RF Transceiver Architectures for W-CDMA

Systems Like UMTS:

State of the Art and Future Trends
R. Weigel1;2 , L. Maurer1 , D. Pimingsdorfer2 , A. Springer1
1
Institute for Communications and Information Engineering, University of Linz, Austria
2
DICE-Danube Integrated Circuit Engineering, Linz, Austria
{weigel, maurer, pimingsdorfer, springer}@mechatronik.uni-linz.ac.at
Invited Paper

Abstract— The standardization phase for wideband CDMA) systems follows. The last section gives a per-
CDMA systems like UMTS is running towards its final- spective of future trends in transceiver front-end design
ization. As is typical for mobile communication systems for 3G systems.
standardizing, sufficient RF performance has been as-
sumed and most efforts have been put to baseband issues. In 1985 the ITU (International Telecommunications
This is especially true for the pocket phone transceivers Union) started work on 3G systems unter the acronym
the RF part of which is, although its baseband part is FPLMTS (Future Public Land Mobile Telephone Sys-
much more complex in terms of number of devices, still tem) which was later renamed to IMT-2000 (Interna-
the bottleneck of the entire system. In the RF concept tional Mobile Telecommunications) [1].
engineering of today’s commercial products with their
short time-to-market requirements, a prediction of the The key factors and main objectives for 3G systems
needed RF performance by using RF system simulation include worldwide coverage and roaming incorporating
is meanwhile indispensable. This is in particular the case a satellite component, capacity and capability to serve
with third generation (3G) wireless systems which, from more than 50% of the population [2], multimedia ser-
the RF design point of view, are quite different from 2G vice capability, high-speed access, low-cost operation,
TDMA/FDMA systems due to the fact that the users are
and integration of residential, office, and cellular ser-
now separated in the power domain (using codes) rather
than being separated in the time and/or frequency do- vices into a single system based on one piece of user
main. The present work gives an insight how to derive equipment. Further issues are:
the transceiver requirements of 3G mobiles in terms rec- Æ Packet access. This item is closely linked to the
ognizable by RF designers. above. Most of the traffic in 3G networks will originate
Keywords— Transceivers, CDMA, Radio Communica- from data communications. Therefore, packet switched
tion, Spread Sprectrum Communications. communication must be provided in addition to a circuit
switched mode to ensure efficient resource usage. This
feature is or will be introduced already in 2.5G systems
I. I NTRODUCTION
like GPRS [3].
This work gives an introduction to transceiver de- Æ Evolution. The transition from 2G to 3G will be an
sign for third generation (3G) wireless communications evolutionary path. In the beginning 3G systems and
systems. We start with a review of the transition from services must coexist with todays 2G and 2.5G sys-
second generation (2G) cellular systems to 3G systems. tems, since no one (neither users nor network opera-
The following section introduces basic terms and equa- tors) would be able or willing to afford a hard transition
tions of the spread spectrum technique. The relation from 2G or 2.5G to 3G.
between specific transceiver characteristics like noise From the above mentioned items the basic demands
figure or linearity and testcases from the third genera- for data throughput over the air interface were identi-
tion partnership project (3GPP) specifications are cov- fied as 144 kbps (preferably 384 kbps) with full cover-
ered in the succeeding section. A review of differ- age and high mobility of the user equipment (UE) and
ent receiver and transmitter architectures and their suit- up to 2 Mbps for low mobility and coverage limited to
ability for wideband-code division multiple access (W- high traffic areas. These bit rates were harmonized to
the ISDN (Integrated Services Digital Network) 2B+D II. E VOLUTION FROM 2G TO 3G S YSTEMS
(144 kbps), H0 (384 kbps), and H12 (1920 kbps) chan-
nels [1]. The general vision of 3G systems, is that As pointed out before, there is already existing de-
they should basically ensure communications from any- mand for data rates higher than the few kbps nowadays
where to anybody at any time. possible with 2G systems. With the exception of the
PDC system all 2G standards have provided add-on fea-
Upon a request from the ITU for radio transmission tures supporting higher data rates to account for this
technology (RTT) proposals, different regional stan- traffic demand. A review of these modes, commonly
dardization bodies submitted their proposals for IMT- referred to as 2.5G systems, can be found in, e.g., [10].
2000 in 1998 [4]. Details of these proposals are avail- If the 2.5G systems are fully deployed they will allow
able at [5]. The vast majority of the submitted propos- for data rates up to 384 kbps. With this the mobile
als were based on W-CDMA or at least contained a W- units will evolve from mobile phones to so-called smart
CDMA component. During the evaluation of the differ- phones including, e.g., PDA’s (Personal Digital Assis-
ent proposals by the ITU it turned out that the vision of tants) to better support data applications like email, In-
a global standard with a single radio interface was not ternet or location based services. Also mobile Web pan-
realizable for 3G systems. This was due to the differ- els are under development. If the maximum data rate of
ent 2G technologies used in the different regions in the 1 Mbps in UMTS is available such demanding applica-
world. It would have been impossible to find one tech- tions like video conferencing could be supported. This
nology as evolutionary path for all existing 2G systems. will introduce again a new generation of mobile termi-
nals capable of dealing with multimedia contents.
IMT-2000 Terrestrial
Radio Interfaces III. S PREAD S PECTRUM BASICS

The basic of spread spectrum (SS) technology is

IMT-DS IMT-MC IMT-TC IMT-SC IMT-FT
Direct Multi Time Single Frequency given by Claude Shannon’s well known formula for the
Spread Carrier Code Carrier Time channel capacity [11]:

 
CDMA TDMA FDMA C = W log2 1 + NS : (1)

Fig. 1. The set of IMT-2000 Terrestrial Radio Interfaces

C is the channel capacity in bps, W the bandwidth in
Hz, N the noise power and S the signal power.
Therefore, a family concept was adopted and agreed
upon at the end of 1999 [6]. The five standards included A spread spectrum system must meet two criteria:
in IMT-2000 are shown in Fig. 1. As IMT-DS (Direct
Spread) the UTRA FDD (UMTS Terrestrial Radio Ac-
Æ The transmitted bandwidth is much greater than the
bandwidth or rate of the information being sent.
cess Frequency Division Duplex) mode was adopted in
Europe and Japan, IMT-TC (Time-Code) is a combina-
Æ The spreading signal must be independent of infor-
mation bearing signal (i.e. FM is not SS!).
tion of the UTRA TDD (Time Division Duplex) (Eu-
rope and Japan) and the TD-SCDMA (China) propos- An important reason for using SS is the linear depen-
als, cdma2000 (USA) is found in IMT-MC (Multi Car- dency of the channel capacity C from the bandwidth W
rier), IMT-SC (Single Carrier) corresponds to UWC- in equation (1), whereas C increases only with the log-
136 (USA), and IMT-FT (Frequency Time) is the Eu- arithm of the signal to noise ration S/N. Furthermore,
ropean DECT proposal. These five standards are now SS techniques have an inherent resistance against in-
further developed in the regional standardization bod- terference and jamming. Figure 2 gives an explana-
ies. For the W-CDMA based technologies (IMT-DS tion of this ability. Suppose a narrowband interferer is
and IMT-TC) the third generation Partnership Project present in your received signal. Due to the high corre-
3GPP was created [7]. A similar group was founded lation between the wanted signal (which was spread in
for the development of the cdma2000 based systems, the transmitter by the same sequence) and the locally
termed 3GPP2 [8]. This activity is running in parallel generated code for despreading in the receiver, the sig-
to 3GPP and is coordinated with 3GPP [9]. nal level increases. At the same time the interference
Fig. 2. An interfering signal is spread by a code sequence resulting in a lower power spectral density. The wanted signal level
is increased due to the high correlation of the spreading code and the signal.

signal is spread to a larger bandwidth and therefore the mation data rate (represented by the bit duration Tbit )
interference power in the receiver bandwidth decreases. to the rate of the spreading code (represented by the
To gain this increase of the wanted signal level, the lo- chip duration Tchip ). This ratio ranges for, e.g., the
cally generated code has to be exactly synchronized to 3GPP system from 4 to 512.
the incoming wanted signal.
SS-systems can be classified in three main groups: SF = TTbit (2)
chip
Æ Direct Sequence SS (DSSS): Spreading is done by a
multiplication of the data carrying signal with a code SG[dB] = 10
log SF (3)
sequence of much larger bandwidth.
Æ Frequency Hopping SS (FHSS): Spreading is accom- Let us denote the signal level before despreading the
plished by periodicly changing the carrier frequency. chip energy to interference ratio ( EIc /dB) and the signal
Æ Chirp SS: Carrier frequency varies continuously level after despreading the bit energy to interference ra-
(usually linear) during a time interval. tio ( EIb /dB). Than EIc /dB, EIb /dB and SG are related by

Most SS-systems are based on Direct Sequence-SS. For Eb = Ec + SG + OF:

that reason we will deal in the following exclusively I I (4)
with DSSS (also the term SS will refer to DSSS).
The factor OF/[dB] describes the degree of orthogo-
The best known advantages of SS systems for cel- nality between wanted user signal and interference sig-
lular system design include the possibility of selective nal. The orthogonality factor (OF) for e.g. Gaussian
addressing (Code Division Multiple Access (CDMA)) noise equals 0 dB. Therefore, in a Gaussian noise envi-
and the ability to eliminate the effect of multipath prop- ronment the wanted user signal level is increased by an
agation by using a RAKE receiver in the mobile station. amount of SG dB. For perfectly orthogonal signals OF
Disadvantages incorporate the relatively complex struc- results to 1 dB. Thus the choice of codes employed for
ture of the RAKE-receiver and the need of an accurate the spreading of the user signals greatly influences the
output power control in order to deal with the near-far overall performance of a CDMA system. The orthogo-
problem [12]. The wideband nature of the signal also nality among the spreading codes should be as large as
leads to the necessity of a wideband modem and wide- possible.
band baseband amplifier stages. Furthermore, a fast and
At this point we emphasize once again that these re-
accurately working automatic gain control (AGC) cir-
sults are only valid for perfect synchronization of the
cuitry is a prerequisite for an efficient handling of the
received signal and the locally generated code for de-
multipath signal [13].
spreading. It can be shown, that a timing error of, e.g.,
SS systems can be described by a few equations one half of the chip time results in an SNR loss of 6 dB.
and terms. Important terms when talking about spread Therefore obtaining initial synchronization and keeping
spectrum are the so-called spreading factor SF and the the code synchronized by a code tracking loop can be
spreading gain SG. SF describes the ratio of the infor- considered as key problems in SS system design.
 IV. T RANSCEIVER R EQUIREMENTS AND the center frequency of the assigned channel. Adjacent
3GPP F RONT-E ND T ESTCASES channel selectivity is the ratio of the receive filter atten-
uation on the assigned channel frequency to the receive
The short explanation of some of the 3GPP testcases filter attenuation on the adjacent channel(s).
below should give the reader an introduction of how RF
key parameters can be derived from specifications given Power Spectral
Density
by 3GPP. The complete set of RF specific testcases for
-52 dBm/3.84 MHz
the 3GPP FDD mode can be found in [14]. Further
comments to these testcases are made in [15].

A. Reference Sensitivity Level Testcase -92,7 dBm/3,84 MHz

The reference sensitivity is the minimum receiver in- 10,3 dB

DPCH_Ec=-103 dBm
put power measured at the antenna port at which the bit
error rate (BER) does not exceed a value of 10 3 . This
testcase determines the tolerable noise figure of the re-
ceiver front end. The cumulative value of the incoming
5 MHz
signal power is -106,7 dBm. The wanted user signal Wanted Adjacent
level before despreading EIc is -117 dBm. The reference Channel Channel
channel is a 30 ksps channel which yields an SF of 128.
Fig. 3. Signal levels for the Adjacent Channel Selectivity
According to equation (3) this result in a value for SG testcase.
of approximately 21 dB. Let us assume that the required
E
bit energy to interference ratio bI;req is 5 dB, the inser- The ACS shall be better than 33 dB. Simultaneously,
tion loss (IL) for the baseband implementation is 2 dB the bit error rate shall not exceed 10 3 for the follow-
and the coding gain (CG) is 4 dB. Then the acceptable ing test parameters (see Fig. 3). Received power spec-
interference signal level after despreading (PN ) result tral density of the wanted signal at the terminal antenna
in: connector is -92,7 dBm/3.84 MHz. The wanted user
signal level before despreading EIc is -103 dBm. The
PN = EIc + SG + CG EbIreq IL: ;
(5)
symbol rate of the physical channel is 30 ksps. This re-
sults in an SF of 128 which yields a processing gain
Inserting values in equation (5) results in: of approximately 21 dB. The power spectral density of
the band limited white noise 5 MHz away from the
PN =-117 dBm+21 dB+4 dB-5 dB-2 dB=-99 dBm.
wanted channel PI is -52 dBm/3.84 MHz.
This leaves a margin for the front end noise figure
We assume again that the required bit energy to inter-
(NF) of E
ference ratio b;Ireq is 5 dB, the insertion loss (IL) for the
baseband implementation is 2 dB and the coding gain
NF = PN 10 log(kTB=1 mW): (6) (CG) is 4 dB. This leads us to an acceptable interfer-
ence level PA of
Inserting values for the Boltzmann constant k, tem-
perature T (300 K) and bandwidth B (3.84 MHz), equa-
tion (6) results in a tolerable noise figure NF of PA = EIc + SG + CG EbIreq IL; ;
(8)

NF = 99 dBm + 108 dBm = 9 dB: (7)

resulting in a value of

B. Adjacent Channel Selectivity Testcase

PA = 103 dBm + 25 dB 7 dB = 85 dBm: (9)
Adjacent channel selectivity (ACS) is a measure of
a receiver’s ability to receive a W-CDMA signal at its If the adjacent channel interference signal is treated as
assigned channel frequency in the presence of an ad- Gaussian noise like interference, the required adjacent
jacent channel signal at a given frequency offset from channel selectivity (ACS) can be derived:
 Image Reject
Filter
Interferer

ACS = PI PA = 52 dBm + 85 dBm = 33 dB: Wanted Band Image

11
00
00
11 11
00
00
11
00
11
00
11 00
11
00
11
Channel Select
00
11
00
11 00
11
00
11 Filter
(10) 00
11
00
11 00
11
00
11 11
00
00
11
00
11 00
11
00
11 00
11
00
11 00
11 00
11
00
11
00
11
00
11 00
11
00
11 00
11
The equations used in this section exemplify of how ωwanted ωimage ω
00
11
00
11
00
11
00
11
00
11
the signal levels are influenced by the despreading oper- 00
11
00
11
00
11
ation and by interference sources. However, one should ωIF ωIF 0 ωIF ω

always keep in mind that these results can only serve ωLO ω
as estimates. Further estimations such like the above
mentioned ones can be found in [16]. Fig. 5. Image rejection and channel selection for the hetero-
dyne receiver structure.

V. T RANSCEIVER D ESIGN
Another interesting situation arises with an inter-
A. General Considerations ferer at (!wanted + !LO ) =2. If this interferer expe-
riences second-order distortion and the LO contains
Complexity, cost, power dissipation, and the number a significant second harmonic, then a component at
of external components have been the primary criteria j(!wanted + !LO) 2!LOj = !IF arises. This phe-
in selecting transceiver architectures. As IC technolo- nomenon is called half-IF problem [19].
gies evolve, however, the relative importance of each A major advantage of the heterodyne receiver struc-
of these criteria changes, allowing approaches that once ture is its adaptability to many different receiver re-
seemed impractical to return as plausible solutions [17]. quirements. That is why it has been the dominant
choice in RF systems for many decades. However, the
B. Receiver Architectures complexity of the structure and the need for a large
number of external components (e.g., the IF filter) make
B.1 Heterodyne Receiver problems if a high level of integration is necessary. This
is also the major drawback if costs are concerned. Fur-
Figure 4 shows the heterodyne receiver structure. thermore, amplification at some high IF can cause high
This architecture first translates the signal band down power consumption.
to some intermediate frequency (IF), which is usually
The IMT-2000 proposal specifies an operation mode
much lower than the initially received frequency band.
using two times or four times the base chiprate resulting
Channel select filtering is usually done at this IF, which
in a bandwidth of 7,68 MHz or 15,36 MHz, respectively
relaxes the requirements of the channel select filter. The
(multiband operation). On the other hand the hand-
choice of the IF is a principal consideration in hetero-
sets should also be able to receive GSM signals with a
dyne receiver design (see Fig. 5).
bandwidth of approximately 200 kHz (multimode oper-
As the first mixer downconverts frequency bands ation). Due to the fixed receive bandwidth of the hetero-
symmetrically located above and below the local oscil- dyne receiver structure caused by the external IF-filter,
lator (LO) to the same center frequency, an image reject these multimode and multiband capability can only be
filter in front of the mixer is needed. As depicted in the implemented by using a separate IF section for each
left part of Fig. 5, the filter is designed to have a rela- mode. This would result in high costs and a complex
tively small loss in the desired band and a large atten- receiver structure.
uation in the image band, two requirements that can be
simultaneously met if 2!IF is sufficiently large. Thus, B.2 Homodyne Receiver
a large IF relaxes the requirements for the image rejec-
tion filter, which is placed in front of the mixer (see Fig. The homodyne receiver structure (also called zero-
4). On the other hand it complicates the design of the IF or direct-conversion architecture) entails vastly dif-
channel selection filter (right part of Fig. 5), because of ferent issues from the heterodyne topology. Suppose
the higher IF. In today’s cellular systems the channel se- that the IF in a heterodyne receiver is reduced to
lection filtering is normally done with surface acoustic zero. The LO will then translate the center of the de-
wave (SAW) filters [18]. sired channel to 0 Hz, and the channel translated to
 VGA
I A
Band Image Channel D
Select Reject Select
Filter LNA Filter Filter
0 VCO
90

VCO Q A
D

Fig. 4. Heterodyne receiver structure.

Channel
Select
Filter VGA
I A
D
Preselect
filter LNA
0 VCO
90

Q A
D

Fig. 6. Homodyne receiver structure.

the negative frequency half-axis becomes the image to as serious in a heterodyne receiver. Since in a homo-
the other half of the same channel translated to the dyne topology the downconverted band extends to zero
positive frequency half-axis. The downconverted sig- frequency, offset voltages can corrupt the signal and,
nal must be reconstituted by quadrature downconver- more importantly, saturate the following stages. There
sion (or some other phasing method), otherwise the are three main possibilities how DC-offsets are gener-
negative-frequency half-channel will fold over and su- ated. First, the isolation between the LO port and the
perpose on to the positive-frequency half-channel [20]. inputs of the mixer and the LNA is not infinite. There-
The simplicity of this structure offers two important fore, a finite amount of feedthrough exists from the LO
advantages over a heterodyne counterpart. First the port to the mixer or the LNA input. This “LO leakage”
problem of image is circumvented because !IF = 0. As arises from capacitive and substrate coupling and, if the
a result, no image filter is required. This may also sim- LO signal is provided externally, bond wire couplings.
plify the LNA design, because there is no need for the This leakage signal is now mixed with the LO signal,
LNA to drive a 50
load, which is normally necessary thus producing a DC component at the mixer output.
when dealing with image rejection filters. Second, the This phenomenon is called “self-mixing”. A similar ef-
IF SAW filter and subsequent downconversion stages fect occurs if a large interferer leaks from the LNA or
are replaced with low-pass filters and baseband ampli- mixer input to the LO port and is multiplied by itself. A
fiers that are amenable to monolithic integration. The time varying DC offset is generated if the LO leaks to
possibility of changing the bandwidth of the integrated the antenna and is radiated and subsequently reflected
low-pass filters (and thus changing the receiver band- from moving objects back to the receiver.
width) is a major advantage if multimode and multi- Large amplitude modulated signals that are con-
band applications are concerned. verted to the baseband section via second order distor-
On the other hand the zero-IF receiver topology en- tion of the IQ-mixers also lead to time varying DC off-
tails a number of issues that do not exist or are not set. The spectral shape of this signal contains a signif-
icant component at DC accounting for approximately this architecture has not been used for terminal applica-
50% of the energy. The rest of the spurious signal ex- tions. Nevertheless, it is utilized in base stations where
tends to two times of the signal bandwidth before down- man channels must be received and processed simulta-
converted by the second order nonlinearity of the mix- neously.
ers. The cause for the large signal content at DC is that
every spectral component of the incident interferer is C. Transmitter Architectures
coherently downconverted with itself to DC. In order
to prevent this kind of DC offset, a large second order C.1 Direct Conversion Transmitter
intercept point (IP2) of the IQ-mixer is necessary.
It the transmitted carrier frequency is equal to the lo-
3GPP compliant receivers need approximately 80 dB cal oscillator frequency, the architecture is called “di-
gain. Most of this gain is contributed by the baseband rect conversion”. In this case, modulation and upcon-
amplifiers. That means that even small DC offsets (in version occur in the same circuit. The architecture in
the range of several mV) at the mixer outputs may lead Fig. 7 suffers from an important drawback. Through
to DC levels sufficient to saturate the analog to digital a mechanism called “injection pulling” or “injection
converters (ADC). locking” the transmit LO spectrum is corrupted by the
In time-division multiple access (TDMA) systems power amplifier (PA). The problem worsens if the PA is
idle time intervals can be used to carry out offset can- turned on and off periodically, as it is the case for the
cellation. This would be a practical solution for the 3GPP-TDD mode.
3GPP-TDD mode. It can not be used for offset can- Problems also arise if the system has to fulfil tight
cellation in the FDD mode, because of the continuous requirements on output power range, which is usually
signal reception. Here, the natural solution for DC off- necessary in W-CDMA systems. Most of the gain has
set cancellation is high-pass filtering. Since the signal to be done in the baseband section, leading to high
band extends from DC to approximately 2 MHz, a high- linearity requirements for the baseband filters and the
pass filter with a cut-off frequency of several kHz re- modulator. Furthermore, the LO lies always in the
sults in an acceptable degradation of the system perfor- transmit band, which causes high requirements on the
mance [21]. This approach is only possible because of LO-RF isolation. I/Q phase mismatches are also an is-
the wideband nature of the signal. A system level DC sue when using direct upconversion. Even a low error in
offset compensation approach is described in [22]. the phase shifting network may lead to a severe degra-
I/Q mismatches are another critical issue for the zero- dation of the error vector magnitude (EVM).
IF receiver topology. Fortunately, pilot symbol assisted
channel estimation is done in W-CDMA systems. Irre- C.2 Heterodyne Transmitter
spective of the pilot symbols used (either the time mul-
The second possibility of signal upconversion, which
tiplexed pilot symbols or the common pilot signal), this
circumvents the problem of LO pulling in transmitters,
estimation leads also to a correction of the I/Q phase
is to upconvert the baseband signal in two steps so that
and amplitude mismatch.
the PA output spectrum is far from the frequency of the
VCO’s. An advantage of two-step upconversion over
B.3 Digital-IF Receivers
the direct conversion approach is that since quadrature
In the heterodyne receiver architecture of Fig. 4 the modulation is performed at lower frequencies, I and Q
matching is superior. On the other hand, an IF filter (in
second downconversion and subsequent filtering can be
most cellular applications again a SAW filter) is needed
done digitally. The principal issue in this approach is
which can rise costs considerably.
the performance required from the ADC. To limit the
requirement on the ADC, a sufficiently low IF has to If high integration is an important feature, then both
be chosen, which makes it impossible to employ band- heterodyne transmitters and receivers can cause prob-
pass filtering to suppress the image frequency. Thus, lems. Trying to find intermediate frequencies for the
an image suppression mixer has to be used. The im- transmit and receive section, that do not lead to spuri-
age suppression feasible in today’s systems is limited ous frequencies falling, e.g., in the receive band, may
to a range of 30-55 dB. Due to the high demands on prove to be impossible. This is especially true, if single
the ADC and the image suppression mixer performance chip transceivers are concerned.
 I D/A AGC
RF SAW
LO

90°
+ PA

Q D/A AGC

Fig. 7. Direct upconversion transmitter structure.

I D/A
IF SAW RF SAW
LO IF

90°
+ AGC AGC PA

Q D/A LO RF

Fig. 8. Heterodyne transmitter.

VI. C URRENT D EVELOPMENTS AND F UTURE on-chip voltage regulator for the VCO/buffer. A fixed
T RENDS PLL with reference divider, RF prescaler, lock detect
circuitry and three external elements for 3rd -order loop
filter complete the on-chip synthesizer. The W-CDMA
The first operable UMTS IF transceiver front end was
IF transmitter includes a 5th order active Butterworth
published in Reference [23]. This paper describes a
baseband pre-filter, a quadrature modulator, a variable
fully integrated Si-bipolar IF receiver and transmitter
gain amplifier with >60 dB gain range at a fixed IF fre-
with on-chip synthesizer for use in third-generation W-
quency of 285 MHz. The fully integrated VCO operates
CDMA mobiles. Both devices in a small outline lead-
at a frequency of 1520 MHz.
less package incorporate an on-chip IF synthesizer with
on-chip VCO tuning and tank as well as 6th /5th order Most of the published work on receiver design is
baseband filters and comply with ARIB W-CDMA and based on the direct conversion topology. It seems that
UMTS standards. The IF-chips are fabricated with In- especially the need for high integration restricts the
fineon’s high frequency 0.4 m/25 GHz silicon bipo- receiver architecture to the zero-IF structure. Exam-
lar process. IF receiver and IF transmitter die size is ples can be found in [24][25][26][27][28]. All these
2.33x2.9 mm2 . The devices are designed for low exter- receivers are designed using standard BiCMOS pro-
nal component count and, together with the small pack- cesses. An interesting option is the use of Si/Ge bipo-
age size, minimize the required board area of a com- lar technology for the receiver front end [29][30]. This
plete IF transceiver. The chips operate at 2.7-3.3 V sup- receiver incorporates an I/Q down conversion mixer,
ply, an ambient temperature range of -30 to +85, and broad-band I/Q-generation, fully- integrated VCO,
incorporate several power-down modes for efficient use dual-modulus prescaler, low-noise baseband buffer and
in W-CDMA mobile stations. The W-CDMA IF re- a blocking filter. Integrated in a 75 GHz ft BiCMOS
ceiver includes two complete IF paths for antenna di- technology with 35 m CMOS it draws 33 mA from
versity/service channel monitoring and a common LO a 2.7 V supply. Remarkable are the extremely low
generation and distribution. Each path features a vari- LO leakage of -95 dBm together with a high IIP2 of
able gain amplifier with >95 dB gain range at an IF 55 dBm. This results in very low DC offset values of
frequency of 318 MHz, a quadrature demodulator and less than 20 mV at the baseband output of the IC.
a 5th -order Chebyshev filter and 1st -order all-pass for
the differential I/Q outputs. The IF synthesizer includes With the continuing development of complementary-
a completely integrated on-chip VCO with integrated metal-oxide-semiconductor (CMOS) components and
transformer and varactor diodes, tuning circuitry and processing techniques, CMOS technology is also ex-
pected to become suitable for RF applications within [10] S. Nanda, K. Balachandran, and S. Kumar, “Adaptation Tech-
a few years. Advances with respect to improved de- niques in Wireless Packet Data Services,” IEEE Communica-
tions Magazine, vol. 38, no. 1, pp. 54–64, 2000.
vices, circuit topologies, and system-level architec-
[11] C. E. Shannon, “Communication in the Presence of Noise,”
ture make CMOS a strong contender for implementing Proceedings of the IRE, vol. 37, no. 1, pp. 10–21, 1949.
transceivers for the IMT-2000 global wireless system [12] M. K. Simon, J. K. Omura, R. A. Scholtz, and B. K. Levitt,
[31]. RF CMOS seems to be a promising approach, Spread Spectrum Communications Handbook. McGraw-Hill,
even though the performance of some key components Inc, 1985.
[13] R. C. Dixon, Radio Receiver Design. Marcel Dekker, Inc.,
like PAs, VCOs and filters is not sufficient yet. Consid-
New York, 1998.
ering the rapid development of RF CMOS technology, [14] Technical Specification Group Radio Access Networks, “UE
however, some of these building blocks may become Radio Transmission and Reception (FDD),” Tech. Rep. TS
available within a few years. An interesting project 25.101 V3.5.0, 3rd Generation Partnership Project, Dec.
aimed at RF-CMOS design for W-CDMA transceivers 2000.
[15] Technical Specification Group Terminals, “Radio transmis-
is described in Reference [32]. sion and reception (FDD),” Tech. Rep. TS 34.121 V3.3.0, 3rd
Generation Partnership Project, Dec. 2000.
VII. C ONCLUSION [16] O. K. Jennsen, T. E. Kolding, C. R. Iversen, S. Laursen, R. V.
Reynisson, J. H. Mikkelsen, E. Pedersen, M. B. Jenner, and
T. Larsen, “RF Receiver Requirements for 3G W–CDMA Mo-
A review of cellular phone transceiver concepts for bile Equipment,” Microwave Journal, pp. 22–46, Feb. 2000.
IMT 2000 system use was given. It was shown how [17] B. Razavi, RF Microelectronics. Prentice Hall communica-
basic spread spectrum equations lead to an estimation tions engineering and emerging technologies series, Prentice
Hall, Inc., 1998.
of the transceiver system requirements. Examples ex-
[18] L. Maurer, W. Schelmbauer, H. Pretl, Z. Boos, R. Weigel, and
hibit the influence of 3GPP testcases on key parameters A. Springer, “Impact of IF-SAW Filtering on the Performance
for the transceiver design. Furthermore, a review of the of a W–CDMA Receiver,” in Ultrasonics 2000 Symposium,
most popular receiver and transmitter architectures was Oct. 2000.
given. Their suitability for the W-CDMA system was [19] R. C. Sagers, “Intercept Point and Undesired Responses,”
IEEE Transactions on Vehicular Technology, vol. 32, pp. 121–
evaluated and possible problems were addressed. The
133, Feb. 1983.
last section reviews the state-of-the art and discusses [20] A. A. Abidi, “Direct–Conversion Radio Transceivers for Dig-
future trends of W-CDMA transceiver design. Possi- ital Communications,” IEEE Journal of Solid–State Circuits,
ble advances with respect to improved devices, circuit vol. 30, pp. 1399–1410, Dec. 1995.
topologies, and system-level architecture can make RF [21] J. H. Mikkelsen, T. E. Kolding, T. Larsen, T. Klin-
genbrunn, K. I. Pedersen, and P. Mogensen, “Feasibility
CMOS based transceivers a promising possibility. Study of DC Offsett Filtering for UTRA–FDD/WCDMA
Direct–Conversion Receiver,” in IEEE NORCHIP Confer-
ence, pp. 34–39, Nov. 5th 1999.
R EFERENCES
[22] H. Tsurumi, M. Soeya, H. Yoshida, T. Yamaji, H. Tanimoto,
[1] T. Ojanperä and R. Prasad, Wideband CDMA for Third Gener- and Y. Suzuki, “ System–Level Compensation Approach to
ation Mobile Communications. The Artech House Universal Overcome Signal Saturation, DC Offset, and 2nd–Order Non-
Personal Communications Series, Artech House, 1998. linear Distortion in Linear Direct Conversion Receiver,” IE-
[2] P. Tagghol, B. G. Evans, E. Buracchini, R. D. Gau- ICE Trans. Electron., vol. E82-C, pp. 708–716, May 1999.
denzi, G. Gallinaro, J. H. Lee, and C. G. Kang, “Satellite [23] W. Thomann, J. Fenk, R. Hagelauer, and R. Weigel, “Fully In-
UMTS/IMT–2000 W–CDMA Air Interfaces ,” IEEE Commu- tegrated W–CDMA IF Receiver and Transmitter including IF
nications Magazine, vol. 37, pp. 116–126, Sept. 1999. Synthesizer and on–chip VCO for UMTS Mobiles,” in IEEE
[3] J. Cai and D. J. Goodman, “General Packet Radio Service in Bipolar/BICMOS Technology Meeting, 2000.
GSM,” IEEE Communications Magazine, vol. 35, pp. 122– [24] K. Itoh, T. Katsura, H. Nagano, T. Yamaguchi, and
131, Oct. 1997. Y. Hamade, “2 GHz Band Even Order Harmonic Type Di-
[4] P. Chaudhury, W. Mohr, and S. Onoe, “The 3GPP Proposal rect Conversion Receiver with ABB–IC for W–CDMA Mo-
for IMT–2000,” IEEE Communications Magazine, vol. 37, bile Terminal,” in Microwave Symposium Digest, pp. 1957–
pp. 72–81, Dec. 1999. 1960, 2000.
[5] http://www.itu.int/imt/2_rad_devt/proposals/index-html. [25] A. Pärssinen, J. Jussila, J. Ryynänen, L. Sumanen, and
[6] ITU/99-22, “IMT–2000 Radio Interface Specifications Ap- K. Halonen, “A Wide–Band Direct Conversion Receiver for
proved in ITU Meeting in Helsinki.” ITU Press Release, Nov. WCDMA Applications,” in 1999 IEEE International Solid-
5th 1999. State Circuits Conference, pp. 220–221, 1999.
[7] http://www.3gpp.org. [26] A. Pärssinen, J. Jussila, J. Ryynänen, L. Sumanen, and K. A. I.
[8] http://www.3gpp2.org. Halonen, “A 2–GHz Wide–Band Direct Conversion Receiver
[9] H. Holma and A. Toskala, WCDMA for UMTS. John Wiley & for WCDMA Applications,” IEEE Journal of Solid–State Cir-
Sons, 2000. cuits, vol. 34, pp. 1893–1903, Dec. 1999.
[27] B. Sam, “Direct Conversion Receiver for Wide-band CDMA,”
in Spring Wireless Symposium, 2000.
[28] M. Goldfarb, W. Palmer, T. Murphy, R. Clarke, B. Gilbert,
K. Itoh, T. Katsura, R. Hayashi, and H. Nagano, “Analog
Baseband IC for Use in Direct Conversion W–CDMA Re-
ceivers,” in RFIC Symposium, pp. 79–82, 2000.
[29] H. Pretl, W. Schelmbauer, B. Adler, L. Maurer, J. Fenk, and
R. Weigel, “A SiGe–Bipolar Down–Conversion Mixer for a
UMTS Zero–IF Receiver,” in IEEE Bipolar/BICMOS Tech-
nology Meeting, pp. 40–43, 2000.
[30] H. Pretl, W. Schelmbauer, L. Maurer, H. Westermayr,
R. Weigel, B. U. Klepser, B. Adler, and J. Fenk, “A W–
CDMA Zero–IF Front–End for UMTS in a 75 GHz SiGe BiC-
MOS Technology.” to be published.
[31] J. H. Mikkelsen, T. E. Kolding, and T. Larsen, “RF CMOS
Circuits Target IMT–2000 Applications,” Microwaves & RF,
pp. 99–107, 1998.
[32] http://www.iis.ee.ethz.ch/nwp/lemon/lemon.html.