RF Front-End Modules in Cellular Handsets
`
`Wang-Chang Albert Gu
`TriQuint Semiconductor - Sawtek Division
`1813 S Highway 44] , Apopka, FL 32703, USA.
`Email: agu@sawtek.com
`
`Abstract
`Radio portion afthe cellular handsets have evolvedfrom a
`nearly all-discrete implementation towards a very high
`level of integration in baseband lC, transceiver [Cs and
`front—end (FE) modules. Integration allows handsets de-
`signers Io pack more functionalities into a smaller volume
`with reduced bills of material (BUM) and costs. This paper
`willfirsr review the FE modulesfor the 2nd generation (26)
`handsets followed by a discussion affnture trend of FE
`integration in the emerging multi—bana’, multi-mode phones
`for me 2. 56' and 36‘ systems.
`
`RF FRONT-END IN ze HANDSETS
`RF FE in a cellular handsets have been considerably sim-
`plified in recent years with the development of direct-
`conversion receiver,
`integration of frequency synthesizer
`into transceiver IC, FE switching modules, and power am-
`plifier modules (PAM). A GSM FE consists of antenna
`switch module (ASM),
`receive (Rx) filters, PAMs and
`transceiver [C as shown in Figure l. The widely adopted
`translational loop, or the offset phase-locked loop (OPLL)
`transmitter further eliminates the interstage transmit (Tx)
`filter [l].
`
`ocssm
`
`EN
`
`
`RF Transceiuct Transceiver
`
`
`
`Pcs saw
`
`
` Figure 3. Diagram oi a dual-band ASM tor GSM handsets
`
`
`F‘.FZIF Tn]nacuivor The most common concerns regarding pHEMT switch are
`
`ASM in a GSM system serves two important circuit functions:
`switching between Tx and Rx signals for time division multiple
`access (TDMA), and rejecting harmonics generated by the nonlin-
`car PA Yla 310‘” P355 filter (”Fl Antenna switch 15 predomi-
`nately implemented usmg PlNldtodes 1“ a Single-P916, dou-
`blerthrow (SPDT) 9°"fig_ura‘1°“ as shown in Fleur? 2’
`which provtdes low insertion [055 m the T" path, hlgh ‘50"
`latton between Tx and Rx Signals, and low cost. Drawbacks
`of PlN diode switch include its complexity, especially in
`multt-band configurations, and drawmg 0f “313”de high
`control current of about 10 mA.
`
`V“
`
`3
`
`T:
`|—o
`
`|--—oRx
`
`Ant
`
`N4
`
`Figure 2. Schematics ol 3 PIN diode antenna switch in
`SPDT configuration.
`
`Figure 3(a) illustrates the block diagram of a typical dual-
`band ASM, which consists of a diplexer followed by two
`SPDT switches and LPFs. This dual-band ASM is consid—
`
`four~throw pHEMT
`erably simplified if a single-pole,
`(pseudo-morphic high-electron-mobility transistor) switch
`is used as shown in Figure 3(b).
`
`Des
`
`. ncs
`
`‘ DOS
`
`083 Tit
`
`Gs" Rx
`
`- 65" R)
`
`- Dcs T:
`
`- 63" TI
`
`using (a) PIN diode and (b) pHEMT switches.
`
`its susceptibility to electrostatic discharge (ESD) damage,
`as well as the need f0r a relatively large number of control
`lines, which may not be available from the baseband IC.
`The deployment of external ESD protection circuit can
`easily circumvent the ESD robustness concern, while recent
`development of the enhancement and depletion modes
`pHEM’l‘ (BID-mode pl-IEMT) processes reduces the num-
`
`Figure 1- RF FE diagrams 0’ 93M handsets: (a) single
`band, (bl dual bands, (c) tnple bands, and (d) quadruple
`bands.
`
`0-1303e8615-7l04l320.m ©2004 IEEE.
`
`225
`
`1
`
`2W4 IEEE CSlC Digest
`
`APPLE 1011
`
`APPLE 1011
`
`1
`
`

`

`ber of control lines to the same level as PIN diode switches
`by integrating the control logics directly onto the switch IC
`[2].
`Further integration of GSM FE is currently under way,
`which include combining the low— and high-band PAMS
`into a dual-band PAM, combining multiple band Rx filters
`along with impedance matching circuits into a filter bank
`module, and combining ASM with Rx filters into a front-
`end module (FEM).
`RF FE of CDMA (IS-95) handsets is considerably more
`complex than GSM handsets due to the requirements for
`duplexer and linear transmitter. The transition from superw
`heterodyne to direct conversion receiver started more than a
`year ago, which is a few years behind similar migration in
`GSM handsets. The CDMA transmitter architecture is
`
`dominated by the direct up-conversion, or the so-called
`“direct launch” architecture, which requires an interstage
`Tx filter to suppress spurs from the direct conversion
`modulator [1]. It is predicted that the superheterodyne Rx
`architecture in CDMA handsets will be phased out in the
`next couple of years, and only the ZIF architecture is dis-
`cussed here.
`
`the singlc-band and
`In their minimum implementation,
`dual—band FEs of CDMA handsets are reproduced from
`Qualcomm’s reference designs as shown in Figure 4. For
`each band, the FE consists of a duplexer, an Rx filter, a Tx
`interstage filter, and a linear PAM. In a dual-band FE, ei-
`ther a SPDT pHEMT switch or a diplexer is required. Since
`CDMA system adopts a linear modulation scheme,
`the
`SPDT has very stringent cross-modulation requirement,
`which is typically less than 400 dBc, while the all-passive
`diplexer is inherently linear.
`The position location feature through the Global Position-
`ing System (GPS) is becoming standard in mid- and high-
`tier CDMA phones. The simplest implementation of GPS
`receiver uses a separate antenna, and for all practical pur-
`poses,
`it can be treated as a standalone receiver from the
`rest of the CDMA radio. However, it is desirable to elimi-
`nate the GPS antenna to reduce the size and cost of the
`handset. To share a common antenna, either a SP3T
`pHEMT switch or a triplexer is required. The SP3T CDMA
`switch has similar cross-modulation requirement as the
`SPZT CDMA switch, while the all-passive triplexer, again,
`is inherently linear. The triplexer also offers simultaneous
`receiving of GPS signals while the full-duplexed CDMA
`radio is in use. On the contrary, the receiving of GPS sig‘
`nals can only be achieved when the CDMA radio with
`SP3T switch is idle. The tn'plexer comprises a LPF, a high-
`pass filter (HPF) and a GPS filter, which are all connected
`at the antenna port. The challenge in triplexer design is to
`minimize loading effects of the other two filters on the fil—
`ter’s pass band. The triple-band CDMA FE with triplexer
`implementation is illustrated in Figure 5.
`
`ta]
`
`c-II
`A“ Dualaxor
`
`
`
`
`RFtienoo__
`'RFLeouof
`
`
`RFTEiUD
`
`PCS
`DupluIf
`
`Figure 4. HF FE diagrams of CDMA (ls-95) handsets: (a)
`single band, and (b) dual bands.
`
`Coll
`Duplex-r
`
`aCI5:u!Au.D:
`
`‘
`
`RFREDOD
`
`RFTEiDD
`
`'lrlpllxtr
`
`Figure 5. RF FE diagram of a triple-band, dual-mode CDMA
`handset.
`
`Turning now to duplexer, SAW duplexer dominates the
`cellular-band market, bulk acoustic wave (BAW) duplexer
`enjoys the first-mover advantage over SAW in the US PCS
`band, while the bulky ceramic duplexer is quickly becom-
`ing obsolete. There are currently two competing BAW
`technologies for the PCS band duplexer as well as the PCS
`interstage Tx BPF, they are: film bulk acoustic resonator
`(FEAR) [3] and solidly mounted resonator (SMR) [4] tech-
`nologies. BAW resonator has a higher quality factor (Q)
`over SAW resonator near and above 2 GHz, where SAW
`IDT becomes metal-Q limited as both IDT metal thickness
`and finger width go down with frequency. Higher Q factor
`translates directly into lower insertion loss and steeper filter
`skirt. BAW resonator also exhibits better thermal stability,
`superior ESD robustness and better power handling capa-
`
`226
`
`2
`
`

`

`It should be noticed that the
`bility over SAW resonator.
`performance of SAW filter is more than adequate for hand—
`set applications at lower frequencies, where its maturity
`and commodity pricing are unmatched in the near future.
`As with GSM FE, the level of integration in CDMA FE is
`Currentiy being raised through combining the low- and
`high-band SAW filters into a 2-in-1 package, combining
`PAM with duplexer to optimize transmitter performance as
`well as TX/RX isolation, and finally, add the interstage Tx
`filter to the PA/Duplexer module to form a transmitter
`module.
`
`RF FRONT-END IN 2.5G and SG HANDSETS
`While the cellular industry is recovering from its worst
`recession since its inception in late 705, the cellular services
`is also in the midst of transition from the maturing 2G
`technologies first to 2.5G then to 30 systems. For GSM
`service providers, the migration paths are first to add the
`General Packet Radio System (GPRS) capabilities for im-
`proved data rate, which is basically a software upgrade,
`then migration to the Enhanced Data rates for GSM Evoluw
`tion (EDGE), which utilizes the 31:18 SPSK linear modula-
`tion scheme for higher data rates and improved spectrum
`efficiency
`over
`constant-envelope GSM modulation
`(GMSK), and eventually transition to 3G wideband CDMA
`(WCDMA), or the Universal Mobile Telecommunications
`system (UMTS).
`The hardware implication of EDGE FE is three folds: First,
`the OPLL Tx architecture has to be replaced mostly likely
`by the direct up-conversion one, which will add interstage
`Tx filter back to the Tx line-up; secondly, a linear PA is
`required; and lastly, a tinear antenna switch is needed to
`minimize
`cross—modulation
`interference. Although the
`specifications for EDGE FE blocks will be tightened,
`it
`looks very similar to the GSM FE at
`the block—diagram
`level with only one exception, namely, the interstage Tx
`filter before PA. Therefore, there is no reason to believe the
`paths of FE integration in EDGE handsets to be any differ-
`ent from that ofGSM handsets with the possibility that the
`interstagc Tx filter being added to PAM. Fortunately, linear
`PA and pHEMT switch are not new in 2G handsets, where
`linear transmitters are adopted in both Time Division Mul-
`tiple Access (TDMA) and CDMA systems; ,and linear
`SPZT and SPJT pHEMT switches are widely used in
`CDMA FE, albeit their requirements are different from the
`EDGE system. The GSM industry is pushing for the so—
`called “EDGE—capable“ GSM hardware, and future EDGE
`handsets will most likely be backward compatible to GSM
`system.
`
`The 3G WCDMA system involves new spectrum allocation
`at
`l900MHz (Tx) and ZlGOMHZ (Rx) bands, while its
`handset transceiver architecture closely resembles that of
`the 20 CDMA handsets. The WCDMA FE blocks include
`a duplexer, 21 TX interslage filler, an Rx filter and a linear
`PA. The most
`interesting aspect regarding WCDMA FE
`lies in the fact that WCDMA handsets will carry at least
`one or more GSM/GPRS/EDGE/CDMA transceivers for
`
`backward compatibility in voice communication in the
`foreseeable future. In other words, 3G UMTS phones will
`be universally multi-bands and multi-modes. Figure 6 illus-
`trates an example FE of a UMTS and dual-band GSM
`handset. It is important to note that the FE switch module in
`Figure 6 provides the dual
`functionalities of antenna
`switching for GSM as well as onloff switching for UMTS,
`which has to meet stringent
`linearity requirement
`for
`UMTS.
`
`
`
`Figure 6. FE ot a dual-mode UMTS and dual-band GSM
`handset.
`
`Without new spectrum allocation, the evolutional paths for
`2G CDMA (IS—95)
`system, dubbed CDMAZOOO, are
`mostly backward compatible and relatively straightforward.
`CDMAZOOD 1x reflects the multi-carrier nature of the en-
`hanced system, where Ix indicates the utilization of one
`LZSMHZ channel. The voice capacity is doubled in
`CDMA2000 Ix through incremental improvements in radio
`hardware such as faster power control. reduced rate of
`voice coding, transmitter diversity and coherent demodula-
`tion. The data rates enhancement
`in CDMAZOOO lx is
`
`achieved through less efficient 3x spreading rates (three
`LZSMHZ channels). CDMAZOOO leV-DO offers an evo-
`lutional path for CDMA Ix optimized for data communica-
`tion, which is not compatible to voice channels, and has to
`bc overlaid with the 18—95 or CDMA lx system. Finally,
`CDMA2000 leV—DV, which stands for integrated data
`and voice, brings back backward compatibility to lS-95 and
`CDMA2000 lx systems with a unified air interface for both
`circuit—switched voice and packct~switched data. For the
`purpose of present
`discussion,
`the FE portion of
`CDMA2000 handset
`is identical
`to CDMA handsets as
`shown in Figs. 4 and 5. Qualcomm is proposing further
`capacity improvement over CDMA2000 though advanced
`speech coding techniques, 4-way diversity receivers at has-
`cstation and diversity receivers in handsets. A typical FE of
`CDMA2000 handsets with diversity receivers is illustrated
`in Figure 7.
`
`227
`
`3
`
`

`

`
`Dlpleur
`
`.arsesat:..
`
`
`''3RFLsootii'
`
`
`
`
`-aFTmeo
`
`Figure 7. FE of a CDMAZOOO handset with diversity receiv-
`ers.
`
`CONCLUDING REMARKS
`Integration of RF FE in 2G handsets has brought about
`miniaturized handsets, reduced bills of material (BOM),
`low cost handsets, shortened design cycles, and faster time-
`to-markct. This trend of RF FE integration will undoubt-
`edly continue into the emerging 3G handsets, however,
`there are practical limits on the roads to a fully integrated
`FE such as system on chip (SOC) or system in package
`(SIP). To understand these limitations, one needs to take a
`closer look of the underlying economy of the handset in-
`dustry, where pricing of handsets is dictated by service
`providers. In lieu of free-market competition, service pro-
`viders subsidize handsets in exchange for long—term service
`contracts. This somewhat arbitrary pricing structure only
`worked due to the sheer units volume involved, e.g., more
`than 600 millions handsets are projected to be shipped
`globally this year. Economy of scale is the only saving
`grace throughout
`the food chains of handsets industries,
`
`which allows very little profit margins but demands high-
`tech products of the highest quality. The volume helps
`spreading sizable capital
`investments to a manageable
`level, while opens doors for technology innovations geared
`towards cost reduction in a high volume environment. With
`this background, it is arguable that a highly integrated FE
`module may not be so desirable due to the market fragmen-
`tation nature of such approach. Instead, RF FE modules
`suppliers should focus on FE partitions where economy of
`scale in design and manufacturing can be easily realized.
`Existing FE modules that fall into such category include
`ASM, filter bank module, PAM,
`triplexer, and duplexer.
`Combination of any of these blocks further may scale back
`its total addressable markets (TAM) more than in half from
`its constituents’ TAM. Taking away the last saving grace
`may spell the demise of many of the participants in this
`unforgiving handsets industries.
`
`HEFEFI ENCES
`[1} RF Micro Devices, Greensboro, NC, “Reduced Filter
`Requirements Using an Ultra Low Noise Modulator,"
`Cover Feature, Microwave Journal, vol. 44, no. 1, Jan.
`2001.
`
`[2]
`
`i3]
`
`[4]
`
`“The TQPED pHEMT process enables customers to
`achieve higher levels of integration in RF front-end de-
`sign for cell phones and other wireless systems,” Tri-
`Quint Semiconductor Press Release, Dec. 15, 2003.
`Rich Ruby, “FBAR — From Technology Development
`to Production,” Second International Symposium on
`Acoustic Wave Devices for Future Mobile Communi-
`
`cation Systems, Chiba University, Chiba, Japan, Mar.
`3-5, 2004.
`Robert Aigner, “Volume manufacturing of BAW-
`filters in a CMOS fab,” Second lntemational Sympo-
`sium on Acoustic Wave Devices for Future Mobile
`Communication Systems, Chiba University, Chiba, Ja~
`pan, Mar. 3-5, 2004.
`
`228
`
`4
`
`