For the PDF version of this article, click here.

While backers of RF LDMOS power transistors continue to make gains in frequency, performance and cost to further strengthen their position in the wireless infrastructure space, developers of compound semiconductor material gallium nitride (GaN)-based heterostructure FETs and high electron mobility transistors (HEMTs) are also reporting progress in delivering high frequency, high power and broad bandwidth for emerging wireless protocols like WiMAX and other broadband applications.

For instance, early proponent Nitronex has just released its first commercial products for the WiMAX market based on its unique GaN on silicon (SIGANTIC) technology. The family of WiMAX power transistors will initially consist of 10 W and 50 W devices supporting the 2.5 GHz and 3.5 GHz segments of the WiMAX market. Target customers are OEMS developing RF power amplifiers for fixed and mobile WiMAX infrastructure. The devices are designed to support broadband operation in these segments (2.3 GHz to 2.7 GHz and 3.3 GHz to 3.8 GHz) to allow greater design re-use and flexibility for OEMS. Meanwhile, efforts are under way to introduce devices for the 5.8 GHz segment of the WiMAX market this year. “The release of these products marks an important milestone in the semiconductor industry. It is the first commercial introduction of a GaN-based device grown on silicon for the wireless infrastructure market, said Chris Rauh, vice president of marketing for Nitronex.

The NPT35050 (3.5 GHz, 50 W) is the first device in this product line to be released. It is designed for WiMAX power amplifier output stages. GaN's higher power density enables the design of packaged devices with higher input and output impedances (20 Ω), eliminating the need for output matching, and making the device easy for power amplifier developers to use. Using a typical single-carrier OFDM (802.16d) waveform, the performance at Vds of 28 V and Idq of 1000 mA is 5 W of average output power with an EVM of 2%, efficiency of 18% and gain of 10 dB.

Korea's RFHIC Company is an early adopter of NPT35050 for its upcoming line of high-power modules aimed at the WiMAX, WiBro BTS and repeater power amplifier markets. Additionally, a team of researchers from Nitronex disclosed new achievements at last year's IEEE International Electron Devices Meeting (IEDM). The company demonstrated a pulsed power output of 368 W at 2.14 GHz, with 70% drain efficiency and 17.5 dB of small-signal gain.

Besides Nitronex, others in this race include Eudyna Devices, Oki Electric, Cree Research, RF Micro Devices, and TriQuint among others. Like Nitronex, Japan's Oki Electric is also taking the GaN-on-silicon approach. According to Oki, its GaN HEMT uses a recessed gate structure to offer a 56 GHz cut-off frequency and an fmax of 115 GHz.

Eyeing WiMAX, Eudyna Devices has released three products in each of the 2.5 GHz to 2.7 GHz and 3.4 GHz to 3.6 GHz bands, with output power ranging from 30 W to 180 W. While Nitronex and Oki are taking the silicon route to cut cost, Cree Research has chosen silicon carbide (SiC) substrate to make reliable GaN HEMTs more competitive. Using electric field modification by field plates, Cree scientists have boosted power density to >30 W/mm up to 8 GHz, which was reported at last year's IEDM conference. Although a field plate adds parasitic capacitances to the device and reduces gain, miniaturization of the gate and the field-plate dimensions could cut back the overall capacitances and push the field-plated GaN HEMT into the millimeter-wave region, according to Cree's paper given at IEDM. In fact, efforts continue to improve the reliability of these devices in the millimeter-wave spectrum.

Similarly, RF Micro Devices is improving the reliability and manufacturability of GaN on SiC HEMT transistors (Figure 1). Optimization of the device process and structure for high power operation continues. Key specs include 28 V, 0.5-0.8 A/mm Idss with more than 50% peak power-added efficiency (PAE). Leveraging its experience and knowledge in GaAs production, RFMD is planning to take its GaN HEMTs to production in 2007. Target applications for RFMD's GaN power transistors include WiMAX, WiBro, WCDMA, GSM base stations, and military MMICs.

Meanwhile, TriQuint Semiconductor Inc. has been awarded a multiyear contract from the Defense Advanced Research Projects Agency (DARPA) to develop high-power wideband amplifiers in GaN. The contract is in two phases. The first phase, lasting three years and valued at $15.8 million, covers the development of gallium nitride material and devices to improve performance and reliability. The second, optional phase covers years four and five and is valued at $15.9 million. This phase will develop GaN high-power, wideband amplifiers and package technology for insertion into Department of Defense (DoD) systems. TriQuint (Hillsboro, Ore.) is teaming with BAE Systems, Emcore Corporation, II-VI Incorporated, Lockheed Martin and Nitronex on the program. University partners Michael Shur of Rensselaer Polytechnic Institute and Jesus del Alamo of the Massachusetts Institute of Technology are also participating.

With ever shrinking process geometries, CMOS processes offer tremendous design flexibility for digital and analog circuits. With recent advances, it has become apparent that CMOS can also provide a wide range of RF capabilities. This is reflected in the diversity of markets served by the various semiconductor makers now developing RF capabilities implemented on CMOS.

One approach followed by several vendors is to develop RF capabilities for existing CMOS processes, and then to enhance those capabilities with established core competencies. Microchip, for example, has well-developed digital processing and data storage architectures in its microcontrollers, which can support wireless protocols such as ZigBee. Manufacturing different capabilities on the same process does not necessarily warrant full integration, however. Kobus Marneweck, a senior manager of strategic marketing for Microchip, stated chipset solutions can often provide more flexible RF solutions, especially when addressing processing and memory requirements for the evolving ZigBee protocol. The same processes used for microcontrollers are also used for RF CMOS, which in the future could be used for optimized, integrated microcontroller plus RF solutions. Both Chipcon, recently acquired by Texas Instruments, and Ember have opted for full integration and offer system-on-chip ZigBee solutions. For low-frequency RF applications, Microchip has recently developed a discrete analog front end, the MCP2030, which features three-axis reception and programmable antenna tuning capacitance. In the PIC16F639, this analog front end is interfaced to a microcontroller die in a single package.

Qualcomm approaches CMOS RFIC development with its strength in systems engineering. That is, radio, baseband and software are optimized to provide the best solution. Recent product developments include the RFR6155/RFT6150 chipset in 0.25 µm RF CMOS. According to Jim Tran, senior director of product management, this is the industry's first dual-band chipset with GPS. Another recent product, also implemented in 0.25 µm RF CMOS, is the TR6275. Said Tran, this product is another first, being a single-chip transceiver that supports the GSM, GPRS, EDGE and UMTS protocols. While earlier designs incorporated SiGe process technology, Qualcomm has completely migrated to RF CMOS. According to Tran, the performance achieved in SiGe has been maintained in RF CMOS through architectural innovation. One example of a performance-enhancing feature is called receive diversity, in which a signal is captured by two antennas and reconstructed with greater accuracy. This function is fully integrated into the recently released RFR6500. To enhance performance of the transmit function, Qualcomm will move to transmitters with polar architectures.

In applications where cost is the driving factor, as it is for ultralow-cost wireless handsets, high levels of integration are desirable. Silicon Labs has met this challenge with its AeroFONE, the Si4905, which is essentially a cell phone on a chip. Implemented in a 0.13 µm CMOS, this device includes an ARM 9 processor core, serial and parallel interfaces, a digital baseband processor and a full RF front end (Figure 2). It also includes an integrated dc-dc converter that connects to an external inductor. This design required high levels of subsystem noise isolation, and the ability to do so is a major asset within SiLabs' IP portfolio, said James Kimery, marketing director of wireless products at SiLabs. Another interesting aspect of the AeroFONE is the on-chip digitally controlled crystal oscillator (DCXO), which replaces the temperature-compensated voltage-controlled crystal oscillator (TC-VCXO) modules used in earlier GSM handsets. Kimery stated that the oscillator circuit, featuring a varactor and capacitor bank for fine and coarse tuning, is calibrated against the reference signal embedded in the GSM base station signal. Once calibrated, the oscillator can synthesize the required RF output frequency to within the required 0.1 ppm accuracy using only a single external crystal. This aspect highlights another trend in CMOS RFIC technology, which is the ability to replace specialized discrete components or modules with on-chip circuitry.

RF Micro Device's Polaris 2 radio module uses a purely digital technique implemented in CMOS to eliminate the transmit SAW filter. Brent Wilkins, a marketing manager for RFMD, explained that this technique generates phase and amplitude information fed to a saturated GaAs HBT power amplifier. Operating the power amplifier in saturation mode reduces power consumption. To allow the baseband processor to use the smallest available CMOS process geometry, all analog circuitry is kept in the RF section of the module.

Recent CMOS RF components from Analog Devices also streamline the use of external components, with a special emphasis on programmability for flexibility and accuracy. For example, the ADF7020-1 ISM band FSK/ASK transceiver IC features an internal temperature sensor unit with a ±10 °C accuracy that can be further improved with a single-point calibration. According to Doug Grant, director of business development for RF and Wireless Systems at ADI, the proprietary automatic frequency control (AFC) permits coarse and fine adjustment of the crystal oscillator, thereby supporting the use of lower tolerance crystals. Grant stated that while Analog Devices has other products with integrated inductors, such as the ADF4360 series, the use of an external inductor allows the RF oscillator of the ADF7012 UHF transmitter to support a frequency range of 50 MHz to 1 GHz yet operate within a specific RF band.

One issue with integrating inductors and capacitors in CMOS is the impact of substrate leakage on the quality factor Q of resonant circuits. There are also parasitic losses through active devices at high frequencies, limiting the performance of conventional CMOS switches at high frequencies. To address these issues, a variation to the conventional CMOS process known as silicon-on-insulator (SOI) has been developed. Peregrine's SOI process, UltraCMOS with HaRP technology enhancements, has been the basis for the PE42660 and PE42672 RF switches and continues to advance. The SP6T configurations have an IP3 of 70 dBm, and it is possible to expand to SP8T or SP9T while meeting an IP3 of 65 dBm with margin. While currently at 0.5 µm, which provides a maximum frequency of 50 GHz, the process will be scaled to 0.25 µm in 2006, pushing the maximum frequency to 100 GHz, according to Rod Novak, vice president of marketing for Peregrine. This will not only have the benefit of smaller die, but will improve the on-resistance, off-capacitance, insertion loss and linearity of the switches. While the RF switches are primarily designed as front-end antenna switches for cellular handsets, Peregrine's digital step attenuator product, a switch-tapped resistive divider, is a solution for wireless base stations and broadband communications. Novak observes the only competition for this product in those applications is an EM relay.

Another modification to the CMOS process to enable high-quality RF switching is to accommodate the fabrication of MEMS structures, such as WiSpry's line of RF switches. WiSpry has also developed digital MEMS capacitors, which were successfully manufactured at Jazz Semiconductor's 200 mm wafer fabrication last year.

These MEMS structures, like the RF switches, are electrostatic actuators. On-chip voltage multipliers allow the devices to be operated from a standard 2.6 V CMOS-compatible supply. They can be considered as highly precise digital versions of analog varactors, according to Jeff Hilbert, president and CEO of WiSpry. Their structure has been defined as a standard design cell with adjustable parameters, allowing them to be configured as tunable RF filters and matching networks. They may also lead to new types of circuits, Hilbert continued, such as frequency synthesizers without a traditional feedback loop. They are fabricated on a process technology that is a combination of basic CMOS with added layers for MEMS. An additional passivation layer protects the MEMS structures during the standard CMOS packaging process.

The broad range of capabilities realized in CMOS RFICs presents many design possibilities for the future.

On the passives front, the pressure to provide higher performance in smaller form factors while meeting environmental directives continues unabated. As a result, passive component manufacturers continue to innovate at a rapid pace. Passive components are being readied for the implementation of the European Union's (EU) restriction of hazardous substances (RoHS) directive 2002/95/EC. The RoHS directive restricts the use of six hazardous substances such as lead (Pb). This restriction becomes effective this year for products shipped to the EU. In the near future the directive will be extended to include other countries such as China. In addition, some states such as California may also adopt similar standards this year.

In the directive, the restriction on the use of hazardous materials is by weight percent. With the exception of cadmium (Cd) the allowed permissible weight percent is 0.1% by weight for a ‘homogeneous material.’ The component industry is interpreting ‘homogenous material’ in at least two different ways. The first interpretation is that the percent by weight of the restricted material for an entire component must be less than 0.1%. The second interpretation reads homogenous material to include each portion of a component. For example, the solder can only include 0.1% or less of lead since the solder may be considered a homogeneous material. However, by considering the solder as part of the component, the entire component may have percent weight of less than 0.1% of lead.

The producer of OEM equipment that has many such passive components may need to consider such detailed issues as the producer (the one who sells a product under a product label) who may be liable for violations of the directive. The RoHS directive allows for fines and other costs for selling products that do not conform to the weight percentage.

Recent product announcements from the component industries address such concerns. For example, Vishay Intertechnology recently announced that it has added the RoHS-compliant 153 CRV series to its range of vertical SMD aluminum electrolytic capacitors. Devices in the 153 CRV series are compatible with lead (Pb)-free soldering processes and offer long useful operating life from 200,000 hours at 40 °C, Vishay claimed.

On the product miniaturization front, passive component manufacturers continue to cut equivalent series resistance (ESR) to support higher data rates. For example, KEMET introduced a 7 mΩ T520 (conductive polymer/low ESR), in addition to four new 45 mΩ T520C (6032-28) part types. KEMET also introduced what it claims is low ESR for high-speed microprocessor applications such as six 7 mΩ T520 part types.

AVX Corp. has introduced the Advanced BestCap capacitors, a generation of low ESR double layer capacitors (DLCs), that addressed the high ESR and loss of capacitance in kHz frequency range in older DLCs by employing features such as polymer separators and carbon electrodes, according to the company.

Meanwhile, niobium oxide and polymer conductors continue to make inroads on tantalum as the material of choice for capacitors with space constraints (Figure 3). For example, AVX recently developed tantalum polymer capacitors that deliver the highest capacitance value capability. Designated the TCJ series, the newest capacitors are one to three capacitance value (CV) levels higher than other devices on the market. AVX can now offer 10 uF 6.3 V in 0805, 47 uF/6.3V in 1206 and 100 uF/4V in 1206 size.

In support of the ever-decreasing form factor sizes, baluns have been reduced in size. Baluns are used to couple balanced networks with unbalanced networks. Typically, they were based on transmission lines and were large in size. Anaren Microwave introduced baluns that are in 0404 in addition to the 0805 and 0603 package sizes. The low profile of these baluns makes them suitable for module designs such as SIP solutions, said the maker. In addition to low height, the baluns offer fine pin pitch, excellent electrical performance, and low price. The 0404 baluns feature a land-grid-array (LGA) interface to occupy only 0.0016 in.² when mounted on a printed-wire-board (PWB) substrate. The baluns, which operate past 5 GHz, are suitable for 2.4 GHz Bluetooth and IEEE 802.11b/g wireless local area networks (WLANs), 3.5 GHz WiMAX, 5 GHz IEEE 802.11a WLANs and home cordless applications, and a range of consumer applications. Insertion loss is typically 0.6 dB at 2.4 GHz with better than 23 dB return loss at that frequency.

To address ever-increasing demands for reliability in cellular connections, CTS Corporation introduced the Monoblock LR series of low ripple ceramic filters. These filters are specifically designed to meet the requirements for consumer cellular repeaters.

A cellular repeater solves communication problems in areas with poor signal reception such as inside buildings or outside areas due to line-of-site distance or obstructions. Boosting or amplifying the weak RF signal enhances performance. This is typically accomplished by cascading a series of filters and amplifiers to achieve the rejection requirements required due to the power amplification of the repeater. However, this action increases ripple, causing signal distortion, dropped calls, and possible system damage.

Applications continue to unfold that use SiP technology to compress as many features into as small of an area as possible. For example, Philips Semiconductor has continued to introduce reference designs that use its advanced RF SiP technology, which further exploits its proprietary QUBiC4 BiCMOS process technology. This process is capable of producing highly integrated RF circuits in which critical passive components can be fabricated alongside active transistors on the same piece of silicon, the company said. Less critical passive components, or passive components that require significant silicon area, can be fabricated using Philips' Passive Integration Technologies. Where RF power handling and linearity are required it can bring in gallium arsenic (GaAs) devices.

In many cases, getting these active and passive component networks closer together results in RF performance improvement due to shorter interconnect lengths and fewer parasitics. However, these improvements can only be realized if the quality factor (Q) of the passive components when integrated onto silicon can be maintained. It is the ability to maintain the quality factors of integrated passive components that differentiates one manufacturer's passive component integration capabilities from another's.

RSS    Save to Del.icio.us  Digg This