Tijdschrift Ned. Elektron.- & Radiogenoot.

(NERG), 1
Vol. 70, No. 4, pp. 112-123, Dec. 2005

Multiple Carriers in Wireless Communications –

Curse or Blessing?
Tim C.W. Schenk, Peter F.M. Smulders and Erik R. Fledderus

Abstract— The use of multi-carrier techniques is a natural on the system performance is generally low, when the largest
choice when regarding wireless systems with high bandwidths channel delay is small compared to the symbol time of the
and for which the application environment exhibits severe multi- system. Or when, similarly, the system bandwidth is small
path propagation. These techniques provide a way to cope with
and benefit from the time-dispersive channel. This, however, compared to the coherence bandwidth. For those cases the
comes at the cost of a higher sensitivity to imperfections in influence of the channel can generally be easily removed using
the analogue radio frequency front-end. This paper illustrates an equalizer with a few taps or a rake receiver with several
the sensitivity of the most frequently used multi-carrier tech- fingers.
nique, i.e. orthogonal frequency division multiplexing (OFDM),
to three of the main impairments: phase noise, IQ imbalance
and nonlinearities. Furthermore, it is shown that the use of
digital signal processing can largely compensate the effects of
these non-idealities, overcoming the disadvantages of the use of
multi-carrier techniques.
Index Terms— Wideband communication, multipath propaga-
tion, physical layer, orthogonal frequency division multiplexing
(OFDM), radio front-end impairments, digital compensation.

I. I NTRODUCTION
(a) User scenario (b) Multipath Channel
The application of digital modulation is well established to
convey data between the transmitter (TX) and receiver (RX) Fig. 1. Wireless communication in a multipath environment.
of a wireless communication system. It enables the use of
advanced signal processing and coding techniques to improve As the bandwidth, however, increases and systems move
the transmission quality. Where the application area of these towards more time-dispersive environments (e.g. offices), these
kinds of systems was traditionally in point-to-point links (e.g. solutions become highly complex. Hereto, the use of multiple
in satellite communications and microwave radio links), it carriers was proposed. In these techniques the whole system
has moved over the last decades towards terrestrial wireless, bandwidth is subdivided into several parallel narrow subbands.
mobile and indoor networks. In parallel, the ever increasing When this is implemented in the analogue domain it requires
demand for higher speeds in these kinds of networks, has multiple carriers for frequency conversion and steep bandpass
caused a move from narrowband towards wideband systems. filters to separate the non-overlapping subchannels. Therefore,
Wireless networks are currently deployed in urban areas and efficient implementations in the digital domain were proposed.
in indoor environments, like offices and homes. An example The most applied version of these techniques is based on
for the former are GPRS and UMTS and for the latter wireless the discrete Fourier transform (DFT) and named orthogonal
LAN (Wi-Fi) and Bluetooth. In these environments the trans- frequency division multiplexing (OFDM), the basics of which
mitted signal experiences multipath propagation before reach- will be treated in Section II.
ing the RX. This means that multiple copies of the transmitted Although OFDM exhibits a high spectral efficiency and
signal arrive with different delays and attenuations at the RX, the ability to use the multipath channel to its advantage,
see Fig. 1. The effect is that the response of the wireless it has several disadvantages when compared with traditional
channel is frequency selective and that delayed versions of the single carrier systems. These disadvantages lie mainly in
transmitted symbols leak into neighbouring symbols, causing the constraints it puts on the quality of the analogue radio
inter-symbol interference (ISI). The influence of the multipath frequency (RF) front-end of both TX and RX. The influence of
the most important imperfections on the system performance
Tim Schenk is with the Radiocommunication Group, Department of Elec-
trical Engineering, Eindhoven University of Technology, P.O. Box 513, 5600 are treated in Section III-A, Section IV-A and Section V-A.
MB Eindhoven, The Netherlands (e-mail: [email protected]). Since stringent specifications for the front-end of the re-
Peter Smulders is with the Radiocommunication Group, Department of garded wireless system are required, the analogue part is the
Electrical Engineering, Eindhoven University of Technology, P.O. Box 513,
5600 MB Eindhoven, The Netherlands (e-mail: [email protected]). most expensive part of the system. Furthermore, Moore’s law
Erik Fledderus is with TNO Informatie- en Communicatietechnologie, will influence the digital part in terms of size and price, but has
P.O. Box 5050, 2600 GB Delft and with the Radiocommunication Group, little impact on the analogue part, i.e. the RF front-end; there-
Department of Electrical Engineering, Eindhoven University of Technol-
ogy, P.O. Box 513, 5600 MB Eindhoven, The Netherlands and (e-mail: fore, this part will dominate over time the performance and
[email protected]). price of the radio system. Therefore, this paper also addresses
 2

digital signal processing based algorithms in the baseband part interval (GI), which basically prefixes a copy of the last Ng
of the system, which are designed to suppress the influence samples to the start of the OFDM symbol. If Ng is chosen
of the analogue impairments. Examples hereof are presented sufficiently large compared to the channel length, the ISI is
in Section III-B, Section IV-B and Section V-B. A design contained in the GI of the symbol. Since it is redundant
incorporating these digital compensation techniques allows for information it can be disregarded at the RX, removing the
higher impairment levels, and thus opens the door for cheaper influence of ISI.
and more optimized implementation of the RF front-end, e.g.,
the use of RF CMOS or homodyne transmitters/receivers. This
of course at the cost of higher complexity in the digital part.

II. O RTHOGONAL F REQUENCY D IVISION M ULTIPLEXING

(OFDM)
For a comprehensive description of OFDM and its applica-
tion in different wireless systems, the reader is referred to [1],
[2] and [3]. Here we will treat the basic concept of OFDM
and illustrate the application of the technique for three wireless
systems.
Fig. 3. OFDM signal in the time-domain, showing the addition of the guard
A. OFDM basics interval (GI).
The concept of using the discrete Fourier transform (DFT)
as part of the digital modulation/demodulation in TX and RX Since the addition of a GI decreases the effective datarate
part of the wireless system to achieve parallel data transmis- of the system, the ratio between the number of carriers Nc ,
sion was proposed in the early seventies by Weinstein and which is equal to the symbol length in samples, and the GI
Ebert in their seminal paper [4]. Due to the properties of the length Ng is an important design parameter. It must be chosen
DFT the subchannels are shaped like sin(x)/x. An example in a tradeoff between ISI robustness and effective datarate.
of spectrum of three OFDM subcarriers is shown in Fig. 2, The basic OFDM processing in both TX and RX is sum-
which shows that spectra are partly overlapping, significantly marized in the block diagram of Fig. 4. First the baseband
increasing the spectral efficiency as compared to conventional processing to the input bitstream is applied, e.g., interleaving,
non-overlapping multi-carrier systems. It is clear, however, channel coding, puncturing and mapping to complex symbols,
from Fig. 2 that the separation of the different carriers can here modelled by the block GTX . The complex signal is
not be carried out by bandpass filtering. Therefore, baseband then, after being demultiplexed (DEMUX), fed to the inverse
processing is applied which exploits the orthogonality property DFT, which converts the signal to the time domain. Many
of the subcarriers. This property is apparent from Fig. 2, where efficient implementations of the (I)DFT exist, which overcome
at the the maximum of one subcarrier all other carriers have the previously regarded insurmountable complexity of this
a zero amplitude. operation. Subsequently, a GI is added to the signal. Then the
signal is converted to the analogue domain and up converted
to RF fRF . Then the signal is transmitted through the wireless
(multipath) channel.

Fig. 4. OFDM system block diagram, showing a generic model for the
Fig. 2. Spectra of three subcarrier forming an OFDM signal.
transmitter and receiver.

The same OFDM signal is depicted in the time-domain in The received signal is, subsequently, downconverted to base-
Fig. 3. From this figure it can be concluded that the symbol band by the RF RX front-end. The output of the analogue-to-
length contains {1, 2, 3} periods of the signal on the different digital converter is then passed to the RX baseband processing.
carriers, respectively. Here the number of periods depends on This processing removes the GI, which annuls the influence of
the position of the subcarrier within the OFDM spectrum. the ISI. The DFT processing then separates the signals on the
To increase the robustness of the OFDM system against ISI, different carriers. The multiplexed (MUX) data stream is then
caused by multipath propagation, the addition of a cyclic processed in the RX processing, here depicted as block GRX ,
extension of the symbols was proposed in [4]. Hereto the where, e.g., channel equalization, decoding and deinterleaving
symbol length is prolonged for Ng samples with a guard are performed.
 3

B. Turning multipath into an advantage all subchannels carry symbols from the same modulation
As mentioned above, the influence of the ISI caused by size. Here the subchannels with the lowest SNR determine
multipath propagation is removed at the receiver, when the GI the probability of error and thus the modulation format. In
is chosen long enough. The other effect of the multipath, the contrast, a system based on adaptive modulation can maximize
frequency selective fading, remains. Now, however, since the the total throughput for a certain probability of error.
bandwidth is subdivided into parallel carriers, the subcarrier
bandwidth is smaller than the channel coherence bandwidth.
That is why the channel can be regarded as frequency flat for
a certain carrier. This enables to use of a single tap equalizer
per subcarrier to compensate for the channel influence.
If we assume the noise experienced in the system is white,
the subcarriers having a lower channel transfer experience a
lower signal-to-noise ratio (SNR). The detection of the trans-
mitted symbols on such subcarriers yields a higher probability
of error. This is schematically depicted in Fig. 5. Here the
white carriers have a low probability of correct detection of
the transmitted symbols.

Fig. 6. Adaptive modulation.

The use of adaptive modulation is illustrated in Fig. 6. Here

two subcarriers with a very low channel transfer are assigned
no bits, nine subcarriers with moderate channel transfer carry
1 bit/subcarrier (e.g., by the use of BPSK modulation) and
the eleven subcarriers with the highest channel transfer are
assigned 2 bits/subcarrier (e.g., by the use of QPSK modu-
lation). In the case of conventional OFDM all carrier would
have been assigned 1 bit/subcarrier, clearly showing the gain
Fig. 5. Subcarriers of an OFDM system experiencing a frequency selective of adaptive modulation.
channel. 3) Orthogonal frequency division multiple access: The
multi-carrier technique OFDM can also be used as a multi-
1) Channel coding: The performance of OFDM in fre-
ple access technique: Orthogonal frequency division multiple
quency selective fading can be largely improved by the use
access (OFDMA). In contrast to conventional OFDM, where
of channel coding, yielding coded OFDM (COFDM). Here
all carriers are assigned to one user, in OFDMA the carriers
the bits are encoded in codewords, which are spread over the
are subdivided between the different users. The scenario of
different subcarriers using an interleaver. Since the codewords
such a scheme is depicted in Fig. 7, where multiple users are
are spread over the different carriers, the probability that a
receiving simultaneously.
whole codeword is received on a channel with low transfer is
lower and thus, the resulting probability of error detection is
also lower. It can be shown that the performance of a COFDM
system will improve with increasing frequency selectivity, up
to channel lengths where ISI occurs. In this way the system
benefits from the multipath channel.
2) Adaptive modulation: Another way to cope with the
multipath is the use of adaptive modulation. In this approach
an estimate of the channel is available at the TX side of
the wireless link. This can be obtained by either using the Fig. 7. User scenario for a system applying OFDMA.
reciprocity of the channel or by the use of a feedback channel.
This channel estimate, as mentioned above, relates to the Since all the users have their own location, the signals trans-
SNR experienced at the RX. In adaptive modulation, now, mitted/received by the users experience different multipath
subchannels with high channel transfers, or equivalently high channels. Therefore, also the frequency response of channels
SNRs, are assigned symbols of higher order modulation and corresponding to the different users will differ. In OFDMA we
subchannels with low channel transfer are assigned symbols can exploit this to our advantage by assigning those subcarriers
from a lower order modulation or even no symbols. This is to an user, where its channel transfer is high. Clearly this, as
based on the fact that high order multi-level modulations, e.g., for adaptive modulation, requires some knowledge about the
M -QAM and M -PSK modulations with high M , achieve a channel transfer.
high bit rate per subcarriers, but also require a high SNR The use of OFDMA is illustrated in Fig. 8, where the left
for reliable detection. In conventional OFDM, in contrast, block of carriers is assigned to user 1, the middle to user 2 and
 4

2) Ultra Wideband: Recently, the application of OFDM

has been proposed for ultra wideband (UWB) communica-
tions [7] under the IEEE 802.15.3a PAN framework. This
proposal applies a signal bandwidth of 528 MHz, which is
divided into 128 subcarriers of which 100 are used for data
transmission. The system is based on what is called multiband
OFDM, where consecutive symbols are transmitted in different
frequency bands. Initial deployment is foreseen in the 3.1-4.9
GHz band, but extensions towards bands up to 10 GHz are
envisioned for the future. The proposal is based on QPSK
modulation and a variable coding rate resulting into a data
Fig. 8. OFDMA: a multiple access scheme based on OFDM. rate varying from 53.3 Mbps up to 480 Mbps. Since the
allowed transmit powers for these types of systems are low,
the application is foreseen in systems combining low cell radii
the right block to user 3. It is noted that although also channel and high data rates.
1 has its maximum transfer in the right block of subcarriers, it In parallel to the standardisation by the IEEE, a similar
is assigned to user 3, since channel 3 has a low transfer in the proposal was accepted in May 2005 as the wireless USB [8]
left block, where the transfer of channel 1 is still acceptable. specification. Different vendors are now starting to deliver
Although the assignment is applied in blocks of carriers here, products based on this specification.
other patterns might be more optimal, especially when this 3) Wireless MAN: The use of OFDM has also been stan-
multiple-access technique is combined with channel-coding. dardized for application in outdoor networks, for example un-
For instance individual carriers throughout the spectrum can be der the IEEE 802.16 framework. This standardisation effort is
assigned to a user, resulting into an interleaved carrier pattern focussed on wireless metropolitan-area-networks (WMANs).
for the different users. The initial application was for point-to-point links, with a main
The advantage of OFDMA is that multiple users can be focus on providing a cost effective alternative for the wired
receive/transmit simultaneously, and that the channel transfer local-loop. These systems [9] are often collective referred to
is optimized for the combination of users. as WiMax and can operate in the 2-11 GHz band, providing
speeds up to 75 Mb/s. The bandwidth is flexible, varying from
1.5 - 20 MHz. The system provides an OFDM and OFDMA-
C. Wireless Systems using OFDM
mode with 256 and 2048 subcarriers, respectively.
The application of OFDM is currently standardized for use Recently, an extension to this standard was proposed which
in many different kinds of wireless systems, including wireless extends the application of these systems to mobile networks.
personal-area-network (PAN), local-area-network (LAN) and This extension of focussed on frequency bands below 6 GHz
metropolitan-area-network (MAN), digital audio broadcasting and applies a scalable OFDMA design. Again the bandwidth is
and digital video broadcasting. To illustrate the application flexible, but at a bandwidth of 5 MHz a maximum datarate of
of OFDM in these kinds of systems, the system parameters of 15 Mbit/s can be achieved. This design is being standardized
Wireless LAN, Ultra Wideband and Wireless MAN are treated as IEEE 802.16e and a version is currently being deployed in
here. Korea as Wireless Broadband (WiBro).
1) Wireless LAN: The application of OFDM for wireless
local-area-networks (WLANs) was first standardised in 1999 III. P HASE N OISE
as IEEE 802.11a [5], often referred to as Wi-Fi, where it was
As mentioned in the introduction of this paper, OFDM-
applied to the 5 GHz frequency band. A similar system was
based systems are very vulnerable to radio front-end induced
standardized within ETSI under the name of HiperLAN/2.
impairments. The imperfections of the radio frequency (RF)
Both systems specify a system with 20 MHz bandwidth, 64
oscillator are treated here, since these have been identified
subcarriers and a GI-length of 16 samples (800 ns). A subset
as the major performance limiting factors of OFDM systems.
of 48 carriers is used for data transmission, 4 are used for
First Section III-A treats the influence of imperfect oscillators
synchronization purposes and the remaining carriers (at the
on the OFDM signals and then several digital compensation
sides of the band) are used as guard bands to minimize the
techniques for this influence are reviewed in Section III-B.
out-of-band radiation. The system applies BPSK, QPSK, 16-
QAM and 64-QAM modulation, convolutional coding with
rates varying from 1/2 to 3/4, creating data rates ranging from A. Influence of Phase Noise
6 Mbps up to 54 Mbps. Ideally an RF oscillator would exhibit a power-spectral-
To enable higher speeds in the 2.4 GHz ISM band, where up density (PSD), which resembles a delta-function at frequency
to then the IEEE 802.11b standard was deployed, the design fRF . Any practical oscillator, however, experiences distur-
of the IEEE 802.11a OFDM physical layer (PHY) was ported bances due to thermal noise, making the PSD differ from the
to the 2.4 GHz band in 2003. This was standardized in the ideal. Since in general the disturbance of the amplitude of
IEEE 802.11g standard [6], which combines the OFDM PHY the oscillator output is marginal [10], [11], most influence of
with the IEEE 802.11b PHY. the oscillator imperfection is noticeable in random deviation
 5

of the frequency of the oscillator output. These frequency This influence of PN on the received signal in an OFDM
deviations are often modelled as a random excess phase, and system applying QPSK modulation is illustrated in Fig. 10.
therefore referred to as phase noise. Phase noise (PN) will These results are for a system transmitting several packets
more and more appear to be a factor limiting the performance and not experiencing a channel, i.e., Hm (k) = 1 for all k
of OFDM systems, when low-cost implementations or systems and m. In Fig. 10(a) a typical result is shown for a system
with high carrier frequencies are regarded [12], since it is in having an oscillator characterised by a corner frequency ∆f3dB
those cases more challenging to produce an oscillator with which is low compared to the spacing of the subcarriers.
sufficient stability. It can be concluded that the dominant effect of PN is the
The RF oscillator process used for down-conversion of the rotation. When ∆f3dB is increased, whereas the total noise
received signal can be modelled as power remains unchanged, for the situation in Fig. 10(b), the
rotational behaviour is less pronounced, but a higher additive
a(t) = e−j{2πfRF t−θ(t)} , (1) behaviour is visible.
where j denotes the imaginary unit, t is the time variable and 1.5 1.5
θ(t) denotes the PN process. When (1) models a free-running 1 1
oscillator it can be shown that the corresponding PSD is given
0.5 0.5
by the Lorentzian function [11], as depicted schematically in
Fig. 9. The PN process is then fully determined by the 3 dB

Q
0

Q
0

bandwidth ∆f3dB of the PSD. −0.5 −0.5

−1 −1

−1.5 −1.5
10 log10 ( 2π∆f1 ) −1.5 −1 −0.5 0
I
0.5 1 1.5 −1.5 −1 −0.5 0
I
0.5 1 1.5
3dB
(a) Low ∆f3dB (b) High ∆f3dB

Fig. 10. Influence of PN on the reception of QPSK symbols.

−20 dB/decade
PSD (dBc/Hz)

It can be concluded from Fig. 10 that the lower and higher

frequencies in the PN PSD result in rotational and additive
behaviour, respectively. For more indepth information on this
behaviour and the influence of PN on the system performance
in general, the reader is referred to [13–15].

B. Suppression of Phase Noise

fRF fRF + ∆f3dB
f (log scale) Designing an oscillator with high enough stability to over-
Fig. 9. Single side band representation of the PSD of the oscillator process
centred around fRF .
come the effects presented in Section III-A is very challenging,
especially when low-cost solutions are regarded. In contrast,
here we regard the compensation of these effects by the use
If we, subsequently, regard the influence of PN on the
of signal processing in the digital baseband part of the RX.
reception of the OFDM symbols, we find that the time domain
Different approaches have been proposed in the literature
signal in the baseband receiver is multiplied with the phase
to remove this rotational part. The techniques can basically
noise process ejθ(t) . In the frequency domain this can be seen
be divided into two groups, where the first one uses a data-
as a convolution of the OFDM subcarriers, as depicted in
aided approach and the second one uses detected data-symbols
Fig. 2, with the PSD of the oscillator process, as depicted
to estimate αm . For an example of the former the reader
in Fig. 9. Clearly this degrades the orthogonality between the
is referred to [14]. Basically these techniques exploit that
subcarriers.
known symbols sm (k), i.e., pilot symbols, are transmitted
The received signal, behind the DFT-processing in the RX,
on certain subcarriers. Since the rotation is common to all
on the kth carrier of the mth symbol is then given by
carriers, the location of these pilot carriers is not important,
xm (k) = ejαm Hm (k)sm (k) + ξm (k) , (2) but they are generally equally spaced over the whole system
bandwidth, to avoid that all pilots would fall in a channel
where sm (k) and Hm (k) denote the transmitted symbol and fade. The second group of techniques first compensates the
channel response on the kth carrier of the mth symbol, received signal for the estimated channel response and αm−1 ,
respectively. The effects of the PN are modelled in a phase which was estimated in the previous symbol. Now, the average
rotation common to all carriers, i.e., αm and the parameter rotation of this corrected symbol gives an estimate of ∆αm ,
modelling the leakage into carrier kth, i.e. ξm (k). The latter is which provides the estimate of αm = αm−1 + ∆αm .
often referred to as inter-carrier-interference (ICI) and consists Next to the compensation of the rotational part, also sup-
of a weighted sum of the received symbols on all other pression of the PN-caused ICI component ξm (k) is possible.
subcarriers. The phase rotation ejαm is given by the average Recently, some approaches have been proposed for this ICI
value of ejθ(t) during the mth symbol. compensation [16], [17]. These techniques use that the ICI at
 6

a certain carrier is dominated by the neighbouring carriers. These carriers have the same separation from DC. The signal is
Some gain in performance is achieved here, but the overall up converted to RF and transmitted to the frequency selective
effect is limited at low SNRs and in fading channels. channel, where the received RF signal is depicted in Fig. 12(b).
It is clear that carrier −k is more attenuated by the channel
IV. IQ IMBALANCE than carrier k. Subsequently, the RX signal is down converted
to baseband using the homodyne structure of Fig. 11. Since
In this section we discuss the influence of a mismatch
this structure exhibits IQ mismatch, the mirror is not fully
between the I and Q branch, known as IQ imbalance. This mis-
rejected, and mixes down into the regarded baseband channel.
match occurs due to limited accuracy in the implementation
This is illustrated in Fig. 12(c), which shows that carrier k
of the RF front-end and results into a limited image rejection.
experiences a contribution of the signal received on the mirror
Although IQ imbalance can occur in any quadrature receiver,
carrier −k and vice versa.
we here focus on a homodyne receiver [18], as illustrated
The effect on the received baseband signal can be expressed
in Fig. 11. The advantage of these direct conversion type of
as
RXs is that no costly surface acoustic wave (SAW) filter is
∗
necessary, only two low-pass filters (LPFs), which can more xm (k) = K1 Hm (k)sm (k) + K2 Hm (−k)s∗m (−k) , (3)
easily be integrated.
where ∗ denotes complex conjugation and K1 and K2 model
the imbalance. The effect of mirror leakage as shown in
Fig. 12(c) is also apparent from (3). In general K1 is larger
than K2 and in a system with ideal matching K1 = 1 and
K2 = 0.
To clearly show the effect of the IQ imbalance on the
reception of an OFDM signal, a noiseless system applying 16-
QAM modulation is regarded. The system does not experience
a wireless channel, but has a 10% amplitude and 5◦ phase
Fig. 11. Schematic representation of a direct conversion based RX. imbalance between the I and Q branches of the RX. The
received signal is depicted in Fig. 13, which shows that the
First we will regard the influence of IQ imbalance in transmitted 16-QAM points are distorted by an additive rotated
Section IV-A and then review several digital compensation 16-QAM constellation of lower amplitude. This is due to
approaches in Section IV-B. the leakage of the mirror carrier −k, where the rotation and
reduced amplitude are due to the imbalance parameter K2 .
A. Influence of IQ imbalance Furthermore, one may observe a minor rotation and decrease
in amplitude due to the multiplication of the desired signal
The IQ imbalance in the structure of Fig. 11 has several (on carrier k) with imbalance parameter K1 .
sources, but regarding the influence we can distinguish be-
tween two types of IQ imbalance: frequency independent and
1.5
frequency selective IQ imbalance. An example of the first
group is the mismatch which occurs when the phase shift
1
between the signal used for up/down-conversion of the I and
Q signal is not exactly 90 degrees. The frequency selective IQ
0.5
imbalance for instance occurs due to a mismatch between the
low-pass filters (LPFs) in the I and Q branches, see Fig. 11.
Q

0
In a practical RX the frequency independent behaviour will
be dominant. Hereto, we will focus on this behaviour here.
−0.5

−1

−1.5
−1.5 −1 −0.5 0 0.5 1 1.5
I

Fig. 13. Influence of IQ imbalance on the reception of 16-QAM symbols in

(a) TX baseband (b) RX RF (c) RX baseband a noiseless channel-less scenario.

Fig. 12. The influence of the IQ imbalance on the reception of an OFDM

signal.
B. Suppression of IQ imbalance
The influence of IQ imbalance is shown schematically The occurrence of IQ imbalance can be circumvented by
in Fig. 12. The transmitted baseband signal is shown in increasing the accuracy of the implemented mixing structure,
Fig. 12(a), where two carriers (−k and k) are highlighted. which would result into a more expensive solution. The use of
 7

compensation techniques for this RF impairment is, however, the PAPR becomes higher. Figure 15 illustrates the PAPR
promising, since the IQ imbalance variation with time is dependence on Nc for a system applying 64-QAM modulation.
negligible and the influence of IQ imbalance has a very It, hereto, shows the inverse cumulative distribution function
specific structure, as was shown in Section IV-A. for the PAPR of the OFDM-symbols for different number of
Several compensation approaches have been proposed for subcarriers, which indicates the chance that the PAPR is larger
the compensation of IQ imbalance, mainly focussing on imbal- than the value on the x-axis.
ances in the RX. Again data-aided and blind techniques can be
1
distinguished. An effective data-aided technique was proposed N = 64
c
by the authors of [19], which is applicable for systems 0.9 Nc = 128
applying a block of pilot carrier preceding the data carriers. 0.8
Nc = 512
It applies the fact that the estimate of the wireless channel N = 1024
c
will not be smooth, when IQ imbalance occurs in the down 0.7

P{PAPR> x}
conversion. A blind method was proposed by the authors of 0.6
[20], where the imbalance parameters are estimated by using
0.5
the statistical independence of the datasymbols transmitted on
carrier k and −k. Both methods can be applied to significantly 0.4
suppress the RX based IQ imbalance. 0.3
When the system, however, experiences both TX and RX
0.2
IQ imbalance, these techniques will not suffice. Therefore,
the authors recently proposed a data-aided approach for joint 0.1
estimation of TX and RX induced IQ imbalance [21]. This 0
method effectively compensates for the joint TX and RX IQ 2 4 6 8 10 12
imbalance in the digital baseband part of the RX. x [dB]
An example of results achieved with the blind compensation Fig. 15. PAPR of 64-QAM OFDM symbols for different number of
subcarriers Nc .
method for RX IQ imbalance [20] is shown in Fig. 14. Here a
scenario with severe RX imbalance and experiencing additive
RX noise is regarded. It is clear from Fig. 14(a) that detection It can be concluded from the figure that for 1024 subcarriers
of the receiver 16-QAM points before digital compensation 50% of the symbols possesses a PAPR which is higher than
would result in many errors. The results after the compensation 9 dB. This would not be a disadvantage in a fully linear
are shown in Fig. 14(b). It can be concluded that the rotation wireless system. Any practical system, however, has nonlinear
and attenuation of the signal (by K1 ) has been successfully parts like for instance the power amplifier (PA) in the TX and
removed. Furthermore, the scatter plot of the constellation low-noise amplifier (LNA) in the RX. The influence of these
points are much smaller after compensation, showing that also nonlinearities on the signal is discussed in Section V-A and
the mirror leakage is removed. The remaining impairment is suppression approaches are reviewed in Section V-B.
due to the additive RX noise.
1.5 1.5 A. Influence of Non-linearities
1 1 For illustration purposes Fig. 16 schematically depicts the
0.5 0.5 transfer of a nonlinear PA. It shows that the PA delivers linear
amplification for input powers up to a certain level. Beyond
Q

0 0
Q

this level the amplification decreases compared to the linear

−0.5 −0.5
curve. When the input power is even higher clipping occurs,
−1 −1 i.e., the output power is limited to the maximal output power
−1.5 −1.5
of the PA.
−1.5 −1 −0.5 0 0.5 1 1.5 −1.5 −1 −0.5 0 0.5 1 1.5
I I To minimize the influence of nonlinearities and clipping, we
(a) received baseband signal (b) received signal after compensation would like to locate the signal in the linear region. This linear
region is bounded by the point where the transfer deviates
Fig. 14. Effects of IQ imbalance compensation.
1 dB from the linear response, i.e., the 1 dB compression
point. Although Fig. 16 shows the nonlinear transfer of the
amplitude, often referred to as AM-AM transfer, also AM-
V. N ONLINEARITIES PM behaviour occurs, i.e., phase deviations occur at high input
One of the major drawbacks of an OFDM system over a levels. In the remainder of this section we will focus on the
single-carrier system, is that the time-domain OFDM signal former.
exhibits a large peak-to-average-power ratio (PAPR), which, as When transmitted in the nonlinear region of the system,
will be shown in Section V-A, requires a highly linear system. the time-domain signal will be distorted. This will result in a
The PAPR is dependent on the order of the applied modulation two-fold behaviour. First, in-band distortion of the detected
scheme and the number of subcarriers. If the number of signal occurs, since the nonlinearities basically destroy the
subcarriers Nc or the M -QAM modulation order increases orthogonality between the carriers, which decreases the system
 8

signal processing techniques were proposed to overcome this

problem. The approaches can be split into techniques which
try to reduce the PAPR of the signal and techniques that try
to compensate for the nonlinear behaviour.
Several techniques have been proposed of the last few
years to reduce the PAPR of an OFDM signal. The use of
complementary block codes for this purpose was proposed
in [23]. The authors of [24] propose a technique named
selected mapping, where different codes words are generated
representing the same data. The codeword resulting into the
lowest PAPR is then selected and transmitted. The use of linear
combination of partial transmit sequences was proposed in
[25]. In this approach blocks of carrier are multiplied with
different phase shifts. The combination of phase shifts that
Fig. 16. Transfer of a nonlinear power amplifier. results in the lowest PAPR is selected and the resulting symbol
is transmitted. It is noted that the last two techniques require
some extra information to be transmitted, i.e., which codeword
performance. Furthermore, these nonlinearities introduce out-
or combination of phase shifts is used, and reduce the effective
of-band leakage due to spectral regrowth, i.e., the transmission
datarate in that way. By decreasing the PAPR the system can
levels in neighbouring bands are increased. Since the allowed
apply a lower BO and thus operate more efficient.
transmit powers in neighbouring bands are limited by regula-
Linearisation techniques can be applied in both TX or RX
tions, this posses a problem for the overall system design.
of the OFDM system. When it is placed in the TX it is often
An example of the first effect is shown in Fig. 17, where
referred to as digital predistortion. In this technique the TX
a received 16-QAM modulation is shown for a 64 subcarrier
signal is multiplied with a transfer, which compensates for the
OFDM system having a nonlinear PA, where the AM-AM
nonlinear characteristic of the PA. The combination of their
distortion is modelled like in [22]. Again a channel-less and
transfers results into a linear transfer up to the point where
noiseless scenario is assumed. The signal is fed at different
the PA output power is maximum, where the clipping will
levels to the PA. Hereto an back-off (BO) is applied, which is
again occur. This is illustrated in Fig. 18. The linearisation can
the difference between the average input power level and the
be implemented in different ways: one option is to estimate
input power corresponding to the border between the linear
the nonlinear characteristic in an offline calibration mode and
and nonlinear region. The higher the BO is the more linear
another option is the use of an adaptive algorithm. For more
the transfer is that the signal experiences, but also the lower
information on digital predistortion the reader is referred to
the output power becomes.
[26].

(a) BO = 11 dB (b) BO = 5 dB (c) BO = 3 dB

Fig. 17. The effects of a nonlinear PA on the received constellation for

different BO values.

It can be concluded from Fig. 17 that for a BO of 11 dB

almost no influence of the nonlinearities is visible. This can
be explained by the fact that the chance that the PAPR is lager
than 11 dB is smaller than 10−4 . It is thus very unlikely that
one of the peaks falls in the nonlinear region. This is different Fig. 18. Nonlinear and predistortion characteristics.
for smaller values of BO, as is clear from Fig. 17(b) and
Fig. 17(c), where a growth of the scatter points is observable, Although compensation for the nonlinearities generally is
which corresponds to an increased probability of erroneous done at the TX, it can also be carried out at the RX,
detection of the transmitted symbols. enabling joint compensation of the nonlinearites throughout
the transmission chain. The disadvantage is that it does not
B. Suppression of Non-linearities help to decrease the out-of-band radiation, but it does enable
The conventional method to overcome the influence of the correction for clipped signals, decreasing the probability
nonlinearities, by applying a BO, reduces the output power of erroneous detection.
and is, thus, very energy in-efficient. Therefore, different The individual techniques highlighted in this section might
 9

require a considerable complexity in the baseband part of the [9] IEEE Standard for Local and Metropolitan Area Networks Part 16:
wireless system to significantly reduce the performance degra- Air Interface for Fixed Broadband Wireless Access Systems, IEEE Std
802.16-2004, 2004.
dation due to non-linearities. However, when combinations of [10] M. Lax, “Classical noise. V. Noise in self-sustained oscillators,” Phys.
the different techniques are applied, the complexity can be Rev., vol. CAS-160, pp. 290-307, 1967.
minimized. [11] A. Demir, A. Mehrotra, and J. Roychowdhury, “Phase noise in oscil-
lators: a unifying theory and numerical methods for characterization,”
IEEE Trans. on Circuits and Systems, pp. 655–674, May 2000.
VI. C ONCLUSIONS [12] P.F.M. Smulders, “Exploiting the 60 GHz band for local wireless
multimedia access: prospects and future directions,” IEEE Commun.
The applications multiple carriers can effectively be used Magazine, vol. 40, 2002.
to overcome the impact of multipath propagation on the per- [13] T. Pollet, M. van Bladel, and M. Moeneclaey, “BER sensitivity of OFDM
systems to carrier frequency offset and wiener phase noise,” IEEE Trans.
formance of wireless systems applying high bandwidths. The on Commun., vol. 43, no. 2/3/4, pp. 191–193, Feb./Mar./Apr. 1995.
most applied multi-carrier technique is orthogonal frequency [14] A.G. Armada and M. Calvo, “Phase noise and subcarrier spacing
division multiplexing (OFDM). It is shown that OFDM effec- effects on the performance of an OFDM communication system,” IEEE
Commun. Letters, vol. 2, no. 1, pp. 11–13, Jan. 1998.
tively divides the frequency selective channel into frequency [15] R. Hasholzner, C. Drewes, and J.S. Hammerschmidt, “The effects
flat parallel subchannels. The addition of a guard interval to of phase noise on 26 Mb/s OFDMA broadband radio in the local
the OFDM symbols removes the influence of inter-symbol- loop systems,” in Proc. International Zurich Seminar on Broadband
Communications 1998, Feb. 1998, pp. 105–112.
interference. Furthermore, the application of channel coding, [16] R.A. Casas, S.L. Biracree, and A.E. Youtz, “Time domain phase noise
OFDMA and adaptive modulation turn the time dispersive correction for OFDM signals,” IEEE Trans. on Broadcasting, vol. 48,
channel into an advantage, by exploiting the provided fre- no. 3, pp. 230–236, Sept. 2002.
[17] D. Petrovic, W. Rave, and G. Fettweis, “Intercarrier interference due to
quency diversity. This is the reason why the application of Phase Noise in OFDM - estimation and suppression,” in Proc. IEEE
OFDM is proposed for use in many wireless systems over the VTC Fall 2004, LA, Sept. 2004, pp. 1901–1905.
last few years. [18] A.A. Abidi, “Direct-conversion radio transceivers for digital commu-
nications,” IEEE J. Solid-State Circuits, vol. 30, pp. 1399–1410, Dec.
It is also explained that OFDM is sensitive to imperfections 1995.
in the analogue front-end of the transmitter and receiver. It [19] J. Tubbax, B. Come, L. van der Perre, S. Donnay, M. Engels, and
is shown that phase noise, IQ imbalance and nonlinearities C. Desset, “Joint compensation of IQ imbalance and phase noise,” in
Proc. 57th IEEE Vehicular Technology Conference 2003-Spring, April
impose severe performance degradations to an OFDM system. 2003, vol. 3, pp. 1605–1609.
This paper, however, also highlights that the application of [20] M. Windisch and G. Fettweis., “Standard-independent I/Q imbalance
digital signal processing can largely overcome these disad- compensation in OFDM direct-conversion receivers,” in Proc. 9th
International OFDM Workshop, Sept. 2004.
vantages. A design incorporating these digital compensation [21] T.C.W. Schenk, P.F.M. Smulders, and E.R. Fledderus, “Estimation and
techniques allows for higher impairment levels, and thus opens compensation of TX and RX IQ imbalance in OFDM based MIMO
the door for cheaper and more optimized implementation of systems,” Accepted for IEEE Radio and Wireless Symposium (RWS
2006), Jan. 2006.
the RF front-end. [22] C. Rapp, “Effects of HPA-nonlinearities on a 4-DPSK/OFDM-signal for
Overall it can be concluded that multi-carrier techniques a digital sound broadcasting system,” in Proc. of the second European
conference on Satellite Communication, Liege, Belgium, Oct. 1991, pp.
are a natural choice for wideband systems to be applied in 179–184.
multipath environments, especially since its major drawbacks [23] R.D.J. van Nee, “OFDM codes for peak-to-average power reduction
can largely be solved by emerging digital compensation tech- and error correction,” in Proc. of the IEEE Global Telecommunications
Conference, 1996, Nov. 1996, vol. 1, pp. 740–744.
niques. [24] R.W. Bauml, R.F.H. Fischer, and J.B. Huber, “Reducing the peak-to-
average power ratio of multicarrier modulation by selected mapping,”
ACKNOWLEDGEMENTS Electronics Letters, vol. 32, pp. 2056–2057, Oct. 1996.
[25] S.H. Muller and J.B. Huber, “OFDM with reduced peak-to-average
This work was carried out within the B4 Broadband Ra- power ratio by optimum combination of partial transmit sequences,”
Electron. Lett., vol. 33, pp. 368–369, Feb. 1997.
dio@hand project, which is supported in part by the Dutch [26] S.C. Cripps, Advanced Techniques in RF Power Amplifier Design,
Ministry of Economic Affairs under grant BTS01063 and by Artech House, 2002.
Agere Systems.

R EFERENCES
[1] A.R.S. Bahai and B.R. Saltzberg, Multi-Carrier Digital Communications
- Theory and Applications of OFDM, Kluwer Academic/Plenum, 1999.
[2] R.D.J. van Nee and R. Prasad, OFDM for Wireless Multimedia
Communications, Artech House Publishers, 2000.
[3] J. Heiskala and J. Terry, OFDM Wireless LANs: a theoretical and
practical guide, Sams Publishing, Indianapolis, USA, 2002.
[4] S. Weinstein and P. Ebert, “Data transmission by frequency-division
multiplexing using the discrete Fourier transform,” IEEE Trans. on
Commun., vol. 19, pp. 628–634, Oct. 1971.
[5] IEEE 802.11a standard, ISO/IEC 802-11:1999/Amd 1:2000(E).
[6] IEEE 802.11g standard, Further Higher-Speed Physical Layer Extension
in the 2.4 GHz Band, 2003.
[7] Multiband OFDM Alliance, “MB-OFDM proposal specifications for
IEEE,” available from http://www.multibandofdm.org/, 2004.
[8] “Wireless USB Specification,” available from http://www.usb.
org/, 2005.