TECHNICAL FEATURE

MHz (corresponding to the chip rate). Thedownlink employs quadrature phase-shift key-ing (QPSK) modulation. Root-raised-cosinefiltering is applied to shape the spectrum. Us-ing orthogonal spreading and gold-codescrambling, several CDMA channels are mul-tiplexed onto the same frequency channel.2Hence, the received signal consists of many si-multaneously transmitted channels that usethe same carrier frequency. As a result, largeamplitude variations occur over time. The up-link is similar but uses a more complicated hy-brid-QPSK modulation scheme. Although acombination of code allocation and complexscrambling is used to minimize the number of

O.K. JENSEN, T.E. KOLDING, C.R. IVERSEN, S. LAURSEN, R.V. REYNISSON, J.H. MIKKELSEN, E. PEDERSEN, M.B. JENNERAND T. LARSENAalborg University, RISC GroupAalborg, Denmark

TECHNICAL FEATURE

For RF designers who are experiencedwith 2G TDMA/FDMA wireless sys-tems, the introduction of W-CDMA re-

quires some change of mind. First, rather thanbeing separated in frequency or time, users arenow separated by orthogonal codes. As the useof codes implies a spectral spreading, the treat-

ment of overall signal-to-noiseratio requires considerationsthat are different from those re-quired for TDMA systems. Sec-ond, a single radio channel be-haves more like band-limitednoise than a single sinusoid. Sta-tistical terms like peak-to-aver-age power ratio are thereforenecessary to reflect this newconstellation of signals.

General characteristics ofthe UTRA/FDD system arelisted in Table 1.1 The nominalfrequency spacing between ad-jacent channels is 5 MHz andthe signal bandwidth is 3.84

RF RECEIVERREQUIREMENTSFOR 3G W-CDMA MOBILE EQUIPMENTThe first standardization phase for upcoming third-generation (3G) wirelesscommunications is coming to an end. As is typical for standardization work,sufficient analog performance has been assumed and predominant emphasis hasbeen placed on modulation and coding. However, recent revisions to thestandardization document for the European wideband CDMA (W-CDMA)proposal,1 known as UTRA/FDD, enable a prediction of required performancefor the RF front end. Such requirements are important when preparingcommercial products for the new market. In this article, receiver requirementsfor the mobile unit are derived in terms recognizable by the RF designer.

TABLE IUTRA/FDD W-CDMA

SYSTEM CHARACTERISTICS

Parameter Specification

Uplink frequency band (Tx) (MHz) 1920 to 1980

Downlink frequencyband (Rx) (MHz) 2110 to 2170

Tx-to-Rx frequency separation (MHz) 134.8 to 245.2

Nominal channel spacing (MHz) 5

Chip rate (Mcps) 3.84

mate of duplex cir-cuit performance islisted in Table 2.The 4 dB loss in thereceive path has se-vere implications forthe overall noise fig-ure, however, since asmall physical size ismandatory, it ap-pears inevitable. Thepower level of thetransmitter leakagesignal at the receiverinput is determinedby taking the trans-mitter power class,adding the specified

tolerance (+1/–3 or +2/–2 dB), addingthe transmit path loss (2.5 dB) andsubtracting the expected duplex filterisolation. The result is between –23.5and –34.5 dBm and, since it is still be-ing debated which maximum powerlevels will apply to hand-held units, aspurious transmitter level around –30dBm is expected to be typical. The re-ceiver must be able to handle this sig-nal without significant performancedegradation.

In the next section, the test casesspecified in the standardization docu-ment1 are studied in detail and issuesof noise, second-order distortion,third-order intermodulation, selectiv-ity and oscillator phase noise aretreated. From this treatment, perfor-mance requirements for the W-CDMA receiver are derived. Notethat all derived expressions assumeinsertion in decibel/decibel relative to1 mW (dB/dBm) numbers and thatsignal powers are specified over achannel bandwidth (3.84 MHz) as inthe UTRA/FDD standard.1 A com-plete list of symbols and conventionsused in this paper can be found in thesidebar on the following page. (Re-ceiver requirements are only approxi-mate since they do not relate to aspecific receiver architecture.) Givena detailed architecture, more accu-rate requirements can be derived bystudying the test cases specified bythe standard.1 However, for the pur-pose of assessing challenges faced byRF designers of 3G wireless equip-ment, the treatment is adequate. Thisarticle concludes with a direct-con-version receiver example, which illus-trates a possible implementation thatcomplies with derived requirements.

TEST CASESThe UTRA/FDD standard1 de-

scribes a number of test scenarios inwhich the user bit rate is fixed at 12.2kbps and the bit error rate (BER) mustbe below 10–3. The desired downlinkchannel signal includes two or more or-thogonal CDMA channels, which com-prise the dedicated physical channel(DPCH) carrying the user data, a syn-chronization channel and, in some cas-es, other users’ data channels. The stan-dard specifies total power levels withinthe channel bandwidth and the relativelevel of the DPCH. For simplicity, thedesired channel power is specified asthe DPCH channel power throughoutthis article.

In the baseband receiver, the de-spreading process concentrates thedesired signal energy in a bandwidththat corresponds to the channel sym-bol rate. Since noise and interferenceare uncorrelated with the despread-ing code, noise is not concentrated ina smaller bandwidth. Further, signaldecoding results in a coding gain, andthe total resulting improvement insignal-to-noise ratio is defined as theuser data processing gain given by3

Note that this notation differs fromthe standard CDMA processing gaindefinition, which relates correlationtime to chip time.6 However, the cho-

GMcpskbps

dB

P =

=

103 8412 2

25

10log.

.

TECHNICAL FEATURE

signal nulls,2 the envelope of thetransmitted signal continues to dis-play large amplitude variations. Thesevariations place high linearity re-quirements on the power amplifier,which is believed to be a major RFdesign challenge.

Until the new 3G system deliversthe coverage and services offered bythe well-established 2G systems, mul-timode terminals with both 2G and3G capabilities are required. Such atransceiver system configured for twowireless systems is shown in Figure1. In the transceiver system, a systemselect switch is used for selection be-tween a 2G 900 MHz system (E-GSM) and a 3G W-CDMA system.The 2G system applies time divisionduplex (TDD) as well as frequencydivision duplex (FDD), and a duplexswitch is used to select transmit or re-ceive modes. Since the consideredW-CDMA system only applies FDDand thus employs simultaneous trans-mission and reception, a duplex filteris required to provide isolation be-tween the transmitter and the receiv-er. The continuous presence of thehigh power transmitter signal causesproblems with spurious leakage fromthe transmit band located at a 134.8to 245.2 MHz offset. Unless suffi-cient selectivity is available betweenthe Tx and Rx bands, this spurioustransmitter signal will cause severedynamic range and intermodulationproblems in the receiver chain.

From the previous comments, itshould be clear that a high perfor-mance duplex circuit with good Tx-Rxisolation is needed. Based on data forswitches and ceramic duplex filtersavailable commercially today, an esti-

W-CDMA

TRANSMITTER

RECEIVER

TRANSMITTER

RECEIVER

2G 900 MHz

DUPLEX FILTER

SYSTEMSELECT SWITCH

DUPLEX SWITCH

▲ Fig. 1 Example of a duplex arrangement of a mobile receiver unitincluding 3G W-CDMA and 2G TDMA systems.

TABLE IIANTICIPATED PERFORMANCE

FOR W-CDMA DUPLEX ARRANGEMENT

Duplexer Anticipated Parameter Performance

Duplex filter Tx loss (dB) < 1.5

Duplex filter Rx loss (dB) < 3.0

Loss of system select switchand antenna feed (dB) < 1.0

Combined Tx loss (dB) < 2.5

Combined Rx loss (dB) 2.0 to 4.0

Duplex filter Tx-Rxisolation in Tx band (dB) > 60

Transmitter power classes(hand-held and 33/27/24/21fixed-mounted units) (dBm)

Typical transmitter leakagesignal at receiver input (dBm) –30

sen definition facilitates a more gen-eral comparison of different systems.The required minimum Eb/Nt for aBER of 10–3 is determined from sim-ulations to be 5.2 dB.3 The termEb/Nt is used here instead of the tra-ditional notation Eb/N0 since mosttests include interference in additionto noise. It is also suggested that animplementation margin be added toaccount for various baseband imper-fections.3 The required effectiveEb/Nt is then expressed as

which complies with the chosen defi-nition of processing gain.

NOISE FIGUREThe noise figure (NF) of the

UTRA receiver is calculated from thestandard’s reference sensitivity test.The desired channel power isPR,DPHC = –117 dBm. Using the pre-viously determined (Eb/Nt)eff re-quirement and including the userdata processing gain, the maximumallowable noise power within thechannel bandwidth is calculated to be

When the NF of the receiver and thebandwidth (BW) are known, the ac-tual noise power is determined using

where

k = Boltzmann’s constantT0 = standard noise temperature

Since the actual noise power must belower than or equal to the acceptablenoise power, the NF requirement is

NF P acceptable dBm

dBm dBmdB

N≤ ( ) +

= +=

108

99 1089–

P actual NF

k T BW

NF dBWNF dBm

N ( ) = +

• •( )==

10

138108

10

0

log

––

P acceptableN ( )=

+

= +=

PEN

G

dBm dB dBdBm

R DPHCb

t effp, –

– ––117 7 2599

EN

dBb

t eff

≈ 7

This NF require-ment is for the en-tire receiver. Sub-tracting the loss of 4dB in the duplexcircuit, the NF re-quirement for therest of the receiveris 5 dB. This levelappears to be with-in reach for low cost integrated re-ceivers. It should benoted that the NFmust be met in thepresence of thetransmitter leakage signal.

ADJACENT-CHANNELSELECTIVITY

Adjacent-channel selectivity is de-fined as the relative attenuation of theadjacent-channel power. Selectivity in-cludes filtering at the IF, analog base-band and digital baseband, and the fre-quency sensitivity of the demodulator.A test setting requirement for the firstadjacent-channel selectivity is shown in

Figure 2. In this test, the desired signalpower is PR,DPHC = –103 dBm. Sincethis level is 14 dB above the sensitivitylimit, noise is of minor importance. Thefirst adjacent channel has a power ofPAC1 = –52 dBm centered around a 5MHz offset.

Treating the adjacent-channel sig-nal as noise, the required first adja-cent-channel selectivity can be de-rived. The acceptable interferencelevel, PI, is determined in the same

TECHNICAL FEATURE

MODULATEDADJACENT

CHANNEL −52 dB

ATTENUATED ADJACENTCHANNEL

DESIRED SIGNAL−103 dBm

INP

UT

REF

ERR

ED P

OW

ER (

dBm

)

5 MHz OFFSET

ACCEPTABLEINTERFERENCE LEVEL = −8.5 dBm

Gp(Eb/Nt)eff

SELECTIVITY

▲ Fig. 2 Test for adjacent-channel selectivity.

Note that all power levels and in-tercept points are given in decibelsrelative to 1 mW (dBm), power lev-els of modulated signals are mea-sured within a channel bandwidth(3.84 MHz) and all gain loss valuesare given in decibels (dB).BER: bit error rateBW: channel bandwidth

(3.84 MHz)DPCH: dedicated physical

channelEb/Nt: ratio of average bit energy

to noise and interferencepower spectral density

(Eb/Nt)eff: effective Eb/Nt includingan implementationmargin

GP: user data processing gainIIP2: second-order intercept

point referred to theinput

IIP3: third-order interceptpoint referred to theinput

k: Boltzmann’s constant of 1.38 × 10–23 J/K

NF: noise figure

PAC1: power of the firstadjacent channel

PBLOCK: power of blocker signalPBLeak: power of blocking signal

leaking to thedemodulator

PI: power of intermodulationproduct

PI3: power of third-orderintermodulation products

PINT: interfering signal powerPN: noise powerPN+1: noise and interference

powerPR,DPCH: received DPCH channel

powerPTxLeak: power of transmitter

leakage signalP2DIS: power of second-order

distortion productsP2DISeff: effective power of

second-order distortionproducts (after removal ofDC and componentsabove the signalbandwidth)

T0: standard noisetemperature of 290 K

NOTATION, SYMBOLS AND ABBREVIATIONS USED IN THIS ARTICLE

and lowpass filtered second-orderproducts of the blocker signal withpower P2DISeff and blocker leakagearound 15 MHz at baseband withpower PBLeak. Since the power of thedesired signal in this test is 3 dB high-er than for the sensitivity test, it is as-sumed that noise constitutes 50 per-cent of the total disturbing power. Forsimplicity, the remaining power is di-vided equally (25 percent 6 dB) be-tween the second-order products andthe blocker leakage. The acceptablenoise plus interference level measuredat the antenna input is expressed as

and the acceptable levels are

Note that P2DISeff is the effectivelevel of second-order distortion thatcan be tolerated. However, as a 6 dBimprovement due to baseband filter-ing is assumed, an additional 6 dB ofsecond-order distortion actually canbe accepted. Hence, in the calcula-tion of the required IIP2, the value ofP2DISeff is corrected to P2DIS =P2DISeff + 6 dB. Consequently, the re-quirement to the second-order inputintercept point is obtained using

The necessary selectivity for a chan-nel at 15 MHz offset is found to be

In the specification document, anadditional blocker test is specified with

Selectivity MHz

P P

dBm dBm

dBm

BLOCK BLeak

15

44 102

58

( )≥

= ( )=

–

– – –

IIP MHz

P P

dBm dBm

dBm

BLOCK DIS

2

2

15

2

2 44 102 6

8

( )≥

= ( ) +( )=

–

– – –

P P dB

dBmandP P

P dB

dBm

N N I

BLeak DISeff

N I

==

===

+

+

–

–

–

–

3

99

6

102

2

P PEN

G

dBm dB dBdBm

N I R DPCHb

t effP+ =

+

= +=

, –

– ––114 7 2596

TECHNICAL FEATUREmanner as was used during the NFcalculation:

and the adjacent-channel selectivityrequirement at 5 MHz is

SECOND-ORDER INTERCEPT POINTS

Low even-order distortion, espe-cially second-order distortion, is cru-cial to the receiver’s performance be-cause of the presence of strong mod-ulated signals with time-varyingenvelopes. When a second-ordernonlinearity is exposed to such a sig-nal, a spurious baseband signal pro-portional to the squared envelope isgenerated at baseband, which dis-turbs the reception of the desired sig-nal. Two such groups of disturbingsignals are present: unwanted chan-nels in the receive band (downlink)and the transmitter leakage signal.(The problem of unwanted channels

Selectivity MHz

P P

dBm dBmdB

AC I

5

52 8533

1

( )≥==

–

– –

P PEN

G

dBm dB dBdBm

I R DPCHb

t effP=

+

= +=

, –

– ––103 7 2585

in the receive band is addressed inthe in-band blocker test.1)

The spectral shape of these signals isthe same as for the wanted signal (root-raised-cosine) but the spectral shape ofthe second-order product is different,as shown in Figure 3. A significant DCcomponent is present and the spec-trum is broader than the desired base-band signal. Typically, the DC compo-nent represents 50 percent of the pow-er while 50 percent of the remainingpower lies above the desired signalbandwidth. Consequently, a combina-tion of highpass and lowpass filteringcan improve Eb/Nt by approximately 6dB. Highpass filtering of the W-CDMAsignal can be accomplished with negli-gible degradation of performance dueto the large-signal bandwidth.4 Howev-er, the actual improvement in Eb/Ntdepends on the present signal configu-ration and is different for uplink anddownlink signals. According to simula-tions, the possible suppression of thespurious second-order product rangesfrom 4 to 13 dB.

IIP2 is determined by the in-bandblocker test. The desired signal has apower of PR,DPCH = –114 dBm. Themodulated blocker has a power ofPBLOCK = –44 dBm and is offset infrequency by a minimum of 15 MHz.In this scenario, the demodulator ex-periences three sources of interfer-ence: noise with power PN, highpass

DESIREDSIGNAL

0 5BASEBAND FREQUENCY (MHz)

NOISE

NOISE

MINIMUM 15 MHz OFFSET

(a)

(b)

SECOND-ORDERPRODUCTS

10 15

SIG

NA

L LE

VEL

DESIRED SIGNAL= −114 dBm

MODULATEDBLOCKER SIGNAL

−44 dBm

BLOCKERLEAKAGE

DOWNCONVERSIONAND FILTERING

SECOND-ORDERNONLINEARITIESDOWNCONVERSION

▲ Fig. 3 In-band modulated blocker test; (a) RF spectrum with desired signal and offsetmodulated blocker, and (b) baseband spectrum with desired and disturbing signals.

must be taken into account. Assum-ing that the third-order intermodula-tion product of the two interferersmay be treated as noise, the maxi-mum level of noise and interferenceis found to be

where PN+I is referred to the antennainput.

In this test case, several interferingproducts are created, thus the allow-able noise and interference power(PN+I) must be distributed. The as-sumed power distribution is: noise, 50percent of power (–3 dB); intermodula-tion, 15 percent of power (–8 dB); CWinterferer’s blocking effect, 15 percentof power (–8 dB); modulated interfer-er’s blocking effect, 15 percent of pow-er (–8 dB); and oscillator noise, fivepercent of power (–13 dB). Second-or-der distortion products are neglected.The power level corresponding to eachof the interfering or blocking productsis then PN+I – 8 dB = –104 dBm. Thispower level, along with the relationshipbetween intermodulation power level

P PEN

G

dBm dB dBdBm

N I R DPCHb

t effP+ =

+

= +=

, –

– ––114 7 2596

and input intercept point, gives theminimum receiver IIP3:

IIP3(10/20 MHz)

In addition, the test results in two se-lectivity requirements are

and

The out-of-band CW blocker test,shown in Figure 5, indirectly sets anIP3 requirement for the receiver. If aCW blocker is present at some fre-quency distance (for example, 67.4

Selectivity MHz

P P dB

dBm dBm

dB

INT N I

20

8

46 104

58

( )≥ ( )= ( )=

+– –

– – –

Selectivity MHz CW

P P dB

dBm dBm

dB

INT N I

10

8

46 104

58

,

– –

– – –

( )≥ ( )= ( )=

+

P P P dB

dBm dBm

dBm

INT INT N I12

8

4612

46 104

17

≥ + ( )( )= + ( )( )=

+– –

– – – –

–

TECHNICAL FEATUREa blocker power of –56 dBm and a 10to 15 MHz frequency offset. Corre-sponding to previous calculations, theIIP2 requirement is derived using

Finally, the required out-of-band IP2is determined by the transmitter leak-age level at the receiver input. Thedisturbance mechanisms are thesame as those shown previously butthe blocker signal is replaced by thetransmitter leakage signal. Since theduplex distance is a minimum 134.8MHz, the direct transmitter leakagethrough the receiver to the demodu-lator is insignificant.

Since the transmitter signal is al-ways present, the second-order prod-ucts should be sufficiently suppressed(for example, 10 dB below the noiselevel). A rough estimate of IIP2 isthen determined using

The noise level has been correctedfor the 4 dB duplexer loss, thusIIP2(Tx) refers to the circuit after theduplexer. Again, a 6 dB improvementdue to highpass and lowpass filteringhas been assumed.

THIRD-ORDER INTERCEPT POINTS

The third-order intercept point ofthe UTRA receiver is determined us-ing the intermodulation test de-scribed in the UTRA standard. Theintermodulation test scenario isshown in Figure 4. The desired sig-nal is at PR,DPCH = –114 dBm, 3 dBabove the minimum sensitivity. Twointerfering signals are offset 10 and20 MHz from the desired signal. Thefirst interferer is a CW signal at PINT= –46 dBm; the second interferer is amodulated signal with a power ofPINT = –46 dBm. As the desired sig-nal is close to the minimum sensitivi-ty level, both noise and interference

IIP Tx P P dB

dB

dBm

dBm

dBm

TxLeak N2 2 6

10

2 30

99 4 6 10

47

( ) ≥ +()

= ( )+( )

=

–

–

– –

– – –

IIP MHz

P P

dBm dBm

dBm

BLOCK DIS

2

2

10

2

2 56 102 6

16

( )≥

= ( ) +( )=

–

– – –

–

Tx-BAND Rx-BANDFREQUENCY (GHz)

−15 dBm −15 dB

−15

−40

2.0 2.1 2.2 2.3

−30 dBm −30 dB

−44 dBm −44 dB

85 MHz 85 MHz

60 MHz 60 MHz

15 MHz 15 MHz

INP

UT

REF

ERR

ED P

OW

ER (

dBm

)

▼ Fig. 5 Out-of-band CW blocker test.

DESIRED SIGNAL−114 dBm

20 MHz OFFSET

ATTENUATEDINTERFERERS

MODULATEDINTERFERER

−46 dBm

10 MHz OFFSET

Gp

(Eb/Nt)eff

ACCEPTABLE NOISE +INTERFERENCE = −96 dBm

INP

UT

REF

ERR

ED P

OW

ER (

dBm

)

THIRD-ORDERNONLINEARITIES

SELECTIVITY(10 MHz)

CWINTERFERER

−46 dBm

SELECTIVITY(20 MHz)

NOISE

▲ Fig. 4 Intermodulation test.

MHz) from the receive band and thetransmitter leakage signal is locatedat the double frequency distance(134.8 MHz), then a third-order in-termodulation product is created atthe receive frequency in the sameway as the intermodulation test de-scribed previously.

The level of the CW blocker and theattenuation in the duplex filter dependon the frequency band. For typical du-plex filters, the CW blocker level afterthe duplex filter is below –45 dBm. Forcalculation of the IIP3, the transmitterleakage signal of –30 dBm and the CWblocker of –45 dBm are replaced withtwo signals of equal level such that

since the interferer closest to the de-sired signal has the highest weight inthird-order intermodulation. Becauseof the large frequency offset, block-ing effects are negligible and the al-lowable interference level is PN+I – 3dB since noise contributes 50 percentof the power. Correcting this levelwith the duplexer loss of 4 dB, anIIP3 is found to be

This number is for the receiver partafter the duplexer and is highly de-pendent on the actual duplexer char-acteristics.

IMAGE REJECTIONThe required image rejection also

can be determined from the out-of-band blocker test. If the image fre-quency is distanced at more than 85MHz from the receive band, the nec-essary rejection is

Image rejection (> 85 MHz)

– –

– – – –

P P dB

dBm dBm dB

dB

BLOCK N I≥ ( )( )

=

+ 3

15 96 3

84

IIP MHz

P P P

dB dB

dBm dBm

dBm dB dB

dBm

INT INT N I

3 67 4 134 8

12

3 4

4012

40

96 3 4

8

. / .

– –

–

– – –

– – –

–

( )≥ + ((

))= + (

( ))≈

+

P dBm dBm

dBm

INT = ( ) + ( )=

13

3023

45

40

– –

–

In the calculation it has been notedthat noise constitutes 50 percent ofthe disturbing power PN+I.

OSCILLATOR PHASE NOISEThe presence of blocking signals

sets requirements for the LO noisesidebands. The worst-case scenario isprobably set by the intermodulationtest where a –46 dBm CW blocker ispresent at a 10 MHz offset. Allowingfive percent of the disturbing powerin this test to come from the oscilla-tor noise, the LO noise power mustbe below PN+I – 13 dB = –109 dBm.This level corresponds to –63 dBcmeasured over a 3.84 MHz band-width when taken relative to the CW-blocker carrier. This number can betransferred to the LO since it repre-sents a cross-modulation phenome-non. The spectral shape of the LOnoise sidebands is unknown, however,taking the case of a flat spectrumseems reasonable because of thelarge frequency offset. With this as-sumption, the –63 dBc over 3.84MHz correspondsto a power spectraldensity of –129dBc/Hz. This speci-fication must bemet for an offsetfrom the LO carrierof more than 10MHz – BW/2 ≈ 8MHz.

REQUIREMENTSUMMARY

A summary of re-ceiver requirementsfor the entire re-ceiver and the partof the receiver thatfollows the duplexcircuits is listed inTable 3.

RECEIVER EXAMPLEBecause of its high noise and se-

lectivity performance, the heterodynereceiver architecture has been thepreferred architecture for the 2Gmarket. However, as issues of powerconsumption, cost and size becomemore critical than ever, other archi-tecture candidates are being ex-plored. Recently, homodyne (or di-rect-conversion) receivers haveemerged for the GSM market andmany consider this architecture theproper choice for 3G systems. A par-ticular nicety is the fact that the largesignal bandwidth makes W-CDMAsystems less susceptible to the 1/fnoise and DC off-set problems inher-ent in homodyne receivers.

An example of an UTRA/FDD di-rect-conversion receiver is shown inFigure 6. An interstage bandpass fil-ter is used to provide sufficient atten-uation of the transmitter leakage sig-nal. Sufficient selectivity with low in-sertion loss is possible withcommercial ceramic or surface

TECHNICAL FEATURE

TABLE IIISUMMARY OF REQUIREMENTS

Requirement Entire Receiver After Duplex

Noise figure (dB) ≤ 9 ≤ 5

In-band selectivity (dB)first adjacent channel (5 MHz) ≥ 33 ≥ 33CW interferer (10 MHz) ≥ 58 ≥ 58third adjacent channel (15 MHz) ≥ 58 ≥ 58modulation blocker (> 15 MHz) ≥ 58 ≥ 58

Intercept points (dBm)IIP2 (10 MHz) ≥ –16 ≥ –18IIP2 (15 MHz) ≥ 8 ≥ 6IIP2 (Tx) ≥ 47IIP3 (10/20 MHz) ≥ –17 ≥ –19IIP3 (67.4/134.8 MHz) ≥ –8

Image rejection (> 85 MHz) (dB) ≥ 84 n/a

Oscillator noise sidebands at> 8 MHz offset (dBc/Hz) ≤ –129

DUPLEX LNA1 LNA2

MIXERBB-FA

LO0°

90°

Q

I

BPF

▲ Fig. 6 Direct-conversion receiver including filter.

acoustic wave (SAW) filters. The filtermay have a single-ended or balancedoutput, however, the second lownoise amplifier (LNA) stage shouldhave a balanced output to facilitategood even-order distortion perfor-mance of the mixers.

Overall block requirements arelisted in Table 4. Due to the inter-stage bandpass filter, the overall out-of-band IIP3 performance is domi-nated by the first LNA and this blockmust comply with the overall require-ment of –8 dBm. Note that the IIP3has been set slightly higher at –3dBm to relieve the performance re-quirement to the mixer and BB-FAblocks. For the direct-conversion re-ceiver, the mixer and baseband unitare critical components for the over-all linearity and noise while most gainis placed with the baseband unit. IIP2and IIP3 requirements to the mixerand baseband units are difficult butcertainly realistic. With respect tonoise, the baseband unit constitutes amajor challenge if low cost technolo-gies such as CMOS are to be used.However, recent work has shownmuch progress in this area.5 In theexample, 1 dB of gain tolerance ofthe first receiver blocks is assumed. Ahigher tolerance makes it difficult tomeet any worst-case specification.

Most of the Tx signal suppression isobtained in the duplex and interstagebandpass filters. The interstage band-pass filter also attenuates out-of-bandblockers but, since the filter character-istics are typically less steep in the up-per stop band, only limited selectivitytoward these disturbing signals can be

assumed. Hence, the BB-FA unit mustimplement most of the required selec-tivity derived in this article. In-banddisturbances can be suppressed only atbaseband. In this sense, one of themost fundamental decisions is the dis-tribution of filtering requirements be-tween analog and digital hardware.With the high chip rate employed in3G W-CDMA systems, constraints oflow power limit the available analog-to-digital converter (ADC) resolution.

Assuming a sampling frequency of15.36 MHz (corresponding to foursamples per chip), it is not possible touse digital filtering with the third adja-cent channel and the first modulatedblocker (> 15 MHz) due to aliasing.Hence, these interferers must be sup-pressed in analog hardware prior tosampling. With the first and second ad-jacent channels each additional bit en-ables 6 dB of digital selectivity. Prelimi-nary analyses7 indicate that four to fivebits are required as minimum ADCresolution (no digital selectivity), and 8to 10 dB bits immediately appears tobe a necessary choice. This selectivitymay be combined with a fourth-orderButterworth analog filter (2.5 MHzcutoff frequency) to achieve the listedselectivity requirements.

Issues other than those consideredhere are critical to the implementationof the direct-conversion receiver. Suchissues include DC offsets, in-phase andquadrature gain/phase imbalanceproblems, LO-to-antenna leakage andsusceptibility to 1/f noise. Basebandhighpass filtering should be combinedwith adaptive correction for sufficientperformance of this architecture.4

CONCLUSIONDetailed specifications for a W-

CDMA receiver were derived fromthe tests presented in the standard-ization documents. It was shown thata direct-conversion receiver architec-ture with reasonable block require-ments is capable of meeting derivedspecifications. The problems causedby the continuous presence of thetransmitter leakage signal emphasizethe need for a high performance du-plex filter. There are still implemen-tation challenges left for RF design-ers, but the specifications seem rea-sonable enough to facilitate thedesign of the low power, low cost,hand-held multimedia products thatthe world is awaiting.

ACKNOWLEDGMENTThe RISC Group would like to ac-

knowledge the helpful discussionswith local industry partners, includingBosch Telecom Denmark, MaxonCellular Systems (Denmark), RTXTelecom, Telital and Texas Instru-ments Denmark. ■

References1. Third Generation Partnership Project

(3GPP), “UE Radio Transmission and Re-ception (FDD),” Technical Specification25.101, Vol. 3.0.0, October 1999.

2. Third Generation Partnership Project(3GPP), “Spreading and Modulation(FDD),” Technical Specification 25.213,Vol. 3.0.0, October 1999.

3. TSG-RAN Working Group 4, Nokia Mo-bile Phones, “MS Receiver Sensitivity inUTRA FDD Mode,” Document TSGW4#1(99)005, January 1999.

4. J.H. Mikkelsen, T.E. Kolding, T. Larsen, T.Klingenbruun, K.I. Pedersen and P. Mo-gensen, “Feasibility Study of DC OffsetFiltering for UTRA/FDD/WCDMA Di-rect-conversion Receiver,” Proceedings ofthe 17th IEEEE NORCHIP Conference,Oslo, Norway, November 1999, pp. 34–39.

5. J. Jussila, A. Pärssinen and K. Halonen,“An Analog Baseband Circuitry for aWCDMA Direct Conversion Receiver,”Proceedings of the 25th European Solid-state Circuits Conference (ESSCIRC),Duisburg, Germany, September 1999, pp. 166–169.

6. A.J. Viterbi, CDMA: Principles of SpreadSpectrum Communication, Addison-Wes-ley Publishing Co., 1995.

7. B.N. Vejlgaard, P. Mogensen and J.B.Knudsen, “Performance Analysis forUMTS Downlink Receiver with PracticalAspects,” Proceedings of IEEE VehicularTechnology Conference (VTC), Amster-dam, Netherlands, September 1999, pp. 998–1002.

TECHNICAL FEATURETABLE IV

BLOCK SPECIFICATIONS FOR A DIRECT-CONVERSION RECEIVER

Block Duplexer LNA1 BPF LNA2 Mixer BB-FA Combined Required

Gain in Rx band (dB) –3 ±1 15 ±1 –2 ±1 8 ±1 10 ±1

Rx/Tx selectivity (dB) ≥ 60 ≥ 0 ≥ 21 ≥ 0 ≥ 0 ≥ 0

Noise figure (dB) ≤ 4 ≤ 2.5 ≤ 3 ≤ 3 ≤ 15 ≤ 25 ≤ 8.8 ≤ 9

IIP2 (15 MHz) (dBm) ≥ 37 ≥ 47 ≥ 8.5 ≥ 8

IIP2 (Tx) (dBm) ≥ 37 ≥ 47 ≥ 48.5 ≥ 47

IIP3 (67.4/134.8 MHz) (dBm) ≥ –3 ≥ –3 ≥ –8

IIP3 (10/20 MHz)(dBm) ≥ –3 ≥ 3 ≥ 10 ≥ 20 ≥ –16.8 ≥ –17

Note: The numbers for IIP2 (Tx) and IIP3 (67/135 MHz) are for the receiver part after the duplexer. Other numbers are combined values for the entire receiver, including duplex filter and system select switch.

Ole Kiel Jensen received his MSc degree inelectrical engineering from Aalborg University.He is now an associate professor at AalborgUniversity.

Troels Emil Kolding finished his PhDdissertation in 1999 and is currently a designmanager with RISC Group.

Christian Rye Iversen received his MSEEdegree from Aalborg University in 1995 and iscurrently working toward an industrial PhDdegree. He is also an RF design engineer atBosch Telecom.

Soren Laursen received his MSEE degree in1997 from Aalborg University. He is now aPhD student at Aalborg University.

Ragnar V. Reynisson received his MSEEdegree from Aalborg University in 1999. He iscurrently pursuing his PhD degree.

Jan H. Mikkelsen received his MSEE degreein 1995 from Aalborg University. He iscurrently working as a research engineer atAalborg University.

Erik Pedersen received his MSEE degreefrom Aalborg University in 1988, and iscurrently working toward an industrial PhDdegree. He is also an RF design engineer atMaxon Cellular Systems.

Michael B. Jenner received his MSEE degreefrom Aalborg University in 1995 and iscurrently working as a research engineer at thesame university.

Torben Larsen received his Dr Techn degreein 1999. He is currently general manager ofRISC Group.

TECHNICAL FEATURE