Optimum Technology Matching® – More Relevant Than Ever
By Ben Thomas, RFMD®

There continues to be much fanfare surrounding the use of various Silicon (Si) Complementary Metal–Oxide–Semiconductor (CMOS) technologies for implementing RF functionality. There are numerous published white papers and magazine articles touting a path toward the ultimate goal of providing a monolithic, Si CMOS-based communications solution from digital baseband to RF output. Certainly over the last decade, we have seen a dramatic improvement in the use of Si CMOS for RF functional integration, replacing everything from discrete circuits once found on printed circuit boards (PCBs) to complex Silicon Germanium (SiGe) compound semiconductor-based cellular transceivers. In fact, this monolithic goal has been achieved with Si CMOS systems-on-chip (SoCs) in WLAN and Bluetooth® and, in recent years, we have seen the emergence of SoCs which combine basebands and transceivers in the cellular communications arena.

Despite this rapid advancement in Si integration, there remains a technology battleground in the area between the cellular transceiver output and the antenna input, better known as the cellular front end. While the aim of providing a single placement front end is seen as feasible in the near term, a single placement, monolithic cellular radio is not. The reasons for this are straightforward and few, based on the fundamental tenet that cellular radios are embedded in consumer devices which require both low cost and long battery life. A single semiconductor technology path to simultaneously achieve these two goals in an environment of increasingly complex cellular communication standards has not been presented. However, a highly integrated, multi-technology module has, and the ability to architect and optimize technology selection within these modules will be the key to success in this new decade.

Optimum Technology Matching® Revisited
Almost a decade ago, RF Micro Devices® (RFMD®) issued an advertising campaign describing their philosophy surrounding the choice of semiconductor technologies for various uses in radio frequency (RF) and wireless products. It was termed Optimum Technology Matching® (OTM). OTM illustrated that rather than being focused on a single semiconductor technology to support all RF functions and applications, it would be best to consider that each of the available semiconductor technologies offers unique performance and cost trade-offs. Careful consideration of these factors in combination with the attributes of target markets should determine which semiconductor technology to utilize in order to meet stringent customer demands. At the time, this was a particularly important message to RFMD’s customers, as RFMD was quickly ramping internal production of a high-volume Gallium Arsenide (GaAs) Heterojunction Bipolar Transistor (HBT) wafer fabrication facility and many customers believed this would be their default semiconductor technology choice.

Since the inception of the OTM strategy, the wireless semiconductor industry has changed dramatically and no sector has more so than cellular RF. In today’s age of highly integrated module solutions, RFMD’s development lens is focused on functional technology choice within a module rather than the historical focus on technology choice for each discrete component to be placed on an original equipment manufacturer’s (OEMs) PCB. Nonetheless, the task of mating the proper technology to the function is no less important. In fact, it would seem technology optimization is even more critical since a single product supplier must now balance these choices while assembling an increasingly complex product and yielding the high volumes needed to meet consumer end market demands.

Why Not a Single Semiconductor Technology Like Si CMOS?
Innovations in mixed signal Si CMOS technology have certainly had a dramatic impact in the cellular RF industry. However, the ability to realize Si SoCs that combine cellular baseband and transceiver functions did not come without its trials and tribulations. Many would agree that this combination was much delayed and not nearly as ubiquitous as once expected. During the race toward cellular SoCs there was also a great deal of hype generated around “the next logical step” of integrating the power amplifier (PA), primarily by new market entrants.

Achieving this “next logical step” turned out to be a giant leap forward, a leap that has yet to be attained. Despite the dramatic improvement in RF functionality in Si CMOS, efficiently delivering the power output required by high-volume cellular standards like GSM/GPRS proved difficult in standard Si CMOS process technologies. To satisfy these basic needs, many new market entrants have turned to customized Si process technologies and/or unique module substrate materials to achieve reasonable performance. Some such suppliers are shipping in low-volume production as stand-alone GSM/GPRS power amplifiers today, serving markets that value cost over performance.

However, with this move away from the bulk CMOS technologies used for cellular SoCs, the end goal has shifted from the possibility of full cellular RF integration to one of competing head-to-head with existing cellular PA suppliers based on a promise of a low cost structure. Another aggravating factor is that SoCs are not standing still; instead, they are continuing to drop to the next lowest technology node, 65 nanometer (nm), with 45 nm and 32 nm in their sights. SoCs must remain on this path in order to meet the high baseband processing demands of complex 3G and 4G communication standards. As a result, these 3G and 4G standards are evolving at a rate that outpaces the wireless semiconductor technology’s ability to deliver monolithic integration. This is driven primarily from both the increasingly complex cellular standards for data transmission and the increasingly stringent requirements of the frequency spectrum allocated to the task. As such, opportunities for monolithic integrated circuits (ICs)—particularly in the area of cellular RF front ends—are very limited and perhaps unattainable. To simplify, these small Si geometry nodes are not conducive to achieving the levels of high-power RF performance required by cellular front ends, thus pushing the idea of complete cellular radios on a single Si CMOS technology even farther from reality.

Nevertheless, OEM demand for increasing integration has not slowed and Si CMOS is present in most, if not all, cellular PA and front end modules. However, the Si processes used are the ones more optimized for low-cost control functions and not those utilized for cellular SoCs. In fact, Si CMOS’s presence is growing in cellular front ends, particularly as the world of cellular RF embraces digital control standards like the Mobile Industry Processor Interface (MIPI) alliance, which includes leading companies in the mobile industry that share the objective of defining and promoting open specifications for interfaces inside mobile terminals. Additionally, there is a trend of cellular SoC providers demanding FE products become more self-regulating in response to their operating environments as well as providing more feedback to the SoC for high-level calibration and control schemes.

All of this is leading to an increasing dependence on Si CMOS within cellular front end modules, but in a supporting role rather than the primary semiconductors’ roles of amplification, filtering, and switching. In RFMD’s case, our focus on compound semiconductor-based power amplification, most notably GaAs and Indium Gallium Phosphide (InGaP), has enabled continued advancements in amplifier performance while lowering the costs of both design and manufacturing. Since RFMD has an internal source of GaAs and InGaP die, the largest fabrication facility of its kind, these become cost-effective technology choices. Given the clear performance benefits of compound semiconductor technologies in RF applications, the optimum blend for RFMD, and indeed the cellular front end industry, has become a split between digital and low frequency analog circuits in Si CMOS and RF circuits in compound semiconductors.

Is OTM Feasible?
So, will we see a single placement cellular front end in the near future? Absolutely. Mobile Internet Device (MID) OEMs are demanding minimal placements as the cellular RF section of handsets are driven to consume minimal PCB real estate while also contributing minimal impact to final product yields. Additionally, an OEM’s device development model demands minimal resource investment in designing and implementing the RF section of MIDs. Rather, OEMs have turned to spending their R&D dollars on what the consumer sees—an improved user experience through extensive applications, an intuitive interface, and broadened functional capability.

Will this single placement cellular front end be a monolithic IC? It’s highly unlikely. Let’s take a multimode (MM), multi-band (MB) MID for example. The cellular RF section of this device must support GSM, GPRS, and EDGE over four frequency bands as well as multiple WCDMA data rates over five frequency bands. To answer this need, RFMD’s latest 2G/3G/4G MM, MB PA module integrates GaAs HBT, Si CMOS, and GaAs pHEMT technologies on an advanced, coreless module substrate technology (see Figure 1).

Taking module-based integration to the next logical level, the single placement cellular front end would need to operate as a digitally-controlled “black box” of RF functionality, accomplishing what today requires a handful of already highly integrated RF components, a relatively large PCB area, and a moderate number of surface mount devices (SMDs). The result is a single placement front end that contains all functions needed from the output of a cellular transceiver to the antenna input. Beyond the MM, MB PA example above, this module would require high-order throw count antenna switching, perhaps through silicon on insulator (SOI), as well as filtering (7-9 filters: 2-4 GSM receive filters and 5 WCDMA duplexers) utilizing either SAW or BAW filters that require the likes of lithium tantalite (LiTaO3) and other highly specialized material processes. This cellular front end module will clearly not be a monolithic IC, but rather a highly integrated module which assembles no less than four separate semiconductor technologies, each optimally chosen for their specific attributes and designed to work as a seamless system (see Figure 2).

In looking at such examples it becomes clear that there is an increasing need to consider applying a strategy such as OTM to the task of developing these modularized, system-level, front end solutions. In such a product, the ability to choose technologies according to their performance characteristics and the ability to architect a system which implements varied functions across multiple technologies are both critical to delivering innovation. Each technology carries unique and desirable characteristics that aid the module supplier in meeting the stringent cellular performance metrics required while also delivering a product that OEMs will value.

In weighing the diversity of RF functionality and the performance requirements of the cellular industry against the risks of attempting monolithic integration, it quickly becomes evident that we have a significant time period ahead in which assembling multiple semiconductor technologies in a single module will be the most effective and efficient route of cellular front end product development. This reality drives the need for a product development strategy such as OTM along with the technology portfolio breadth and systems architecture expertise to enact it.

Rest assured the coming decade will bring rapid advancements in semiconductor technologies, likely faster than in the previous decade. But it will only be those suppliers who can balance the various semiconductor technologies at the module level and who possess the systems expertise to integrate them successfully who will, in turn, lead the evolution of cellular RF.

About the Author
Ben Thomas is the director of marketing for 3G/4G Cellular Front Ends at RFMD.

RFMD
www.rfmd.com
Email this article to a friend!