The Economics of Handset RF Front-end Integration

The mobile device has come a long way from its humble beginnings as a cell phone. Today, it is becoming the primary means of communication for most people around the world; serving as both a voice and data connection, emulating many of the functions once reserved for the personal computer and ushering in a new class of personal media devices. In addition, mobile technology is playing a key role in new embedded wireless and machine-to-machine (M2M) communications. These are the trends impacting the mobile device industry, fueling its growth and ultimately defining the needs of the radio inside, which will be as diversified as the trends driving these applications.

For example, an RF front-end solution covering a specific geographical region with a one or two band low-cost entry phone is considerably different than the front-end in a handset targeting global coverage with a high-bandwidth data/voice connection. The former calls for a high volume front-end with RF content at the $1 and below price point, while the later requires a multi-band, multi-mode radio with RF content that can run from $8 to $10.

Figure 1 Projection of mobile phone unit sales by standard.

Market Dynamics: 2G Entry, 3G Emerging and 3G/4G Smartphone Sectors
Worldwide mobile phone sales totaled 315 million units in the first quarter of 2010, a 17 percent increase from the same period in 2009, according to Gartner Inc. Industry analysts are forecasting handset unit volumes to grow approximately 10 percent this year to around 1.5 billion units (see Figure 1). Smartphone sales, which reached 54.3 million units, increased 48.7 percent over the first quarter of the previous year.¹ The growth rate of Smartphones, which contain 3 to 6 times more RF dollar content than voice-only phones, is forecast to continue growing at 42 percent through 2010, and 35 percent in 2011, vastly outpacing middle-tier "feature" phones, which had relatively flat growth and should eventually be displaced by feature-limited 3G entry phones.

While Smartphone market growth has been phenomenal this year, 2G entry phones represent another sizable market opportunity. Ralph Quinsey, President and CEO of TriQuint Semiconductor, expects this market, which is mostly in China but increasingly includes India, to account for "500 to 600 million low-end, voice-only (2G—probably dual-band GSM) phones this year." Quinsey referred to this sector [2G and eventually 3G entry phone] as the second wave of market opportunity (behind the first wave of revenue growth represented by Smartphones).

With voice-only entry phones exceeding new landline connections, Greg Waters, Executive Vice-President and General Manager of Front-end Solutions for Skyworks, said that "the company expects voice-only phones to continue to grow as they replace landline phones, with the cost of voice-only services continuing to drop. A combination of CMOS and GaAs technologies are required to continue to drive aggressive price points in this type of solution. Skyworks believes they have over 50 percent market share today in 2G and that their share will increase."

Eric Creviston, President of the Cellular Products Group at RFMD, confirmed that "2G has just done extremely well in the emerging market category. For this sector, the company developed a transmit module (PA and switch) focused on China and Korea handset manufacturers." The strength of this market was clearly evident when the company surpassed 100 million units in cumulative shipments of its 2G dual-band and quad-band transmit modules this past spring, less than one year into production shipments, putting the 71xx transmit module product family among the company's all-time most successful product launches. Creviston was not in agreement over which company enjoys the largest market share for 2G, believing that RFMD, at 40 percent, holds the largest market share.

By 2015, it is predicted that five billion people will be connected via communication networks, with well over half enjoying broadband access. The majority of new users will come from emerging markets and 3G/HSPA will be the core technologies to connect them to the Internet.²

In general, the 2.5G or "feature" phone category has not really been showing a lot of growth year over year. There is a strong likelihood that the 2.5G category will transition into a new, low-end 3G Smartphone category, one that is different than the current high-end phones with less RF content but a minimum of 2X content over 2G entry (voice-only) phones, (see unit sales by radio standard in Figure 1). Quinsey believes that access to the Internet is the "killer-application", a term repeated by Waters of Skyworks as he described the five year, 131 percent CAGR of mobile data traffic, driven largely by audio, video, P2P and data (see Figure 2).

Creviston also sees the breakout 3G emerging markets as a very fast growth category: "What's behind all this is the need for higher and higher data throughput for mobile applications and economics are driving these finer segmentations. In other words, how cheaply can one provide as much data as possible to as many people as possible. While the 3G/4G or Smartphone category is growing very nicely, to around 250 million handsets, we believe the fastest growth will be the 3G entry category, which to a certain extent will replace the 2.5G, or EDGE feature phone."

Creviston went on to comment that, "the 3G emerging markets will be built around 1 to 2 bands of UMTS and then in some cases just GPRS for the 2G backward compatibility and in other cases EDGE. This is in contrast to the 3G/4G/Smartphone market that typically has 5 bands of UMTS and 4 bands of EDGE. So what vendors are trying to do (for the emerging market) is create a high volume solution for a very targeted region, such as a part of India, for example."

Creviston shifted his attention to the connected device category (M2M and personal media). "This is not really a new segment but one that is growing very rapidly and becoming quite meaningful to all handset front-end suppliers. This segment includes usb data modems, which from an RF perspective are similar to a Smartphone or high-end multi-mode phone with the exception that they typically work at max power level." In contrast, a Smartphone communicating data at the edge of a cell site might be at max power, but they also have a lot of use in voice communication where the power is backed-off to preserve battery life, which leads to a wide spread of power ranges. This results in different design criteria between M2M and handset front-ends.

Will 2G and lower data rate technologies be used for M2M? Not necessarily. "Currently, the majority of M2M applications are at low data rates, and are using CDMA and GSM for asset tracking or remote vending applications which require little data.But more and more applications are using high amounts of data such as video surveillance. So M2M, as a category, will also include 2G, 3G and 4G as well," remarked Creviston.

Quinsey referred to M2M as the third-wave of general mobile device trends "that will provide lift for the RF suppliers. M2M applications such as e-readers, remote health sensors, power grid management and security surveillance will need RF connectivity. Over the next decade, I see billions of wireless nodes, outstripping the demand for person-to-person communications in the long run."

Waters agrees that embedded RF may overtake cellular in pure volume in this decade. "Evolving mobile applications and network traffic that includes broadband data such as video is already calling for a shift in product architecture and will require new innovation and the best of CMOS and GaAs to improve battery life for high data rates. At the heart of the challenge is to find a way to support multiple bands and air interfaces that offer a small enough form factor while improving performance, battery life and heat dissipation. Skyworks is already beginning volume shipments of new products tailored for high speed data and video applications."

The projected contribution of each of these sectors on the cellular terminal market is shown in Figure 3, indicating the growth expected from Smartphones and embedded wireless devices to the overall number of units sold.

Standards and market drivers of Front-End technology
The high-tier Smartphone market needs to provide simultaneous voice and data transmission with broadband functionality for multimedia and international roaming (which allows phone manufacturers to realize economy of scale) while maintaining support of existing networks. To do so, these UMTS (W-CDMA) terminals may support up to 15 frequency band allocations, backward compatibility with the earlier GSM standards, and support for additional non-cellular wireless interfaces such as WiFi, Bluetooth and GPS. All this requires additional RF content (amplifier, filtering and switching), which adds to the complexity of the RF front-end architecture.

Over the past 20 years, the handset radio has evolved from a nearly all-discrete solution toward highly integrated baseband and transceiver ICs and integrated RF front-end modules (FEM). RF functional blocks between the transceiver and antenna include filtering, amplification and switching (with impedance matching incorporated between components where needed). The electrical requirements for each component as well as the overall topology are dictated by the particular standards requirement and band support (see Figure 4).

The increase in RF components as band support moves from single to dual, triple and quad bands in a time-division duplex GSM phone is shown in Figure 5. Progressive integration might lead to a dedicated filter bank (e) or incorporation of the switch and filter into a single Antenna Switch Module (ASM) module (f). Individual band specific power amplifiers could also be combined into a power amp module (PAM).

Where time-division duplexed systems such as GSM rely on switching between receive and transmit functions, CDMA and W-CDMA use full-duplex communication with receive and transmit functions operating simultaneously at slightly different frequencies. The addition of a duplexer in the transmit/receive chain is one of the principle FE architecture differences between these standards. Traditionally, handset designers seeking to support multiple air interface standards in the same device have resorted to high-throw switching and stacked radio architectures, which can lead to front-end duplication of components. An FE diagram for a phone utilizing discrete components and supporting two GSM bands and one UMTS band is represented in Figure 6. The BOM costs for W-CDMA handsets are typically double that for EDGE handsets and nearly triple those for GSM/GPRS devices.⁴

Trends in FEM Partitioning
How front-end components are partitioned into modules is driven largely by economics and suppliers' access to technology. Handset OEMs need to lower the bill of materials and limit the amount of engineering they invest in RF design at the board level. Driving up the cost (and complexity) is the need to increase revenue potential by way of multi-band, multi-mode platforms with broader market appeal. Scale drives integration. In other words, the less specialized the phone's band requirements, the more cost effective it is to integrate multi-mode RF components into a module. Partitioning functional blocks based on standards requirements and reducing costs has led to a number of module products.

Common FE modules (see Table 1) include:

• Antenna Switch Module (ASM): Antenna switch and low pass filters. The ASM appears mostly in low-end GSM handsets. Renesas is the only vendor using pin diode switches and LTCC; the other vendors use hybrid modules based on PHEMT MMIC or SOI switches and SAW or BAW filters.
• Switch Filter Module (SFM): For GSM phones, the SFM integrates the Rx bandpass filter and the ASM. Demand for this module should exceed that of the Tx module when combined with a duplexer module for the UMTS/GSM multi-mode market. It will likely become obsolete as duplexers are integrated directly into the Switch Duplex Module. Samsung, LG, Apple and Huawei are major customers of the SFM.
• Tx Module: ASM and PA module. This module is largely dominated by PA manufacturers who tend to have a cost advantage. This module is mostly adopted in GSM terminals, although Nokia uses this module in its UMTS/GSM handsets as well. The Tx module will likely lose out to the Switch Duplexer and multi-duplexer modules for UMTS/GSM and LTE/UMTS/GSM platforms.
• PA-Duplexer Module: Mostly found in the iPhone 3G/3GS, Samsung Wave/Galaxy, Motorola's Droid, RIM's Bold, HTC's EVO and Huawei's WWAN Module/dongle. TriQuint believes that with the performance advantages and the flexibility that PA- Duplexer modules offer, this integrated solution will solve the large number of band combinations their customers have to provide while minimizing the use of board space. Yet, PA-Duplexer module shipments are expected to decrease as multi-mode PAs begin to ramp in the first half of 2011.
• Switch Duplexer Module (Antenna Switch and Duplexers): Nokia plans to expand adoption of this architecture for its multi-band handsets, which will likely entice other manufacturers.
• SFM and Duplexer Module (Switch filter module and Single Duplexer): This module is the level of integration expected for the future.

Table 2 lists the global suppliers of these various module types.

Figure 7 diagrams various ways in which integration has been adopted. For a single band UMTS/GSM phone, one common architecture will be based on partitioning around a Switch filter module, PA-Duplexer module (shown with single band UMTS) and dual band GSM PAs (see Figure 8a). On other occasions, the duplexer will be incorporated into the Switch Duplexer Module (see Figure 8b).

The multi-duplexer module is an attractive front-end configuration for multi-band UMTS/GSM terminals due to the adoption of multi-band, multi-mode PAs among the top tier handset OEMs. Figure 9a shows the high and low band PAs (two pairs supporting UMTS and GSM) as separate modules from the narrow band filter/duplexer switch module, which toggles between bands (UMTS) and modes (UMTS/GSM). The tri-band UMTS configuration shown could be expanded through module development to support more bands (via additional duplexers and higher throw switching). However, given the addional distribution switch required to route the duplexers, the PAE will degrade, requiring additional current. If required, 4G phone support is achieved with an additional, dedicated LTE PA-Duplexer Module and separate antenna.

An alternative configuration is based on PAs that are integrated with band specific duplexers for UMTS and a transmit module/filter bank combination (see Figure 9b) or a switch filter module (see Figure 9c) for GSM support. Using a GSM Tx Module with Rx filter bank and individual PA-Duplexer modules is an approach found in some Nokia phones and Samsung platforms (note: the Rx SAW filter and LNA are integrated into the UMTS/GSM transceiver), while the GSM Switch filter module, individual PA-Duplexer module and Rx filter bank is found in the iPhone 3G/3GS and certain phones from Samsung, LG and Huawei. As band support increases, stacking PA-duplexer modules becomes increasingly inefficient.

At issue will be the need to improve individual component performance within the multi-mode module solution while making the changes to the fundamental multi-band, multi-mode architecture so that it (the combination of PAs, filters, switches and signal paths) is more efficient and reduces redundancy. This is critical in order to cut size and cost, but can challenge performance if not approached systematically.

The PA Module
Leading GaAs PA manufacturers include RFMD, Skyworks, TriQuint, Avago, Anadigics and Renesas. Analysts report that Skyworks has a slight market share lead over RFMD with each company supplying more than a third of the world's handset PAs, followed by TriQuint supplying around 15 percent. Handset OEMs often use multiple suppliers. For example, Samsung uses W-CDMA PAs from TriQuint, RFMD and AVAGO, GSM/GPRS/EDGE PAs from RFMD, Skyworks and TriQuint, and AVAGO and Anadigics in its CDMA products. Likewise, LG uses PAs from Skyworks and TriQuint in its W-CDMA phones and Skyworks and RFMD in its GSM phones.

"Our (Skyworks) handset customers place a high value on architectures that help them to reduce size and help them resolve noise problems in many applications. In many cases, they are looking to reduce the bill of material complexity and otherwise, enable their competitive differentiation. We have been capitalizing on the increase in handset front-end module content, leveraging our in-house building blocks to produce cost-effective solutions that help decrease the handset manufacturers overall bill of material costs. Because of the complexity, the competitive landscape is narrowing with fewer and fewer suppliers able to support both the technical requirements and the manufacturing scale required by the leading handset OEMs." At this point, Skyworks reports that the company is supporting all five top tier handset OEMs as well as all key Smartphone suppliers and was first (early 2009) to introduce a family of power amplifiers and front-end modules (FEM) for 4G, long-term evolution (LTE) applications, supporting 13 bands.

"With front-ends supporting an increasing number of bands and modes, there is a trend toward using converged multi-mode, multi-band PA modules, rather than having redundant PAs for each band specific signal path (due to the BPF) in the Tx module. Companies with multi-mode, multi-band PA module portfolios and meaningful market positions in GPRS, EDGE and W-CDMA should do well as this market trends plays out," stated RFMD.

Creviston explained the dynamics impacting the entire industry. "There exists a highly integrated 2G backbone with quad-band EDGE plus the TR switch in one box that has been driven up the maturity curve pretty high because of all the volume on the 2G side. But then with 3G, handset manufacturers stacked discrete power amplifiers and duplexers for each band, which made perfect sense when there was only one band of UMTS and was still manageable as dual-band. But inside today's Smartphones, with support for 4 or 5 UMTS bands stacked up next to 2G, the solution is no longer a bolt-on, it's becoming a majority of the content. It calls for a different way of thinking about it and so leaders in this industry began to look at breaking that switch out because it's band-specific, routing that signal through a specific duplexer and filter network. What's not band-specific are the power amplifiers because they can be made broadband, and the power management associated with the front-end is also band independent. This is where the need for the multi-band, multi-mode power amplifier comes from. For example, handset OEMs require effective power management to optimize handset energy consumption while maintaining signal quality. Traditionally, PAs have been two-state GaAs devices, switching the current between high and low power levels. RFMD has implemented patent-pending DC-DC conversion technology to dynamically control PA operating conditions, maximizing efficiency across power levels, across data rates and during non-ideal load conditions, quickly responding to load and line transients."

RFMD is addressing the multi-mode, multi-band trend with its Power -Smart™ power platform. This reflects the company's drive to integrate all RF functionality into two module placements for multi-mode, multi-band 3G/4G segments and a single module placement for all entry level phones. The two module placements are partitioned along the RF configurable power core, which includes PA, PA power management and RF mode switching and Antenna Switch, Switch Filter and Switch Duplexer Modules, which integrate all switching and filtering into a single placement.

Skyworks has also developed multimode PA modules for next-generation Smartphones with the release of its SKY776XX product family. Waters discussed how these new multi-mode and multi-band PA modules meet the need for increased frequency bands while reducing board space by integrating the functionality of multiple discrete PAs into a single package. These devices have been designed to operate efficiently in quad band GPRS and EDGE, and support bands 1, 2, 5 and 8 for W-CDMA and HSUPA modulation. These PA modules have been designed for improved performance under mismatch conditions and have reduced current consumption over the entire power range of the handset. The company's custom BiCMOS controller and interface IC provides the integrated (accurate and fast closed loop circuitry) power amplifier control function, called iPAC™, using a low power control slope. Other innovative technology cited by Waters includes their developments in merged HBT-FET (BiFET) devices, copper backside process, and stacked die and flipped chip technologies.

Likewise TriQuint is developing a scalable 3G/4G converged RF architecture for multi-mode, multi-band mobile devices by converging functionality into one PA module, a move that should offer up to a 50 percent size reduction over today's discrete approach. The TriQuint Unified Mobile Front-end (TRIUMF) is architected to support numerous frequency bands and air interfaces used in 3G mobile devices supporting modes like GSM/GPRS/EDGE for voice and lower-data-rate applications and W-CDMA/HSPA/LTE for high-speed data. On the multi-band side, it will handle traditional quad bands (GSM850/900/DCS1800/PCS1900) unified with options for 3GPP-designated bands 1 through 17. In doing so, it will enable worldwide W-CDMA/HSPA/LTE coverage. In addition, TriQuint's power amplifier modules utilize the combination of a BiHEMT process for higher integration, and performance and proven flip-chip technology that enhances their devices' thermal characteristics.

TriQuint reports that they have been able to simplify the RF front-end while reducing size through a multimode module solution using a converged PAM for the traditional bands and PA-Duplexers Modules for LTE. Quinsey commented that their proprietary design techniques have an advantage in that they do not compromise current consumption when compared to today's discrete architectures.

AVAGO improves power efficiency in the low output power range with an active bypass PA technology called CoolPAM, while ANADIGICS' High-Efficiency-at-Low-Power (HELP) technology uses an integrated pHEMT switch based on its BiFET InGaP processes to allow different amplifier chains to be chosen in the PA, depending on the output power required. AVAGO has plans to develop multimode, multi-band PA structures and is working toward that approach.

The Switch
In multi-band, multi-mode handsets where different RF paths are required for the appropriate filtering and amplification of the signal, the RF switch is required to provide access to the shared antenna. The RF switch must be capable of switching up to 15 paths of high-power RF signals between different duplexers (and GSM filter-paths). Performance-wise, the switch must have low insertion loss (so as not to degrade the effective PAE of the PA), high isolation (to prevent leakage across channels) and exceptional linearity (it is generally agreed that the switch needs an IP3 of better than +65 dBm).

While a PIN diode switch has very low insertion loss and harmonic distortion, a multi-throw PIN diode switch requires quarter-wave transmission lines that make them unacceptably large for today's applications. GaAs PHEMT switches dominate today's market share, but have low ESD tolerance and need additional decoding and DC-blocking circuitry.

TriQuint leverages its unique E/D pHEMT process to address the demanding performance requirements of 3G switches where insertion loss, linearity and harmonics must be optimized.

RF CMOS is making inroads into front-end modules, either as bulk CMOS and silicon-on-insulator (SOI) or silicon-on-sapphire (SOS). RFMD's RF CMOS switches use high-resistivity silicon substrates from a leading foundry to meet or exceed the stringent linearity and isolation requirements for 3G and 4G Smartphones, while providing excellent ESD performance (HBM data rated at 2000 V). By integrating the controller and RF switch, the company has been able to improve performance while reducing size and cost.

Skyworks also employs multiple generations of HEMT material as well as 0.25 μm SOS and 01.3/0.18 mm RF CMOS on SOI for its switches. The company offers a family of GaAs and SOI antenna switch modules for 2/3/4G handsets, including a single pole ten throw CMOS switch that supports five 3/4G transmit-receive ports, three receive ports and two GSM transmit paths with LPF.

Filter/Duplexer Technology
RFMD and Skyworks both obtain filter/duplexer technology from multiple leading sources (Murato and TDK Epcos, Panasonic, Taiyo Yuden and AVAGO), allowing them to optimize performance for a given application. RFMD does not participate in the discrete market since duplexers are mostly becoming integrated into switch duplexer and PA-duplexer modules. Both companies state that this approach (buying filter/Duplexers) allows them to select the best filter technology available and integrate it into their products.

Taking a different position, TriQuint believes it can best achieve the same objective with its own in-house development of filter technology. Quinsey sees TriQuint's filter technology as a significant differentiator for his company. "As you transition from 2G to 3G, from voice to data, duplexers become very important because it becomes a frequency domain duplex market as opposed to time domain based on switches. The market for 3G and 4G duplexers is going to be billions of dollars in the future. Other front-end suppliers buy duplexers and integrate them into modules, whereas we custom design our building block for each integrated module using in-house technology such as SAW, TC-SAW (temperature compensated) and BAW duplexers, switching and power technology. This enables us to create highly integrated modules and long-term roadmaps, focused on integration and performance, size and cost."

Receive filters are typically based on surface acoustic wave (SAW) technology and implemented in single-end to balanced topology for better receiver sensitivity and dynamic range. They are designed with three basic topologies: the ladder filter for low pass-band insertion loss, the coupled resonator filter (CRF) for high out-of-band rejection, and hybrid forms combining ladder and CRF elements.

SAW technology dominates the cellular-band duplexer market, while bulk acoustic wave (BAW) duplexers based on film bulk acoustic resonator (FBAR) and solidly mounted resonator (SMR) technologies dominate in PCS band duplexers as well as the PCS inter-stage transmitter BPFs. A BAW resonator has a higher quality factor (Q) than a SAW resonator near and above 2 GHz, which results in lower insertion loss and a steeper filter skirt. BAW resonators also exhibit better thermal stability, superior ESD robustness and better power handling capability compared to a typical SAW resonator; however, at lower frequencies, SAW filter performance is more than sufficient for handset applications, where its maturity and commodity pricing give it the advantage (see Figure 10).

RF SAW filter packages and assembly processes have evolved through three major generations in recent years: chip & wire, flip-chip and chip-scale package (CSP). The CSP generation is considerably more complicated, which has led to diverse manufacturing solutions. The only common threads among major SAW suppliers remain footprints, pin-outs and the hermeticity requirement. Currently, all major SAW suppliers are developing "packageless SAW" through wafer-level packaging (WLP), including Murata, Avago, TriQuint, etc. Multi-band and multi-mode support will require larger quantities of both to be integrated into filter or duplexer banks if not directly into front-end modules using flip-chip technology and WLP.

CONCLUSION
Driven by global demand for 2G/3G entry phones, a strong smart phone market and M2M communications, the economic outlook for suppliers of RF FEMs for mobile devices is very promising. Supporting multiple bands and standards have allowed suppliers to increase the RF dollar content of their solutions provided they have the R&D scale to meet the engineering challenges. Each major player is developing technologies ranging from flip-chip, copper metallization, devices (HBT, PHEMT, RF CMOS, BiFET, BiHEMT, etc.), packaging and architecture that will help them achieve better performance at reduced size and cost. The good news for RF engineers is that all three executives recognized the absolute necessity of having talented technologists on staff and were very vocal about their efforts to recruit and retain them. For more comments from RFMD, Skyworks and TriQuint on how these companies are approaching the engineering challenges for tomorrow's mobile device front-ends, go to www.mwjournal.com/mobile_FEM_2010.

References

1. http://www.gartner.com/it/page.jsp?id=1372013.
2. http://w3.nokiasiemensnetworks.com/na/Insight/3G+for+emerging+markets/.
3. Y. Andoh, Navian Inc., "Market & Technology of RF Modules: Focusing on Front-end Modules for Cellular Terminals," CS MANTECH Conference, May 17-20, 2010, Portland, OR.
4. D. Pilgrim, "Simplifying RF Front-end Design in Multi-band Handsets," RF Design, February 2008.
5. K.Y. Hashimoto, RF Bulk Acoustic Wave Filters for Communications, Artech House, Norwood, MA, 2009.