Talk of silicon completely displacing GaAs in the RF front-end of cellular phones is premature and dead wrong, argues TriQuint’s Phil Warder. In his view, the future will instead involve smart RF suppliers collaborating closely with OEMs and chipset partners to offer complete RF solutions that exploit the best technologies for each application.

When it comes to cellular phones, GaAs technology is synonymous with RF design. That’s been the case for many years, because design engineers have been able to rely on GaAs for its combination of superior performance and small die size. This material is used to manufacture HBTs used for building power amplifiers (PAs) that lie at the heart of most RF front-end designs. These GaAs-based amplifiers deliver very high levels of efficiency, which hold the key to long battery life in small form factors. Today, these GaAs PAs are facing competition from silicon CMOS versions, which have already made some inroads in the 2G handset market. However, performance concerns have hampered the broad global adoption of this rival technology. So, as long as battery life and small size remain paramount in the eyes of handset designers, our view, at TriQuint Semiconductor, is that GaAs will retain its dominance in the mid-range and high-end 3G/4G market, thanks to its superior performance and significantly smaller size. Increasing complexity in the RF front end The RF front-end is defined as all the components between the digital baseband transceiver and the antenna. Its basic building blocks include PAs that boost RF signals, switches that direct the path of those signals and filters that block unwanted noise. Increasing uptake of 3G/4G data-enabled smartphones, which are replacing 2G voice-only phones, is creating new challenges for design engineers. Along with meeting more stringent performance requirements, designers must accommodate a rapidly increasing number of frequency bands and filters within each mobile device. The ever-more crowded RF spectrum has also fueled demand for high-performance filters for mitigating the resulting interference issues. In addition to DC-DC conversion that is used to reduce the current drain at lower output powers, design engineers are looking to enhance the front-end through the addition of complementary functions, such as envelope tracking modulators and antenna tuning. Many of these functions can be implemented in silicon with adequate performance. While there has been significant improvement in silicon performance, especially in switches and low-noise amplifiers, GaAs PA current drain still sets the industry benchmark. Silicon’s attractive credentials One of the biggest appeals of silicon technologies is their potential to unlock the door to higher levels of integration. This includes the promise of a single chip that integrates the RF front-end with the transceiver. The allure of an entity that incorporates the transceiver, PA, antenna switch and filters is hard to deny, but CMOS struggles to maintain efficiency at higher powers. This means that an-all silicon chip may only appeal to designers who are willing to sacrifice performance in favor of CMOS integration. Now that III-V suppliers are starting to face a potential threat from silicon, how will they respond? By continuing to do what they’ve always done: To deliver what their customers need. Handset designers are focused on optimizing the overall performance, size and cost of their front-ends across their broad product lines, and they don’t concern themselves with the specific technologies used. For III–V suppliers, the key to success is delivering a complete RF solution, which in many cases is an integrated module, not an individual die; utilizing the best technology for each application; leveraging both III-V and silicon technologies to continue pushing performance and cost frontiers; and recognizing that complete front-end solutions require advanced filtering technology.  GaAs power amplifiers (PAs) deliver superior efficiency, providing longer battery life for mobile devices — at about a third the size of comparable CMOS PA die

While there will be variations within the vast global RF market, in many cases the best way forward is to combine the merits of GaAs and silicon. A good example of this strategy is our multi-mode, multi-band power amplifier module (MMPA): This combines high-performance GaAs PAs with a CMOS controller and silicon-on-insulator (SOI) switches. MMPAs provide a highly integrated approach for today’s increasingly complex RF design, and they equip designers with more room on the circuit board while minimizing engineering time and resources. MMPAs can support more frequency bands than discrete architectures, while trimming board space by 20 percent. What’s more, these multi-band amplifiers feature a versatile design, allowing manufacturers to adopt a common platform for releasing new products at a faster pace, while keeping a lid on design and manufacturing costs. More than 90 percent of new smartphones and cellular phones use GaAs power amplifiers (PAs) to deliver longer battery life, although CMOS PAs are beginning to make inroads in entry-level applications   Delving deeper To select the best parts in these modules, one must evaluate the relative merits of GaAs and silicon technologies on a component-by-component basis. When it comes to PAs, GaAs continues to outperform silicon designs significantly in terms of current drain and die size. Due to this, GaAs will continue to be widely used for mid-range and high-performance applications, with silicon PAs targeting lower-end sockets where performance is not as important. Nonetheless, silicon will still have a home within even high-performance, GaAs-based MMPA modules. For example, silicon controllers and distribution switches can enhance MMPAs. Additionally, silicon is used for DC-DC converters and envelope trackers that further optimize battery current drain to improve the overall performance of both GaAs- and silicon-based RF architectures. For years, designers leveraged another GaAs-based technology for its efficiency advantages in RF switches: pHEMTs. Now that steady progress in SOI switches provides comparable performance, this alternative is more widely used in mobile device designs. GaAs pHEMT switches will be reserved for applications where a superior cost or size tradeoff can be achieved by integrating the switch with a GaAs PA die, rather than using two separate die. The RF front end for mobile devices such as smartphones is becoming increasingly complex as a growing number of bands are added to support 2G/3G/4G voice and data services, as well as global roaming. This is driving demand for superior efficiency, as well as high-performance filters  Providing control and biasing circuits within amplifier modules is one area where silicon has been used for many years. In addition, power detectors, temperature sensors and regulators have a long silicon history. These silicon circuits often comprise one die within a multiple-die module. Recently, module control has been transitioning from a few dedicated functional digital pins to a control bus architecture. This change is driven by the increasing number of bands and functions in front-end modules, as well as the desire to minimize the required control pins out of the transceiver or baseband. Silicon will remain the preferred choice for control buses as the MIPI front end interface becomes more widely adopted.  TriQuint’s multi-mode, multi-band power amplifier module (MMPA) mixes GaAs and silicon technologies to achieve best-in-class performance One of the trends within cellular technology has been a steady increase in the number of frequency bands. This has made it more challenging to achieve good radiated performance in compact form factors, due to the expanded bandwidth. Making matters even worse, there is a desire for multiple antennas in MIMO (multiple-input and multiple-output) applications, and this is pushing space constraints. To address all of this, designers are exploring tuning technologies to optimize antenna performance. There are several competing variations in the RF space, including some silicon-based components; the market has yet to throw its weight behind one particular technology. Filtering out the noise Filters play a crucial role in the RF front end, because they selectively pass certain frequencies while rejecting unwanted noise. Unlike PAs, which can cover multiple bands, filters are band specific, so growth in phone band counts leads directly to growth in the number of filters or duplexers within each device. The deployment of 4G LTE networks is driving band counts and increasing demand for high-performance filter technologies like BAW and TC-SAW Many of the new bands allocated for LTE present tough, technical problems associated with filter design. Amid a global spectrum crunch, new 4G bands are being squeezed next to pre-existing bands, often with minimal guard bands. To mitigate the resulting interference issues, it is essential to employ advanced filter technology. Traditional surface acoustic wave (SAW) technologies have been adequate in the past, but the most challenging 3G/4G frequency bands need advanced filter technologies, such as bulk acoustic wave (BAW) or temperature-compensated SAW – we offer all three. In addition, service providers want to increase network capacity through the introduction of aggregation techniques, and high-performance filters can make this possible. Due to the rapid deployment of LTE, shipments of filters for the RF front-end are forecast to outpace the growth in PA content. These filters can be discrete components, or they can be integrated as filter banks or filter banks with switches. They also have the potential to be combined in components with higher levels of front-end integration. This represents an expanding market opportunity for III-V suppliers with advanced filter technologies – they are the key to delivering a complete RF solution.