency, gain, and linearity advantages over MESFETs. Motorola is using its E-mode pHEMTs to prepare a dual-band, dual-mode integrated power amplifier (IPA) for GSM handset applications, slated for release this summer.
Handles GSM, DCS Needs
Meeting the output power requirements of GSM and DCS handsets, this dual-band IPA is exptected to offer approximately a 4% to 5% improvement in efficiency over the older MESFETs. Typically, in the GSM mode the E-mode pHEMT based IPA will meet the GSM power requirement of 35.8 dBm, while achieving a PAE of 55%. Likewise, in the DCS mode, it will provide an output of 34 dBm and a PAE of 45%.
Another backer of E-mode pHEMTs for GSM and CDMA handset power amplifiers is Agilent Technologies. In fact, Agilent has transferred its proprietary E-mode pHEMT process to the manufacturing environment. The company is also prepping power modules as part of its RF chipset solution for CDMA applications. Battery life and product size were given utmost attention in developing these power modules, according to Agilent. Concurrently, it is extending the technology to power amplifiers for GSM applications.
In comparison to HBTs, both these players are touting higher efficiencies for their enhanced pHEMTs. And they have demonstrated linearity good enough to challenge both current and new-generation HBTs on the cellular handset turf. However, in applications such as W-CDMA, where the need for linearity is greater, HBTs will continue to dominate, says Motorola. Consequently, suppliers will continue to suport both the technologies in their product portfolios.
Interestingly, Motorola is closing the chapter on MESFETs, but not without making it competitive against others in the ring. For that, the last member of its GaAs MESFET family, the MRFIC1859, comes with an on-chip NVG to alleviate the problem of using dual supplies. "This dual-band (GSM/DCS) IPA implements unique RF rectification circuitry to realize an NVG that's free of spurious outputs," Civiello contends. In his opinion, it marks an end of a MESFET era for Motorola.
Several efforts are under way to pump more power out of SiGe HBTs with comparable efficiency, gain, and linearity in the desired cellular bands. Anadigics and TEMIC Semiconductors make up one team that's close to accomplishing this. Together, the two partners developed a power amplifier in a multichip-module (MCM) format for cellular CDMA modulation. The adjacent channel interference for this module was adequate at 27 dBm of output power, but more work is required to improve its efficiency, linearity, and gain at higher output levels. Furthermore, some developers are investigating the idea of incorporating SiGe HBTs into a biCMOS process to achieve higher levels of integration around the power amplifier.
In summary, RF power transistors enabling power amplifiers for a variety of wireless communications applications come in many flavors. Over time, the choices just continue to increase. No single technology is an undisputed leader for any specific RF wireless communications application. In other words, no one transistor meets all needs. It's the designer's job to make the tradeoffs and compromises that best fit the design and the end-system requirements.
|Companies That Contributed To This Report|
RF Micro Devices
(800) 879-4963, ext. 209