964

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 64, NO. 3, MARCH 2015

Joint PAPR Reduction and Sidelobe Suppression

Using Signal Cancelation in NC-OFDM-Based
Cognitive Radio Systems
Chunxing Ni, Tao Jiang, and Wei Peng

AbstractIt is well known that the high peak-to-average

power ratio (PAPR) and large spectrum sidelobe power are two
main drawbacks at the transmitter in noncontiguous orthogonal
frequency-division multiplexing (NC-OFDM)-based cognitive
radio (CR) systems. In this paper, we propose a novel signal
cancelation (SC) method for joint PAPR reduction and sidelobe
suppression in NC-OFDM-based CR systems. The key idea of the
proposed method is to dynamically extend part of the constellation
points on the secondary user (SU) subcarriers and add several
SC symbols on the primary user (PU) subcarriers to generate the
appropriate cancelation signal for joint PAPR reduction and sidelobe suppression. Then, the SC method formulates the problem of
the joint PAPR reduction and sidelobe suppression as a quadratically constrained quadratic program (QCQP) to obtain its optimal cancelation signal. Moreover, we also propose a suboptimal
SC (sub-SC) method to efficiently solve the QCQP optimization
problem with low computational complexity. Simulation results
show that the proposed SC method and the sub-SC method can
provide both significant PAPR reduction and sidelobe suppression
performances.
Index TermsCognitive radio (CR), noncontiguous orthogonal
frequency-division multiplexing (NC-OFDM), peak-to-average
power ratio (PAPR), sidelobe suppression.

I. I NTRODUCTION

ECENTLY, cognitive radio (CR) has drawn significant

attention from the academic and industrial communities
to meet the ever-growing needs of spectrum resources and
high-data-rate communication [1][3]. For CR systems, noncontiguous orthogonal frequency-division multiplexing (NCOFDM) is an attractive physical-layer technology due to its
considerable high spectrum efficiency, multipath delay spread
tolerance, immunity to the frequency-selective fading channels,
and high power efficiency [4].

Manuscript received December 19, 2013; revised April 13, 2014; accepted
May 16, 2014. Date of publication May 30, 2014; date of current version
March 10, 2015. This work was supported in part by the National Science
Foundation for Distinguished Young Scholars of China under Grant 61325004,
the National Science Foundation of China under Grant 61172052 and Grant
60872008, the National High Technology Development 863 Program of China
under Grant 2014AA01A704, and the Key Program Research Plan of the National Natural Science Foundation of Hubei in China under Grant 2013CFA055.
The review of this paper was coordinated by Dr. J.-C. Chen.
The authors are with the Department of Electronics and Information
Engineering, Huazhong University of Science and Technology, Wuhan 430074,
China (e-mail: chunxingni@gmail.com; tao.jiang@ieee.org; pengwei@
ieee.org).
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TVT.2014.2327012

Fig. 1.

NC-OFDM-based CR system coexisting with the PU and SUs.

Fig. 1 shows the coexistence of the NC-OFDM-based CR

system with both secondary users (SUs) and primary users
(PUs). For the NC-OFDM-based CR system, the SUs are
cognitive unlicensed users that detect and utilize the subcarriers
located in the unused spectrum band for data transmission.
Moreover, the PU subcarriers, which are located in the PU
spectrum band, need to be turned off to create spectrum
notches to limit the interference to the PUs in the conventional
NC-OFDM-based CR system [5].
Although the NC-OFDM-based CR system has many advantages, it still suffers from two main drawbacks. The first main
drawback is its high peak-to-average power ratio (PAPR) of the
transmitted NC-OFDM signals. Since the high-power amplifier
(HPA) used in the NC-OFDM-based CR system has a limited
linear range, the NC-OFDM signals with high PAPR will be
seriously clipped, and nonlinear distortion will be introduced,
resulting in serious degradation of the bit error rate (BER)
performance [6], [7]. Moreover, the high PAPR leads to out-ofband radiation, which causes serious adjacent channel interferences [8]. The second main drawback of the NC-OFDM-based
CR system is its large spectrum sidelobe. The large spectrum
sidelobe introduces interference to the adjacent PUs, resulting
in the serious performance degradation of the adjacent PUs [9].
Hence, it is highly desirable to suppress the spectrum sidelobe
as much as possible in the NC-OFDM-based CR system.
Recently, various methods have been proposed to reduce
the PAPR for the NC-OFDM-based CR system in the literature, such as clipping [10], partial transmit sequence [11],
[12], active constellation extension [13], and tone reservation
[14], [15]. These PAPR reduction techniques can efficiently
reduce the PAPR of transmitted signals; however, they do not
take sidelobe suppression into account.
Moreover, to suppress the sidelobe of the NC-OFDM-based
CR system, many schemes have been proposed in the literature,

0018-9545 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

NI et al.: PAPR REDUCTION AND SIDELOBE SUPPRESSION USING SC IN NC-OFDM-BASED CR SYSTEMS

such as active interference cancelation (AIC) [16], extended

AIC (EAIC) [5], constellation adjustment (CA) [17], pulse
shaping (PS) [18], spectrum precoding (SP) [19], and sidelobe
suppression with orthogonal projection (SSOP) [20]. The AIC
method utilizes some subcarriers as the guard band to suppress
the sidelobe, resulting in the decrease in spectrum efficiency.
The EAIC method utilizes the tones spaced closer than the
interval of NC-OFDM subcarriers to generate the cancelation signals for the sidelobe suppression in the NC-OFDMbased CR system. However, the EAIC method destroys the
orthogonality of the NC-OFDM-based CR system, which will
introduce serious interferences. Thus, the EAIC method leads
to the degradation of the BER performance in the NC-OFDMbased CR system. For the CA method, some subcarriers are
multiplied by some appropriate weights to generate alternative signals, and then, the signal with the minimum spectrum
sidelobe is chosen as the transmitted signal. Thus, the CA
method must send the selected weights as side information to
the receiver for data recovery, resulting in the decrease in the
data rate. Both the PS and SP methods can efficiently suppress
the sidelobe of the NC-OFDM-based CR system by shaping
the waveform of NC-OFDM signals. However, they suffer from
high computational complexity. The SSOP method utilizes
an orthogonal projection matrix for sidelobe suppression and
adopts several reserved subcarriers for recovering the distorted
signal in the receiver. Thus, the SSOP method suffers from the
decrease in the data rate. Furthermore, although these sidelobe
suppression methods can suppress the sidelobe power in the
NC-OFDM-based CR system, all of them do not consider
PAPR reduction.
Since the PAPR reduction methods may affect the spectrum
sidelobe of the NC-OFDM-based CR system, we must jointly
consider PAPR reduction and sidelobe suppression. To the
authors best knowledge, few researches investigate the joint
PAPR reduction and sidelobe suppression in the NC-OFDMbased CR system. In [21], the selected mapping (SLM) technique is employed to jointly reduce the PAPR and suppress the
sidelobe in the NC-OFDM-based CR system, and its key idea
is to generate several alternative signals and select the signal
with low PAPR and sidelobe as the transmitted signal. However,
the SLM method cannot achieve significant PAPR reduction
and sidelobe suppression performances. Moreover, the SLM
method must reserve several subcarriers to transmit the side
information, resulting in the decrease in the data rate.
In this paper, we propose a novel scheme, which is named
signal cancelation (SC), for joint PAPR reduction and sidelobe
suppression in the NC-OFDM-based system. For the proposed
SC method, part of the outer constellation points on SU subcarriers is dynamically extended, whereas several SC symbols are
added on the PU subcarriers, to generate the appropriate cancelation signal for joint PAPR reduction and sidelobe suppression.
Then, the SC method formulates the problem of the joint
PAPR reduction and sidelobe suppression as a quadratically
constrained quadratic program (QCQP) [22], and the optimal
cancelation signal can be obtained by convex optimization.
Moreover, we also propose a suboptimal SC (sub-SC) method
to efficiently solve the QCQP optimization problem with low
complexity. Simulation results show that the proposed SC

965

method and the sub-SC method can provide both significant

PAPR reduction and sidelobe suppression performances.
The rest of this paper is organized as follows. We briefly
describe the NC-OFDM-based CR system and introduce the
definition of PAPR and spectrum sidelobe in Section II. In
Section III, we propose the SC method to jointly reduce the
PAPR and suppress the sidelobe in detail. The sub-SC method
is discussed in Section IV, and simulation results are shown in
Section V, followed by conclusions in Section VI.
II. P ROBLEMS IN THE N ON -C ONTIGUOUS O RTHROGONAL
F REQUENCY-D IVISION M ULTIPLEXING -BASED C OGNITIVE
R ADIO S YSTEM
A. PAPR
For the NC-OFDM-based CR system, the PUs utilize the
PU spectrum band for data transmission, whereas the SUs can
detect and utilize the unoccupied spectrum band of the PUs
(i.e., SU spectrum band) for data transmission. Thus, the SUs
utilize the SU subcarriers located in the SU spectrum band for
data transmission. Moreover, in the conventional NC-OFDMbased CR system, the SUs must deactivate the PU subcarriers
that are located in the PU spectrum band to create spectrum
notches to limit the interference to the PUs. As shown in
Fig. 1, for a conventional NC-OFDM-based CR system with
N subcarriers, among which L = (b a + 1) subcarriers are
occupied by the PUs (from the ath to the bth subcarriers)
and the remaining N L subcarriers (from the zeroth to the
(a 1)th subcarriers, and from the (b + 1)th to the (N 1)th
subcarriers) are utilized by the SUs for data transmission.
Denote R = {0, 1, . . . , a 1, b + 1, b + 2, . . . , N 1} as the
subset that consists of indexes of the SU subcarriers and Rc =
{a, . . . , b} as the subset that consists of indexes of the PU
subcarriers.
For an NC-OFDM symbol X = [X(0), X(1), . . . , X(N
1)]T , it is obvious that X(k) = 0 for k  R. Then, the timedomain OFDM signal x = [x(0), x(1), . . . , x(N 1)]T can
be generated by employing N -point inverse discrete Fourier
transform (IDFT) to X, i.e.,
N 1
1 
X(k)ej2knf /fs
x(n) =
N k=0
N 1
1 
=
X(k)ej2kn/N ,
N k=0

n = 0, 1, . . . , N 1

(1)

where f is the frequency interval between adjacent subcarriers, and fs = N f is the total bandwidth.
To better approximate the PAPR of continuous-time NCOFDM signals, the NC-OFDM signal samples are obtained by
J-times oversampling. Thus, (1) can be rewritten as
N 1
1 

X(k)ej2knf /Jfs
x(n) =
N k=0
N 1
1 
=
X(k)ej2kn/JN ,
N k=0

n = 0, 1, . . . , JN 1. (2)

966

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 64, NO. 3, MARCH 2015

Moreover, (2) can be rewritten in matrix form as

x = FX

(3)

where
F CJN N is the IDFT matrix with the (r, s)th element
as (1/ N )ej2rs/JN .
In general, the PAPR of the NC-OFDM signal x is defined
as the ratio of the maximum instantaneous power to the average
power, i.e.,
P AP R =

max |x(n)|2
x2


= 1 2
E |x(n)|2
JN x2

0nJN 1

(4)

where E[] represents the expectation, and  2 and  

denote the 2 norm and infinity norm, respectively.
The complementary cumulative distribution function
(CCDF) is widely employed to measure the PAPR reduction
performance, which is defined as the probability that the PAPR
exceeds a given threshold P AP R0 , i.e.,
CCDF = Pr{P AP R > P AP R0 }.

(5)

Since the high PAPR of NC-OFDM signals leads to serious

nonlinear distortion and power efficiency degradation of the
HPA, it is necessary to reduce the PAPR of the NC-OFDM
signals.
B. Spectrum Sidelobe
For an NC-OFDM symbol, to measure the spectrum sidelobe
power in the PU spectrum band, we calculate the spectrum at
the sample points {f0 , f1 , . . . , fM 1 }, and the sampled sidelobe at frequency fm is given as
N 1
1 
Sd (m) =
x(n)ej2nfm /fs , m = 0, 1, . . . , M 1.
N n=0
(6)

Substituting (1) into (6), we have

N 1 N 1
1 
Sd (m) =
X(k)ej2n(kf fm )
N n=0

Fig. 2.

Block diagram of the SC method.

For the conventional NC-OFDM-based CR system, the PU

subcarriers are turned off at the occupied PU band to create
spectrum notches to limit interference to the PUs. However,
the total sidelobe power of the NC-OFDM signals can still
cause nonnegligible interference when the PU subcarriers are
turned off [5]. Thus, some techniques have been proposed to
suppress the sidelobe of the NC-OFDM-based CR system, such
as the AIC method. Different to the conventional NC-OFDMbased CR system, which turns off the PU subcarriers, the AIC
method suppresses the sidelobe by adding several interferencemitigation symbols on the PU subcarriers. Nevertheless, the
sidelobe suppression performance of the AIC method is still
unsatisfied, and the AIC method must utilize some spectrum
bands as the guard band to suppress the sidelobe, resulting in
the decrease in spectrum efficiency.
In summary, the high PAPR of the transmitted signals and
the large sidelobe power are two main drawbacks of the NCOFDM-based CR system, which seriously impact the quality
of communication services. Therefore, we propose a novel SC
method to jointly reduce the PAPR and suppress the sidelobe
power for the NC-OFDM-based CR system.

k=0

m = 0, 1, . . . , M 1.

(7)

Thus, (7) can be rewritten in matrix form as

Sd = Pd X

(8)

where the sampled sidelobe matrix Sd is denoted as Sd =

[Sd (0), Sd (1), . . . , Sd (M 1)]T . Pd is an M N matrix with
 1 j2kn/N j2nfm /fs
e
.
the (m, k)th element as (1/N ) N
n=0 e
Moreover, the total sidelobe power SL in the target spectrum
band can be measured by the sum of the sampled energy
at frequency points {f0 , f1 , . . . , fM 1 }, which can be expressed as
SL =

M
1

m=0

|Sd (m)|2 = Sd 22 .

(9)

III. P ROPOSED S IGNAL C ANCELLATION M ETHOD

Here, we propose a novel SC method for joint PAPR reduction and sidelobe suppression in the NC-OFDM-based CR system, and Fig. 2 shows the diagram of the proposed SC method.
Different to the conventional NC-OFDM-based CR scheme,
which turns off the PU subcarriers to create spectrum notches
to limit the interference to PUs, the SC method utilizes both
the SU and PU subcarriers to generate the cancelation signals
for joint PAPR reduction and sidelobe suppression. Moreover,
for the SC method, part of the outer constellation points is
dynamically extended on SU subcarriers, whereas several SC
symbols are added on PU subcarriers. Note that the SC symbols
on PU subcarriers are simply for joint PAPR reduction and
sidelobe suppression in the NC-OFDM-based CR system, and
SUs cannot utilize PU subcarriers for data transmission. Due
to the extended constellation points on SU subcarriers and the

NI et al.: PAPR REDUCTION AND SIDELOBE SUPPRESSION USING SC IN NC-OFDM-BASED CR SYSTEMS

Fig. 3.

967

Constellation region on SU subcarriers with 4-QAM and 16-QAM, respectively.

SC symbols on PU subcarriers, the SC method achieves better

PAPR reduction and sidelobe suppression performances than
the conventional turning-off method.
Fig. 3 shows the constellation extension regions when the
quadrature amplitude modulation (QAM) is employed for the
SC method. For the SU subcarriers, the constellation points
are classified as inner points, boundary points, and corner
points. Inner points cannot be extended, boundary points can
be extended only outward in one direction along the arrowed
lines, and corner points can be extended to the shaded regions
[13]. As shown in Fig. 3, it is obvious that the minimum
Euclidean distance of the modified constellation points can
be maintained the same as that of the original constellation
points; thus, the BER performance of the SC method can be
maintained.
For the proposed SC method, on the SU subcarriers,
the constellation adjustment symbol Cs = [Cs (0), Cs (1), . . . ,
Cs (N 1)] is added to the original data symbol X, where
Cs (k) = 0 for k  R and the constellation adjustment symbol Cs (k) must obey the principle rule shown in Fig. 3.
Denote M as the constrained space for Cs (k), it is obvious
that Cs (k) M for k R. Moreover, on the PU subcarriers,
the SC symbol Cp = [Cp (0), Cp (1), . . . , Cp (N 1)] is added,
where Cp (k) = 0 for k R. If we denote C = Cs + Cp as the
total SC symbol of the SC method, then we have

C(k) =

Cs (k),
Cp (k),

kR
k  R.

(10)

 of the SC method can be

Thus, the transmitted symbol X
expressed as
 = X + Cs + Cp = X + C.
X

(11)

According to (3) and (11), the transmitted time-domain sig of the SC method can be obtained as
nal x
 = FX + FCs + FCp = FX + FC.
x

(12)

Therefore, the joint PAPR reduction and sidelobe

suppression problem can be formulated as an optimization
problem, i.e.,
min
C

FX + FC

(13a)

subject to Pd X + Pd C2 Pd X2

X +

C22

(13b)
(13c)

C(k) M, for k R

(13d)

where (13a) is the objective function that represents the PAPR

minimization with the proposed SC method. Equation (13b)
represents the sidelobe power constraint of the NC-OFDMbased CR system, and denotes the parameter to control the
maximum sidelobe power, which is calculated according to desired system characteristics. In (13b), the sidelobe power of the
proposed SC method should be equal to or less than times the
sidelobe power of the conventional NC-OFDM-based CR system. Since the total SC symbol C is added to the original data
block X, the total power of the SC method can be expressed as
X + C22 ; then, the SC method slightly increases the power of
the transmitted NC-OFDM signals. Thus, the total power of the
SC method must be restrained, and (13c) represents the power
constraint of the SC method, where  denotes the maximum
power. Equation (13d) represents the constellation modification
constraint, and the extended points on SU subcarriers must obey
the principle rule shown in Fig. 3. For example, when 4-QAM
is employed for the proposed SC method in NC-OFDM-based
CR systems, (13d) can be rewritten as
Re (X(k)) Re (C(k)) 0,

kR

Im (X(k)) Im (C(k)) 0,

kR

(14)

where Re(u) and Im(u) denote the real part and the imaginary
part of a complex number u, respectively.
Obviously, (13a) is a convex function [23], whereas
(13b)(13d) are convex sets. Thus, the joint PAPR reduction
and sidelobe suppression problem (13) is a convex QCQP.

968

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 64, NO. 3, MARCH 2015

Moreover, the convex optimization problem (13) can be

reformulated as a second-order cone program (SOCP) [24], i.e.,
min
C

subject to

(15a)


x(n)2 p,

n = 0, 1, . . . , JN 1

(15b)

 = FX + FC
x

(15c)

Pd X + Pd C2 Pd X2

(15d)

X +

(15e)

C22

C(k) M, for k R

(15f)

p 0.

(15g)

The given convex SOCP model can be solved by the standard

interior-point method [22], and the computational complexity
of the standard interior-point method is O(N 3 ) during each
iteration [23]. Moreover, one IDFT operation is needed during each iteration, which has the complexity of JN log JN .
Therefore, the total computational complexity of the SC method
can be expressed as O(N 3 + JN log JN ).
IV. S UB -S IGNAL C ANCELLATION M ETHOD
As discussed in Section III, the proposed SC method can
efficiently reduce the PAPR and suppress the sidelobe in
the NC-OFDM-based CR system. However, its computational
complexity is a bit high. Thus, here, we propose a sub-SC
method to efficiently reduce the PAPR and suppress the sidelobe with low computational complexity in the NC-OFDMbased CR system.
For the SC method, the constellation adjustment symbol Cs
on SU subcarriers and the SC symbol Cp on PU subcarriers
are both for the joint PAPR reduction and sidelobe suppression
in the NC-OFDM-based CR system. However, for the sub-SC
method, the constellation adjustment symbol Cs on SU subcarriers is only employed to generate the peak-canceling signal
for PAPR reduction, and it does not consider the sidelobe
suppression. Meanwhile, the SC symbol Cp on PU subcarriers
is only utilized for the sidelobe suppression, and it does not take
PAPR reduction into account.
Thus, the joint PAPR reduction and sidelobe suppression
problem (15) can be reformulated as two problems: The first
problem is to reduce the PAPR by dynamically extending
part of the constellation points on SU subcarriers according
to the principle rule shown in Fig. 3; the second problem is
to suppress the sidelobe power by adding SC symbol Cp on
PU subcarriers. Solving one of these two problems separately
without considering the other can impact the performance of the
other problem. For example, when the PAPR is independently
reduced, the sidelobe suppression performance will be affected.
Thus, the proposed sub-SC method is an iterative algorithm,
and it serially solves the two problems during each iteration.
Furthermore, the power constraint of the sub-SC method should
also be maintained, as discussed in (15e), for the sub-SC
method. To maintain the power constraint, the sub-SC method
terminates the iteration whenever the power constraint (15e) is
not satisfied.

A. PAPR Reduction
For the proposed sub-SC method, we should first initialize
the SC symbol Cp = 0 and the constellation adjustment symbol
Cs = 0. During each iteration, we first solve the first problem
and then solve the second problem. For the first problem, the
SC symbol Cp on PU subcarriers is seen as a constant, which
is obtained by (22) in the last iteration. Then, the constellation adjustment symbol Cs on SU subcarriers is employed to
generate the peak-canceling signal for PAPR reduction in the
NC-OFDM-based CR system. Moreover, Cs must obey the
principle rule shown in Fig. 3. Thus, the first problem can be
formulated as follows:
(P1) :

FX + FCs + FCp 

(16a)

subject to Cs (k) M, for k R.

(16b)

min
Cs

For (P1), Cs can be obtained by the projection onto convex

sets algorithm [25] as follows:
1) Take a JN -point IDFT operation to the data block X +
 = [
x(0),
Cp to obtain the discrete time-domain signal x
x
(1), . . . , x
(JN 1)].
2) Clip any signal whose amplitude is larger than A, where
A is the preset clipping threshold. Thus, the clipped timedomain signal x can be obtained by

x
(n),
|
x(n)| < A
x(n) =
(17)
Aej(n) , |
x(n)| A
where x
(n) = |
x(n)|ej(n) .
3) Calculate the clipping noise w = [w(0), w(1), . . . ,
w(JN 1)] by
w(n) = x(n) x
(n).

(18)

4) Take a JN -point DFT operation to w to obtain the

frequency-domain clipping noise W = [W (0), W (1),
. . . , W (N 1)].
5) Set Cs (n) = W (n), if n R and Cs (n) M; otherwise,
Cs (n) = 0. Then, the constellation adjustment symbol
Cs can be obtained.
B. Sidelobe Suppression
During each iteration, after we solve the first problem and
obtain the appropriate Cs , we then solve the second problem.
For the second problem, the constellation adjustment symbol
Cs obtained from (P1) is seen as a constant, and the SC symbol
Cp on PU subcarriers is utilized as a variable to suppress the
sidelobe in NC-OFDM-based CR systems. Thus, the second
problem can be expressed as
(P2) :

min Pd X + Pd Cs + Pd Cp 2 .
Cp

(19)

 p as an L 1 vector, which squeezes out zero from

Denote C
the N 1 vector Cp and only leaves the weights on the position
 p to Cp , i.e.,
Rc . Moreover, T is a matrix that transfers C
 p.
Cp = T C

(20)

NI et al.: PAPR REDUCTION AND SIDELOBE SUPPRESSION USING SC IN NC-OFDM-BASED CR SYSTEMS

Thus, the closed-form solution of (P2) can be expressed as

1

 p = (BH B) BH Pd (X + Cs )
C

(21)

where B = Pd T, and BH denotes the complex conjugate

transpose matrix of B.
Substituting (20) to (21), we have
1

Cp = T(BH B) BH Pd (X + Cs )
= Q(X + Cs )

(22)

where Q = T(BH B)1 BH Pd .

Thus, we obtain the total added symbol C as
C = Cp + C s .

969

For (P2), the N N matrix Q needs to be calculated only

once in advance, and it is stored in memory. Then, Q does
not need to be calculated during each iteration. Moreover,
during each iteration, Cp is calculated according to (22), which
requires one matrix multiplication. Thus, the complexity of
solving (P2) is O(N 2 ) in the worst case. In summary, the
computational complexity of the sub-SC method is O(N 2 +
2JN log JN ) in the worst case.
As discussed in Section III, the computational complexity of
the SC method is O(N 3 + JN log JN ). Therefore, compared
with the SC method, the sub-SC method has much lower
complexity.

(23)

C. Power Constraint
Since the sub-SC method also increases the power of the
NC-OFDM signals, the sub-SC method should follow the
power constraint as shown in (15e). To maintain the power constraint (15e) for the sub-SC method, the iteration is terminated
whenever (15e) is not satisfied. This way, the sub-SC method
can jointly reduce the PAPR and suppress the sidelobe power
while maintaining the power constraint.
Therefore, the iterative sub-SC method can be summarized
in Algorithm 1.
Algorithm 1: The sub-SC method
1: Set the maximum iteration number K and the clipping
threshold A.
2: Set i = 0, and for the i = 0th iteration, set C0s = 0,
C0p = 0, and C0 = 0. Moreover, randomly generate the
original data block X = [X(0), X(1), . . . , X(N 1)] as
discussed in (1).
3: Take the SC symbol Cip on PU subcarriers as a constant
and obtain the constellation adjustment symbol Ci+1
by
s
solving (P1) as discussed in Section IV-A.
on SU
4: Take the constellation adjustment symbol Ci+1
s
subcarriers as a constant and obtain the SC symbol Ci+1
p
by (22).
5: Update the total SC symbol Ci+1 by (23) and calculate the total power X + Ci+1 22 . If the power con=
straint X + Ci+1 22  is not satisfied, output x
F(X + Ci+1 ) as the transmitted signal and terminate the
algorithm.
6: Set i = i + 1, if i < K, go to Step 3; Otherwise, take
a JN -point IDFT operation to X + Ci+1 to obtain the
 and terminate the algorithm.
transmitted signal x

D. Complexity Analysis
The computational complexity of the sub-SC method mainly
depends on the cost of solving (P1) and (P2). For (P1), during
each iteration, one IDFT operation and one DFT operation are
needed; thus, the complexity of solving (P1) can be expressed
as O(2JN log JN ).

V. S IMULATION R ESULTS
Here, some simulations have been conducted to evaluate
the capability of the proposed SC and sub-SC methods, including the PAPR reduction and the sidelobe suppression performances, where the NC-OFDM-based CR system is considered
with N = 64 subcarriers, and the subcarrier interval is f =
22.5 kHz. Moreover, the target spectrum band occupied by PUs
is from 29f to 38f , in which 37 frequency sampling points
for the evaluation of the total sidelobe power are placed with the
equivalent space of (f )/4. Thus, a = 29, b = 38, M = 37,
and the remaining spectrum band is utilized by SUs. In addition,
104 independent NC-OFDM symbols modulated by 4-QAM
and 16-QAM are randomly generated with the oversampling
factor J = 2, and the Welch method with Blackman window is
used to calculate the power spectrum density of the transmitted
signals. Moreover, for the proposed SC and sub-SC methods,
the increased power threshold is set to be  = (1 + )X22 in
the simulations, where is the parameter to adjust the power
threshold . For comparison, we also conduct simulations using
the conventional method, which simply turns off the subcarriers
in the target band, the AIC method [16], and the SLM method
[21], respectively. In the simulations, the SC method solves the
convex optimization problem (15) using the public software
CVX [26].
Fig. 4 shows the time-domain signal of the SC method with
= 0.5 and = 0.5 during one NC-OFDM symbol duration
when 4-QAM is employed. For the turning-off method, SLM
method, and SC method, the peak amplitude is 3.25, 3.00, and
2.51, respectively. Thus, compared with the turning-off method,
the AIC and SC methods can reduce the PAPR by 0.70 and
2.24 dB, respectively. Therefore, the proposed SC method
offers better PAPR reduction performance than the turning-off
and SLM methods. Moreover, the SC method does not need to
transmit side information to the receiver; as a result, the data
rate of the NC-OFDM-based CR system can be maintained.
Fig. 5 shows the PAPR reduction of the proposed SC method
with = 0.5 when is different for 4-QAM. As shown in
Fig. 5, the AIC method has the same PAPR as the turningoff method. Moreover, when CCDF = 103 , for = 0.1, 0.2,
0.3, 0.4, 0.5, and 0.6, the PAPR of the proposed SC method
is 10.0, 9.2, 8.3, 7.7, 7.3, and 7.1 dB, respectively. Therefore,
the proposed SC method can significantly reduce the PAPR of
the transmitted NC-OFDM signals, compared with the turningoff and AIC methods. Moreover, it is obvious in Fig. 5 that

970

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 64, NO. 3, MARCH 2015

Fig. 6. PAPR reduction with the SC method when = 0.5 and is different
for 4-QAM.
Fig. 4. Amplitude of time-domain signal generated by the SC method with
= 0.5 and = 0.5 during one NC-OFDM symbol duration when 4-QAM is
employed.

Fig. 5. PAPR reduction with the SC method when = 0.5 and is different
for 4-QAM.

Fig. 7. Power spectrum density of the NC-OFDM signals with the SC method
when = 0.5 and is different for 4-QAM.

the PAPR reduction improves as increases, i.e., the PAPR

reduction of the SC method improves when the power of the
SC method increases.
Fig. 6 shows the PAPR reduction of the proposed SC method
with = 0.5 when is different for 4-QAM. As shown in
Fig. 6, the SLM method can slightly reduce the PAPR of the
transmitted NC-OFDM signals, and its PAPR is 9.1 dB when
CCDF = 103 . The proposed SC method with different
values also achieves significant PAPR reduction. For example,
when CCDF = 103 , for = 0.2, 0.3, 0.4, and 0.5, the PAPR
of the proposed SC method is 10.2, 9.5, 8.3, and 7.3 dB,
respectively. Moreover, the PAPR reduction performance of the
SC method improves when the value of increases. Because
there is a tradeoff between PAPR reduction and sidelobe suppression, and the sidelobe suppression constraint becomes more
relaxed with the increase of , this leads to the improvement

of the PAPR reduction performance and the degradation of the

sidelobe suppression performance.
Fig. 7 shows the normalized power spectrum of the NCOFDM signals with the proposed SC method when = 0.5
and is different for 4-QAM. As shown in Fig. 7, the turningoff method can achieve a spectrum notch of only 13 dB in the
target spectrum band from 29f to 38f , which means a considerable interference to the PUs. The AIC method can obtain
a spectrum notch that is 7 dB deeper than that of the turningoff method. In addition, the SLM method can achieve the same
sidelobe suppression performance with the AIC method. Moreover, it is obvious that the sidelobe suppression performance
of the proposed SC method is much better than those of the
turning-off, AIC, and SLM methods. For example, compared
with the turning-off method, when = 0.2, 0.3, 0.4, and 0.5, the
SC method can improve the sidelobe suppression by 20, 17, 12,

NI et al.: PAPR REDUCTION AND SIDELOBE SUPPRESSION USING SC IN NC-OFDM-BASED CR SYSTEMS

971

Fig. 8. PAPR reduction with the SC method when = 1 and is different

for 16-QAM.

Fig. 10. PAPR reduction with the sub-SC method when 4-QAM is employed.

Fig. 9. Power spectrum density of the NC-OFDM signals with the SC method
when = 1 and is different for 16-QAM.

Fig. 11. Power spectrum density of the NC-OFDM signals with the sub-SC
method when 4-QAM is employed.

and 9 dB, respectively. Thus, as shown in Figs. 6 and 7, the SC

method can achieve much better PAPR reduction and sidelobe
suppression performances than the turning-off, AIC, and SLM
methods. Furthermore, the sidelobe suppression performance
of the SC method improves when the value of decreases,
since there is a tradeoff between PAPR reduction and sidelobe
suppression. Therefore, the SC method can provide both sufficient PAPR reduction and sidelobe suppression performances
by adjusting parameters and .
Fig. 8 shows the PAPR reduction of the proposed SC method
with = 1 when is different for 16-QAM. As shown in
Fig. 8, the PAPR of the SLM method is 9.1 dB when CCDF =
103 . Moreover, when CCDF = 103 , for = 0.4 and 0.5,
the PAPR of the proposed SC method is 9.9 and 8.5 dB,
respectively. Thus, when 16-QAM is employed, the SC method
with = 1 and = 0.5 can offer better PAPR reduction performance than the turning-off, AIC, and SLM methods.

Fig. 9 shows the normalized power spectrum of the NCOFDM signals with the proposed SC method when = 1 and
is different for 16-QAM. As shown in Fig. 9, the SLM method
can achieve a spectrum notch that is 8 dB deeper than that of
the turning-off method. Moreover, it is obvious that the sidelobe
suppression performance of the proposed SC method is better
than those of the turning-off, AIC, and SLM methods. For
example, compared with the turning-off method, when = 0.4
and 0.5, the SC method can improve the sidelobe suppression
by 12 and 9 dB, respectively.
Fig. 10 shows the PAPR reduction comparison of the sub-SC
method with the turning-off, AIC, SLM, and SC methods with
= 0.5. In addition, the maximum iteration number of the subSC method is K = 6. As shown in Fig. 10, the sub-SC method
can provide better PAPR reduction than that of the turning-off,
AIC, and SLM methods. For example, when CCDF = 103 ,
the PAPR of the turning-off, AIC, SLM, sub-SC with = 0.3,

972

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 64, NO. 3, MARCH 2015

and sub-SC with = 0.6 methods is 10.5, 10.5, 9.1, 8.4, and
7.9 dB, respectively. Thus, the sub-SC method can also provide
significant PAPR reduction. Although the PAPR reduction of
the sub-SC method is a bit worse than the SC method, the subSC method offers lower computational complexity than the SC
method. Therefore, the sub-SC method is efficient and offers
significant PAPR reduction with low computational complexity.
Fig. 11 shows the normalized power spectrum of the NCOFDM signals with the turning-off, SC with = 0.5, and subSC methods, respectively, and the maximum iteration number
is K = 6 for the sub-SC method. In Fig. 11, it is obvious that
the sub-SC method can provide better sidelobe suppression
performance than the turning-off method, and it can improve
the sidelobe suppression by 5 dB. Although the sidelobe suppression performance of the sub-SC method is not as good as
that of the SC method, the sub-SC method has much lower
computational complexity than the SC method. Therefore, the
sub-SC method is an efficient method for joint PAPR reduction
and sidelobe suppression with low computational complexity.
VI. C ONCLUSION
In this paper, we have proposed an SC method for joint
PAPR reduction and sidelobe suppression in NC-OFDM-based
CR systems. The proposed SC method dynamically extends
part of constellation points on the SU tones and adds several
SC symbols on the PU tones to jointly reduce the PAPR and
suppress the sidelobe of NC-OFDM signals. Moreover, we also
proposed a sub-SC method to efficiently reduce the PAPR and
suppress the sidelobe with low computational complexity. Simulation results show that both the SC and sub-SC methods can
provide significant PAPR reduction and sidelobe suppression
performances.
R EFERENCES
[1] S. Haykin, Cognitive radio: Brain-empowered wireless communications, IEEE J. Sel. Areas Commun., vol. 23, no. 2, pp. 201220,
Feb. 2005.
[2] Y. Zou, Y. Yao, and B. Zheng, Cooperative relay techniques for cognitive
radio systems: Spectrum sensing and secondary user transmissions, IEEE
Commun. Mag., vol. 50, no. 4, pp. 98103, Apr. 2012.
[3] X. Huang, G. Wang, and F. Hu, Multitask spectrum sensing in cognitive radio networks via spatiotemporal data mining, IEEE Trans. Veh.
Technol., vol. 62, no. 2, pp. 809823, Feb. 2013.
[4] Z. Li and X.-G. Xia, Time-domain interference cancellation for
ALAMOUTI-coded cooperative OFDM systems with insufficient CP,
IEEE Trans. Veh. Technol., vol. 59, no. 6, pp. 31313136, Jul. 2010.
[5] D. Qu, Z. Wang, and T. Jiang, Extended active interference cancellation for sidelobe suppression in cognitive radio OFDM systems with
cyclic prefix, IEEE Trans. Veh. Technol., vol. 59, no. 4, pp. 16891695,
May 2010.
[6] J.-C. Chen, M.-H. Chiu, Y.-S. Yang, and C.-P. Li, A suboptimal tone
reservation algorithm based on cross-entropy method for PAPR reduction
in OFDM systems, IEEE Trans. Broadcast., vol. 57, no. 3, pp. 752756,
Sep. 2011.

[7] J.-C. Chen and C.-P. Li, Tone reservation using near-optimal peak
reduction tone set selection algorithm for PAPR reduction in OFDM
systems, IEEE Signal Process. Lett., vol. 17, no. 11, pp. 933936,
Nov. 2010.
[8] T. Jiang, C. Li, and C. Ni, Effect of PAPR reduction on spectrum and
energy efficiencies in OFDM systems with Class-A HPA over AWGN
channel, IEEE Trans. Broadcast., vol. 59, no. 3, pp. 513519, Sep. 2013.
[9] D. Qu, J. Ding, T. Jiang, and X. Sun, Detection of non-contiguous
OFDM symbols for cognitive radio systems without out-of-band
spectrum synchronization, IEEE Trans. Wireless Commun., vol. 10,
no. 2, pp. 693701, Feb. 2011.
[10] I. Gutman, I. Iofedov, and D. Wulich, Iterative decoding of iterative
clipped and filtered OFDM signal, IEEE Trans. Commun., vol. 61, no. 10,
pp. 42844293, Oct. 2013.
[11] L. Li and D. Qu, Joint decoding of LDPC code and phase factors for
OFDM systems with PTS PAPR reduction, IEEE Trans. Veh. Technol.,
vol. 62, no. 1, pp. 444449, Jan. 2013.
[12] T. Jiang and C. Li, Simple alternative multi-sequences for PAPR
reduction without side information in SFBC MIMO OFDM systems,
IEEE Trans. Veh. Technol., vol. 61, no. 7, pp. 33113315, Sep. 2012.
[13] B. Krongold and D. Jones, PAR reduction in OFDM via active constellation extension, IEEE Trans. Broadcast., vol. 49, no. 3, pp. 258268,
Sep. 2003.
[14] H. Li, T. Jiang, and Y. Zhou, An improved tone reservation scheme with
fast convergence for PAPR reduction in OFDM systems, IEEE Trans.
Broadcast., vol. 57, no. 4, pp. 902906, Dec. 2011.
[15] Y. Wang, W. Chen, and C. Tellambura, Genetic algorithm based nearly
optimal peak reduction tone set selection for adaptive amplitude clipping
PAPR reduction, IEEE Trans. Broadcast., vol. 58, no. 3, pp. 462471,
Sep. 2012.
[16] H. Yamaguchi, Active interference cancellation technique for MBOFDM cognitive radio, in Proc. 34th Eur. Microw. Conf., Oct. 2004,
vol. 2, pp. 11051108.
[17] D. Li, X. Dai, and H. Zhang, Sidelobe suppression in NC-OFDM systems
using constellation adjustment, IEEE Commun. Lett., vol. 13, no. 5,
pp. 327329, May 2009.
[18] R. Xu, H. Wang, and M. Chen, On the out-of-band radiation of DFTbased OFDM using pulse shaping, in Proc. IEEE Int. Conf. Wireless
Commun. Signal Process., Nov. 2009, pp. 15.
[19] W.-C. Chen and C.-D. Chung, Spectral precoding for cyclic-prefixed
OFDMA with interleaved subcarrier allocation, IEEE Trans. Commun.,
vol. 61, no. 11, pp. 46164629, Nov. 2013.
[20] J. Zhang, X. Huang, A. Cantoni, and Y. Guo, Sidelobe suppression with
orthogonal projection for multicarrier systems, IEEE Trans. Commun.,
vol. 60, no. 2, pp. 589599, Feb. 2012.
[21] A. Ghassemi, L. Lampe, A. Attar, and T. A. Gulliver, Joint sidelobe and
peak power reduction in OFDM-based cognitive radio, in Proc. IEEE
72nd Veh. Technol. Conf. Fall, Sep. 2010, pp. 15.
[22] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge, U.K.:
Cambridge Univ. Press, 2004.
[23] Y.-C. Wang and Z.-Q. Luo, Optimized iterative clipping and filtering for
PAPR reduction of OFDM signals, IEEE Trans. Commun., vol. 59, no. 1,
pp. 3337, Jan. 2011.
[24] M. Lobo, L. Vandenberghe, S. Boyd, and H. Lebret, Applications
of second-order cone programming, Linear Algebra Appl., vol. 284,
no. 13, pp. 193228, Nov. 1998.
[25] A. Gatherer and M. Polley, Controlling clipping probability in DMT
transmission, in Proc. 31st Asilomar Conf. Signals, Syst., Comput.,
Nov. 1997, vol. 1, pp. 578584.
[26] M. Grant and S. Boyd, CVX: Matlab Software for Disciplined Convex
Programming (Web Page and Software), Oct. 2013. [Online]. Available:
http://cvxr.com/cvx

Authors photographs and biographies not available at the time of publication.