Spectral Efficiency Analysis in OFDM and

OFDM/OQAM based Cognitive Radio Networks

Haijian ZHANG, Didier LE RUYET, Michel TERRE
Electronics and Communications Laboratory, CNAM
Paris, France
[email protected], [email protected], [email protected]

AbstractThe future wireless communication is expected to

be able to improve the efficiency of spectrum usage. To solve
the challenge of spectrum shortage, an innovative opportunistic spectrum access strategy, called Cognitive Radio (CR) has
been proposed. Conventional Orthogonal Frequency Division
Multiplexing (OFDM) has already been suggested as a physical
layer candidate for CR system. Herein another potential candidate for CR, OFDM Offset Quadrature Amplitude Modulation
(OFDM/OQAM) is introduced, and its spectral efficiency for
coded multicarrier transmission is compared with Cyclic Prefix
based OFDM (CP-OFDM) and Raised Cosine windowed OFDM
(RC-OFDM) in CR context. Simulated results of Spectral Efficiency Comparison (SEC) for different multicarrier systems are
interpreted by theoretically analyzing the out-of-band radiation
of their prototype pulses shaping. Both theoretic analysis and
experimental results can show that OFDM/OQAM is a more
natural candidate than CP-OFDM and RC-OFDM for CR
networks application. 1

I. I NTRODUCTION
Multicarrier modulations attract a lot of attention ranging
from wireline to wireless communications compared to single
carrier modulation because of theirs capability to efficiently
cope with frequency selective fading channels. Much of attention in the present literature emphasizes on the use of
conventional OFDM, which is able to avoid both intersymbol
interference (ISI) and interchannel interference (ICI) making
use of a suitable cyclic prefix. OFDM has also been suggested
as a candidate for multicarrier based Cognitive Radio (CR)
systems [1]. However, in spite of these advantages, it is
very sensitive to fast time variations of the radio channel
and to frequency offsets due to imperfect synchronization.
Moreover, OFDM systems sacrifice data transmission rate
because of the insertion of CP. The objective of this paper is on
another alternative filter bank based multicarrier modulation
OFDM/OQAM [2] [5], which shows higher robustness to
residual frequency offsets than CP-OFDM by taking advantage
of the spectral containment of its modulation prototype filters.
Filter bank based multicarrier system is already considered as
a physical layer candidate for CR [6].
In a real OFDM based CR system problems arise from the
IFFT/FFT operation resulting in additional interference from
the CR system to the primary system and vice versa [1][7].
Using the IFFT transmitter implementation, the temporal pulse
1 Part

of this work has been supported by PHYDYAS UE FP7 project.

shape of one symbol is rectangular, resulting in a sincshaped frequency response on each subcarrier, thus OFDM
systems suffer from high side-lobe radiation. Herein the spectral efficiency of OFDM/OQAM based multicarrier systems
allowing for much lower out-of-band radiation are analyzed
and compared with OFDM. Recently, the spectral efficiency
analysis for OFDM and filter banks has been considered in [8],
where the trade-off between the level of interference caused
by secondary transmission to primary user and the spectral
efficiency of secondary user is studied in the CR context.
Numerical results show that using filter bank approach, a lowinterference secondary transmission can be obtained without
sacrificing the bandwidth efficiency.
The objective of this paper is to further focus on a complete
spectral efficiency analysis (SEA) in the CR context. The SEA
for a CR system contains two factors: time-frequency resource
exploitation (or throughput) of CR users and interference
level from CR user to licensed users. In practice, we usually
balance these two factors to maximize the spectral efficiency
under the premise to ensure compatibility between primary
and secondary systems. The out-of-band radiation property
is elaborated for various filter pulse shapes, the characteristics of which are confirmed by final simulated results. Two
typical OFDM/OQAM pulseshapes are used in this paper:
The Isotropic Orthogonal Transform Algorithm (IOTA) [4]
(IOTA/OQAM) and the reference filter [5] applied in the
project PHYDYAS [9] (PHYDYAS/OQAM).
The remainder of this paper is organized as follows: In
Section II, interference levels from OFDM/OQAM and OFDM
based CR systems are represented by analyzing the outof-side radiation of different prototype pulseshapes. Coded
multicarrier transmission is considered for spectral efficiency
analysis, for this purpose in section III we introduce a coded
SEA strategy. Simulation results are given in Section IV.
Finally, section V concludes this paper.
II. O UT- OF - SIDE R ADIATION A NALYSIS
In this section, four different multicarrier modulation systems are compared by quantifying the interferences of sidelobes radiation:
A. CP-OFDM based Spectrum Pooling Systems
We only investigate the interference from rental user (RU)
to licensed user (LU). In real OFDM systems cyclic prefix

978-1-4244-2517-4/09/$20.00 2009 IEEE

is a cyclic extension of the transmitted signal in the time

domain in order to overcome ISI in a multipath radio channel.
This temporal extension of one OFDM symbol results in a
narrower spectrum of the signals on each subcarrier. So the
power density spectrum (PDS) of each single carrier ofdm
signal is represented in the form [7]:
of dm (f ) = A2 Ts

 sinf Ts 2

Pof dm (n) =

Ptot

(n+1/2)fs

(n1/2)fs

of dm (f )df

(2)

where Ptot is the total transmit power emitted on one subcarrier and n represents the distance between the considered
subcarrier and the neighboring LU subband. In a pure OFDM
system the power in (2) does not cause any interference to
the adjacent subcarriers as their signals are orthogonal and
can be separated in the OFDM receiver [7]. In general, we
cannot assume that a potential licensed system is OFDMbased. Even if it is, it would have to use the same subcarrier
spacing and need to be synchronized with the rental system,
which contradicts the assumptions in Cognitive Radio. So the
signals of the licensed users are not orthogonal to the rental
users (RU). Thus the sidelobes of the sinc-shaped spectra on
each subcarrier fully interfere with the licensed users.
Calculating (2), the values of interference for different n are
displayed in the first column of Table I. In bad cases the mean
interference power one LU encounters from the RU can be as
large as 4 percent of the power transmitted on one subcarrier.
One measure for the mitigation of this mutual interference is
the introduction of adaptive guard bands in the rental system. It
implies the additional deactivation of one or more subcarriers
lying adjacent to allocated subbands of the licensed system.
Unfortunately, this measure sacrifices bandwidth of the rental
system.
TABLE I
INTERFERENCE POWER RATIO FROM



1
1
t

2 + 2 cos + Ts ,
prc (t) = 1,

 (tTs ) 
1 1
,
2 + 2 cos
Ts

f or
f or
f or

(1)

f Ts

where A denotes signal amplitude and Ts is the symbol duration which consists of the sum of useful symbol duration Tu
and guard interval. The bandwidth of one subcarrier spacing
is fs = 1/Tu , so the mean relative interference power to one
LU is defined as [7]:
1

pulses on each subcarrier and lower the sidelobes. A common

used raised cosine window is defined in [7]:

RU TO LU

Different Systems

CP-OFDM

IOTA/OQAM

PHYDYAS/OQAM

n=1
n=2
n=3
n=4

4.05%
0.68%
0.39%
0.25%

11.83%
2.78e-2%
5.56e-3%
3.04e-5%

6.38%
3.14e-5%
9.33e-7%
1.20e-7%

B. RC-OFDM based Spectrum Pooling Systems

Another interesting way to reduce the interference power
to the licensed system is the use of time domain windowing
techniques for the OFDM symbols of the rental system. Such
as the root raised cosine windows can sharpen the spectral

0 t < Ts
Ts t < Ts
Ts t < (1 + )Ts
(3)

where denotes the rolloff factor. = 0 corresponds the

typical CP-OFDM, herein RC windows with =0.25, 0.5,
0.75, 1 are considered for performance comparison. Using
the same way as CP-OFDM, the values of interference for
different n at each rolloff factor are displayed in Table II.
Obviously, the interference decreases as the increase of rolloff
factor .
TABLE II
RU TO LU FOR RC-OFDM

INTERFERENCE POWER RATIO FROM

Different
rolloff factor

= 0.25

= 0.5

= 0.75

=1

n=1
n=2
n=3
n=4

3.70%
0.34%
7.26e-2%
8.84e-3%

2.52%
2.16e-2%
1.29e-3%
2.41e-4%

1.52%
2.78e-3%
1.15e-4%
3.48e-5%

0.89%
5.07e-4%
4.46e-5%
5.83e-6%

We can see that RC-OFDM owns positive effect to reduce

out-of-band radiation. But this method can only be carried out
at the expense of a longer symbol duration, which reduces the
throughput of CR system.
C. IOTA/OQAM based Spectrum Pooling Systems
Le Floch [4] gives an overview of the main features concerning IOTA. The closed-form expression z,0 ,0 can be found
in [10]. An orthogonalization method generates the function
z,0 ,0 :
z,0 ,0 (t) = O0 F1 O0 Fg (t)

(4)

where F is the Fourier transform operator, 0 and 0 are the

frequency and time real parameters of the modulation system
such that 0 0 = 12 , Oa is an orthogonalization operator with
a equals to 0 or 0 , which transforms a function x into a
function y according to:
y(u) =

x(u)
|x(uka)|2

(5)

and finally g (t) = (2)1/4 et with , which is the

spreading parameter.
The function z,0 ,0 satisfies: Fz,0 ,0 = z1/,0 ,0 , and
in the special case where = 1 and 0 = 0 , we get
the function  such that F = . Therefore the time
and frequency properties of  are nearly isotropic, which is
called the Isotropic Orthogonal Transform Algorithm (IOTA)
prototype function. In order to derive z,0 ,0 from (4), one
has to use the following closed-form expression:

z,0 ,0 =

1
2


dk,,0

Considering the general situation, we assume that one LU

band with l subcarriers is surrounded by RU band with
subcarriers length n on both sides. The mean interference
power density from CR users is normalized [7]:


g (t + k0 ) + g (t k0 )

k=0

dl,1/,0 cos(2l t0 )

(6)

P RL =

l=0

where the dk,,0 are real coefficients, which can be computed

according to a set of rational coefficients [10]. So the Power
Density Spectrum of IOTA signal on one single carrier band
and the mean relative interference power can be similarly
derived:

2
iota (f ) = A2 z1,2/2,2/2
 (n+1/2)fs
1
Piota (n) = Ptot
iota (f )df

(7)

(n1/2)fs

where A is the signal amplitude and fs is the subcarrier

spacing. The second column in Table I lists the interference
values for different n. We can see that the sidelobes of IOTA
function decline rapidly and cause very small interference to
the neighboring users for n > 1.

n
BPSK
Guard
band

2
M )

(8)

n=1

Similarly, the PDS of OQAM signal on one subcarrier

and the mean relative interference power to LU subbands are
calculated by:
phy
Pphy


2
= A2 F0 (f )
 (n+1/2)fs
1
phy
oqam (n) = Ptot
oqam (f )

(n1/2)fs

oqam (f )df

4QAM
Noise

16QAM

Primary
User

The PHYDYAS/OQAM prototype coefficients are defined

as h[n] [11], with n = 0, . . . , L 1, where L = KM + 1,
K is the length of each polyphase components and M is the
number of subcarriers (M is even). We define the variable =
KM/2 and assume that the prototype function is symmetry
around th coefficient, this means that h[n] = h[KM n].
The individual amplitude of the th subcarrier is then given
by:

(9)

The interference values for LUs are shown in the third

column of Table I. We can see that almost perfect zero interference can be derived in this special OQAM-based system
for n > 1.
III. S PECTRAL E FFICIENCY A NALYSIS
In this part we introduce a spectral efficiency analysis
strategy corresponding to coded multicarrier transmission.
SEA of one CR system includes two performance parameters:
interference from CR user to LU user and throughput of
CR user.

(10)

where PRL is the accumulated interferences on the l primary

subcarriers from the n active secondary subcarriers. It is already proved in [7] the wider the LU band, the less interference
from CR users.
In practical cognitive radio system, each RU subcarrier
carries different bits according to required BER under adaptive
modulation strategy. A concise explanation for adaptive modulation strategy in cognitive radio system is illustrated in Fig. 1,
where for the five RU subcarriers under different interference
and noise situations, some criteria can be used to adapt each
subcarrier to their optimal constellation size.

D. PHYDYAS/OQAM based Spectrum Pooling Systems


F () = h[] + 2
h[ n]cos n(

PRL
l

Primary
User

CR
Sub1

Fig. 1.

CR
Sub2

CR
Sub3

CR
Sub4

CR
Sub5

Different modulation scheme choices for each RU subcarrier

When a return channel is available from receiver to transmitter, a more powerful feature than adaptive modulation without
coding known as Adaptive Coding and Modulation (ACM)
can be applied. With ACM it is possible to dynamically
modify the coding rate and modulation scheme for every single
frame, according to the measured channel conditions where the
frame is received. This technique provides more exact channel
protection and exploits the capacity of broadband frequency
selective channels.
In order to reduce the signaling overhead, it has been
recommended that the minimum unit (single frame) for linkadaptation and resource partitioning is a rectangular area in
time and frequency denoted as chunk. The size of one chunk
should be chosen based on the coherence time and bandwidth
of the channel such that the fading within a chunk is essentially
flat. The number of payload bits a chunk can offer depends on
the modulation format and hence, on the channel quality. In
[12], the authors present an adaptive coding and modulation
scheme for bit-interleaved coded OFDM systems using RateCompatible Punctured Block-circulant Low-Density ParityCheck codes (RCP-BLDPC). This error correcting code is
known to provide good performance together with efficient
hardware implementation. The throughput is maximized under
a Codeword Error Rate (CWER) constraint. The number of

bits per subsymbol is adapted per chunk depending on its

short-term predicted Channel Gain to Noise Ratio (CNR) level.
Finally, the puncturing rate of the outer code is adjusted to
achieve the target CWER.
For target CW ER = 0.01, paper [12] generates
the required SNR values for the code rate set R =
24 24 24 24 24 24 24
{ 24
48 , 44 , 40 , 36 , 32 , 30 , 28 , 26 } and for information lengths
K = 288 and K = 1152. Herein we use the information
size K = 288 as the chunk size of cognitive radio user.

length equaling 1/4 time of one OFDM symbol, RCOFDM based multicarrier network with the rolloff factor =0.25, 0.5, 0.75, 1, IOTA/OQAM and PHYDYAS/OQAM based multicarrier systems.
3) The licensed system is supposed to have the same
physical layer modulation scheme with its corresponding
CR system.
A. Interference for Licensed User

VXEFDUULHUV

FKXQNL
QG

IUHTXHQF\

Fig. 2.
Adaptive coding and modulation scheme for chunks based CR
subcarriers

Fig. 2 shows the adaptive coding and modulation scheme

for chunks based CR subcarriers, where n is the number of
maximum subcarriers for one CR user and d is the number
of guard band. The ith chunk size 4Ni is chosen based on
the characteristic of CR network, where Ni is the number of
symbols for ith chunk. From Fig. 1 we can see that the SINR
values of 4 subcarriers within one chunk closing to primary
user are obviously different, in this case we consider the worst
SINR value of the 4 subchannels as the mean SNR value of
one chunk.
Therefore, the throughput of one coded cognitive radio
user during one reference transmission time is defined as:
T hcoded =

(Th Td )

n2d
i=1

nTh

Ri log2 Mi

(bits/s/Hz) (11)

where is the useful data ratio (except cyclic prefix or overlapped data), Th is the reference transmission time length
(duration of one spectrum hole), Td is the spectrum sensing
time, n is the number of maximum available subcarriers for
one CR user, d is the length of guard band as shown in Fig.
1, Ri is the code rate and Mi is the constellation size for
ith subcarrier.
IV. S IMULATION RESULTS
In this section, we display numerical simulation results
to evaluate the spectrum efficiency of different multicarrier
modulated systems under an AWGN channel.
The simulations are performed under the analysis scenario
as follows:
1) Assuming the total number of subcarriers is 1024, we
choose one LU and RU bandwidth l = n = 32 for the
following simulations.
2) Four different multicarrier CR systems are investigated:
CP-OFDM based multicarrier network with the CP

Interference power for LU in dB

WLPH

   

1L V\PEROV

0
CPOFDM
RCOFDM =0.25
RCOFDM =0.5
RCOFDM =0.75
RCOFDM =1
IOTA/OQAM
PHYDYAS/OQAM

20

40

60

80

100

120

Guard band length d

Fig. 3.

Interference power for different multicarrier modulation systems

Referring to Table I and II, the normalized interference

power as a function of guard band length is graphed in
Fig. 3. Because of the improved spectral concentration, RCOFDM achieves lower interference level than CP-OFDM. At
the same time it can be observed that the filter bank based
OQAM systems show a very fast and consistent attenuation
of interference power. For a given interference power threshold
(-80dB), PHYDYAS/OQAM and IOTA/OQAM can reach this
threshold with only 2 and 5 widths of guard band, but bigger
guard bandwidth is needed for CP-OFDM and RC-OFDM. On
the contrary, considering a fixed guard band width (d=5), PHYDYAS/OQAM and IOTA/OQAM generate smaller interference
power to neighboring LUs than OFDM, this property makes
OFDM/OQAM more robust for cognitive radio application.
B. Throughput for Rental User
Generally, the spectrum sensing time is negligible comparing to one duration of spectrum hole especially for TV
channels, that is Th Td Th . Defining -30 dB as the
limited interference level on a licensed system, below which
the secondary user will not cause harmful impact to primary
user. According to (11), throughput of RU as a function
of guard bandwidth at a fixed SNR (20dB) with channel
coding is shown in Fig. 4. Obviously, OFDM/OQAM systems
without CP show higher throughput than OFDM systems at
the same guard band length. The low throughput for OFDM
results mainly from repeated and overlapped transmission
data and slightly from big side lobes. Besides, the raised
cosine windowing with increasing clearly pushes down the

another important reason for the limitation performance of

OFDM. Within the same conditions, OFDM/OQAM systems
achieve higher performance comparing with OFDM. Although
it also needs to make a tradeoff between interference and
throughput, big improvements of system performance for
both interference and throughput are observed in Fig. 5. The
prototype filter of PHYDYAS/OQAM has a better spectral
efficiency than IOTA/OQAM because of relative small outof-band radiation. So quasi-perfect spectral efficiency of CR
network is possible by designing small side-lobe prototype
filter.

5.5
CPOFDM
RCOFDM =0.25
RCOFDM =0.5
RCOFDM =0.75
RCOFDM =1
IOTA/OQAM
PHYDYAS/OQAM

Throughput for RU (bits/s/Hz)

5
4.5
4
3.5
3
2.5
2
1.5
1
0.5

Guard band length d

Fig. 4. Throughputs for different guard bandwidths at a fixed SNR=20dB

with RCP-BLDPC

throughput due to increased symbol period, this overhead is

not expected for CR system.
C. Tradeoff between throughput and interference

5.5
5

CPOFDM
RCOFDM =0.25
RCOFDM =0.5
RCOFDM =0.75
RCOFDM =1
IOTA/OQAM
PHYDYAS/OQAM

Throughput for RU (bits/s/Hz)

4.5
4
3.5

R EFERENCES

3
2.5
2
1.5
1
0.5
120

V. C ONCLUSION
In this paper available time-frequency resource exploitation
and corresponding interference power of OFDM/OQAM based
systems are compared to OFDM in the CR context. According
to the out-of-band radiation analysis of different multicarrier
systems and the throughput definition of RU, we demonstrate that as a CR data communication technique, filter bank
multicarrier modulation offers higher spectral efficiency than
OFDM. As a result, OFDM/OQAM is a natural candidate for
physical layer data communication in cognitive radio systems.
Conventional OFDM and raised cosine windowed OFDM
have to compromise between two important performance parameters: throughput of secondary user and interference on
primary user. Although there is a little additional complexity
for OFDM/OQAM system structure due to prototype pulse
shaping, it is profitable for achieving better spectral efficiency.

100

80

60

40

20

Interference power for LU (dB)

Fig. 5. The relationship between throughput and interference at a fixed

SNR=20dB with RCP-BLDPC

The ideal situation for CR network we expect is: maximum

throughput for rental system and minimum interference for
licensed system. Lots of literatures proposed to reach this
Two-Win effect, but we always need to do some compromise
between different performance parameters. Fig. 5 displays the
mutual impact between throughput of RU and interference
of LU of OFDM and OQAM based multicarrier systems for
coded transmission. For CP-OFDM with large side-lobe, in
order to ensure required throughput of RU, it is inevitable to
cause heavy interference level to LU. So it is reasonable to
overcome the interference by windowing the OFDM transmission signal, but the decreased interference gains is obtained
by sacrificing the throughput of RU. As shown in Fig. 5, the
performance curves of RC-OFDM with varying rolloff factors
are like Seesaw Game, the less interference you want, the
more throughput you will lose. Furthermore, cyclic prefix is

[1] T. Weiss, F. Jondral, Spectrum Pooling: An Innovative Strategy for the

Enhancement of Spectrum Efficiency, IEEE Communications Magazine, Vol. 42, no. 3, March 2004, pp. 8-14.
[2] B. Hirosaki, An Orthogonally Multiplexed QAM System Using the
Discrete Fourier Transform, IEEE Trans. On Comm. Tech, Vol. 29,
Issue 7, Jul 1981, pp. 982-989.
[3] P. Siohan, C. Siclet and N. Lacaille, Analysis and design of
OFDM/OQAM systems based on filterbank theory, IEEE Trans. Signal
Processing, Vol. 50, May 2002, pp. 1170-1183.
[4] B. Le Floch, M. Alard, C. Berrou, Coded orthogonal frequency division
multiplex, Proceedings of the IEEE, Vol. 83, June 1995, pp. 982-996.
[5] M. Bellanger, Specification and design of a prototype filter for filter
bank based multicarrier transmission, Proceedings of IEEE ICCASP,
Vol. 1, pp. 2417-2420, 2001.
[6] P. Amini, P. Kempter, RR Chen, L Lin, FILTER BANK MULTITONE:
A PHYSICAL LAYER CANDIDATE FOR COGNITIVE RADIOS,
Proceeding of the SDR Forum technical Conference, Nov. 2005, pp.
14-18.
[7] T. Weiss, J. Hillenbrand, Mutual interference in OFDM-based spectrum
pooling systems, Vehicular Technology Conference, Vol. 4, May 2004,
pp. 1873- 1877.
[8] Tero Ihalainen, Ari Viholainen, Markku Renfors, On Spectrally Efficient Multiplexing in Cognitive Radio Systems, International symposium on wireless pervasive computing, May 2008, pp. 675- 679.
[9] PHYDYAS-Physical layer for dynamic spectrum access and cognitive
radio, Project website: www.ict-phydyas.org.
[10] Pierre Siohan, Christian Roche, Cosine-Modulated Filterbanks Based
on Extended Gaussian Functions, IEEE Transactions on signal processing, Vol. 48, pp. 3052-3061, 2000.
[11] Leonardo G. Baltar, Dirk S. Waldhauser, Out-Of-Band Radiation in
Multicarrier Systems: A Comparison, Lecture Notes in Electrical
Engineering, Vol. 1, pp. 107-116, 2007.
[12] S. Stiglmayr, M. Bossert, E. Costa, Adaptive Coding and Modulation
in OFDM Systems using BICM and Rate-compatible Punctured Codes,
in European Wireless Conf, Paris, 2007.