A Reduced Complexity Subcarrier Switching

Scheme for PAPR Reduction in OFDM System

Sabbir Ahmed1 and Makoto Kawai2

1
Graduate School of Science and Engineering, Ritsumeikan University,
1-1-1 Nojihigashi, Kusatsu city, Japan
[email protected]
2
Graduate School of Science and Engineering, Ritsumeikan University,
1-1-1 Nojihigashi, Kusatsu city, Japan
[email protected]

Abstract. High values of Peak to Average Power Ratio or PAPR is

a fundamental problem of Orthogonal Frequency Division Multiplexing
(OFDM) systems. Literature survey shows that recent research works on
PAPR reduction are proposing schemes focusing on IEEE standard based
systems. One such technique is the null and data subcarrier switching
method- a scheme that can reduce PAPR without any need of side infor-
mation, causes no distortion and can complement other existing methods.
However, the associated computational complexity of this method can be
very high. In this paper, we propose a modied technique based on the
subcarrier switching method that can reduce computational complexity
to a great extent with very little or no loss in the PAPR reduction. In ad-
dition, we also present further PAPR reduction capability of our scheme
with low computational overhead by making compromise on the no-side
information advantage.

Keywords: OFDM, PAPR, Subcarrier switching.

1 Introduction
Orthogonal Frequency Division Multiplexing (OFDM) is a well known tech-
nique for high speed data communication. OFDM exhibits robust performance
in combating inter-symbol interference (ISI) and also gives very good spectrum
eciency. Exploiting these advantages, multiple access schemes like Orthogo-
nal Frequency Division Multiplexing Access (OFDMA) and Multicarrier Code
Division Multiple Access (MC-CDMA) have been proposed. MC-CDMA and
OFDMA both have strong potentials of becoming future generation de-facto
multiple access scheme. In fact, OFDM is being used in various contemporary
communication systems including digital audio and video broadcasting (DAB
and DVB) [1],[2] and wireless LANs [3], whereas OFDMA has been standard-
ized for the PHY layer of IEEE 802.16 wireless metropolitan area networks and
the downlink of 3GPP-LTE [4],[5].
Besides all its advantages, OFDM comes with a fundamental problem of gen-
erating occasional very high peaks in its transmit signal which is often quantied

T.-h. Kim et al. (Eds.): FGCN 2011, Part I, CCIS 265, pp. 6776, 2011.

c Springer-Verlag Berlin Heidelberg 2011
68 S. Ahmed and M. Kawai

by the parameter called Peak to Average Power Ratio or PAPR. As power am-
plier are generally operated near the saturation point, if PAPR is high, e.g.,
the peak power of the signal crosses the operating range of the amplier, it may
get driven well into the saturation region causing severe bit error ratio (BER)
degradation due to non-linear amplication. On the other hand, increasing op-
erating range of the amplier or operating it with high back-os is not practical
for many power constrained devices, specially for hand-held mobile terminals.
Hence PAPR reduction is a crucial issue for any system that uses OFDM.

Transmitter

Serial Parallel
PAPR
Base band To Subcarrier To
IFFT Estimator HPA
Modulation Parallel Switching Serial

Noisy
Channel
Receiver

Parallel Serial
Subcarrier
Base band To To
De- FFT
demodulation Serial Parallel
switching

Fig. 1. OFDM system block diagram with subcarrier switching

Literature survey shows that the investigation of PAPR reduction has been a
very active eld of research for the last couple of years and quite a large number
of techniques have been proposed. Amongst them, some of the fundamental and
generic proposals for PAPR reduction include signal clipping, selected mapping
(SLM) and partial transmit sequences (PTS) etc. [6]-[8]. While signal clipping
is a very simple and eective PAPR reduction technique, it can exert a negative
impact on the spectral containment of the transmitted signal and thus cause
inter-channel interference (ICI) [9]. The PTS and SLM techniques are distortion
less methods but they require transmission of extra information known as side
information to the receiver, which decreases the data throughput.
Recently, other than considering generalized architecture, PAPR reduction
investigations are focusing on systems that are based on dierent standard spec-
ications, e.g., WiMaX (IEEE 802.16) or wireless LAN (802.11) [10]-[12]. In
particular, the scheme described in [11] known as the null and data subcarrier
switching method shows signicant PAPR reduction by switching data subcar-
riers with null subcarriers specied in the IEEE802.11a standard specications
[3]. Unlike signal clipping, it does not distort the transmission signal and also
there is no need for side information transmission as required by SLM or PTS
techniques. But the problem of this method is its high computational overhead.
 Reduced Complexity Subcarrier Switching Scheme for PAPR Reduction 69

In this contribution, we propose a modied technique of null and data subcar-

rier switching which requires much less computations compared to the original
method [11] and yet achieves almost same level of PAPR reduction. Then we
extend our method of low computational complexity for further PAPR reduction
by making a compromise on the advantage of no-side information. We show that
to achieve similar PAPR reduction, the original method with no-side information
requirement is almost impractical due to its extremely high level of computa-
tional overhead. Finally, we take into consideration a non-linear power amplier
and a noisy channel and demonstrate the comparative BER performance of our
method with the original one.

Null 0 0
#1 1 1
#2 2 2

#26 26 26
Null 27 27
Frequency domain input Null IFFT Time domain output
Null 37 37
#-26 38 38

#-2 62 62
#-1 63 63

Fig. 2. Inputs and outputs of IFFT

We organize the rest of the paper as follows. In the following section, we

show and discuss the system model that we have considered. Detail discussion of
our method is given in section 3. Section 4 deals with the performance analysis
results and related discussions. And nally in section 5, we make some concluding
statements.

2 System Model
Figure 1 shows the model of an OFDM system where blocks with solid lines
and dashed lines represent the fundamental components and the components
for PAPR reduction scheme respectively. In this model, at rst binary random
input data is baseband modulated that generates input symbol given by x[i] =
x[0], x[1], ..., x[N 1]. They are then converted from serial to parallel and fed
into the IFFT module. The IFFT module performs the task of multicarrier
modulation. Output of the IFFT is parallel to serial converted and the resultant
discrete time domain symbols is given by Eqn. 1.
N 1
1  2ni
X[n] = x[i]ej N , 0 n N 1 (1)
N i=0
70 S. Ahmed and M. Kawai

14

12

r = inf.
10

8
r=6
Output

r=2
4

0
0 2 4 6 8 10 12 14
Input

Fig. 3. Non-linear power amplier input-output response

The time domain signal is then amplied by a high power amplier (HPA) and
transmitted to the noisy wireless channel. The receiver perform the opposite
operations and nally detects the transmitted symbols.
Now, for PAPR reduction purpose a subcarrier switching module along with
PAPR estimator block are inserted in the transmitter whereas a subcarrier de-
switching module is placed in the receiver. The subcarrier switching module
performs the operation of switching between null and data subcarriers. We de-
scribe in detail its operation in the following section. The PAPR estimator block
determines and stores the PAPR value of all the switching combinations and
transmits the signal with the lowest PAPR. PAPR in dB is expressed by the
following Eqn.

max|X[n]|2
P AP R(dB) = 10 log10 (2)
E [|X[n]|2 ]
where E[] denotes expectation.
As mentioned before, our PAPR reduction scheme is based on the IEEE
802.11a WLAN standard [3]. This standard species a total of 64 subcarriers
that includes 12 null subcarriers. The remaining 52 subcarriers are used for
data (48 subcarriers for user data and 4 subcarriers for pilot data) transmission.
Among the 12 null subcarriers, 6 subcarriers work as guard-band at the low-
frequency edge of the spectral band of the subcarriers while 5 subcarriers serve
as guard-band at the high-frequency edge. The remaining one is placed at the
middle to avoid direct current energy. The IFFT input-output lay out is shown
in Fig. 2 [3].
Finally, for simulating the HPA, we consider the Solid State Power Amplier
(SSPA) model given in [13] where the AM-to-AM conversion characteristics is
expressed by Eqn. 3.
x
F [x] = (3)
[1 + (x/A)2r ]1/2r
 Reduced Complexity Subcarrier Switching Scheme for PAPR Reduction 71

where x is the amplitude of the input signal, A is the saturated output level and
r is the non-linearity level. This model only considers AM-to-AM non-linearity.
The parameter r can be used to tune the level of non-linearity and it is depicted
in Fig. 3. A large value of r turns this amplier into a linear one where as very
small values make it behave as a simple clipping amplier.

Step: 1 Select {N C1 , N C2 , .., N CP }, P number of null subcarriers

out of available L for switching with total available N L data
subcarriers.
Step: 2 Construct data subcarrier combination taking P number of
subcarriers at a time from data subcarrier index vector of length
N L.
Step: 3 Switch the P null subcarrier with data subcarriers generated
in Step 2 such that the order of null subcarrier indices and data
subcarrier indices are same.
Step: 4 Apply IFFT for OFDM operation and calculate PAPR from
the time domain signal.
N L
Step: 5 Repeat Step 2 to Step 4, CP 1 times for all other possible
subcarrier combinations and record the lowest PAPR.
Step: 6 Apply the data subcarrier combination corresponding to the
lowest PAPR as switched data subcarriers for signal transmission.

Fig. 4. Original null and data subcarrier switching

3 PAPR Reduction Method

The main contribution of this paper is a proposal to reduce the computational

overhead involved in searching low PAPR producing data subcarrier switching
positions proposed in [11]. The additional components required to implement
the switching technique is marked by dashed lines in Fig. 1 and we explain the
working principle here. If, we consider an N -size IFFT module with L number
of total null subcarriers, i.e., {N C1 , N C2 , .., N CL }, as per [11], once the number
of null subcarriers for switching is determined, say P , the null subcarriers for
switching {N C1 , N C2 , .., N CP } are needed to be switched with all possible data
subcarrier indices combinations for nding the lowest PAPR. And the underlying
constraint is that the order of data subcarrier and null subcarrier indices must be
same so that no side information is necessary for the receiver to perform the de-
switching operation. As a result CPN L number of search operations are involved
for every input data block, which in turn means that many IFFT operations are
required. For example, a system with N = 64 and L = 12, if P = 4 is considered,
C452 = 270725 number of search operations are required. The time involved in
such calculation may make this technique impractical for many delay sensitive
applications.
On this basis, we were motivated to consider modied approach for searching
that can reduce computational burden without making signicant sacrice in
PAPR reduction. In our proposed method, we apply incremental search by con-
sidering one null-data subcarrier switching at a time. For example, for P = 2,
72 S. Ahmed and M. Kawai

we at rst only consider N C1 and search C1N L , i.e., N L times to look for
the data subcarrier position, say D1 yielding lowest PAPR when switched with
N C1 . After this, we apply the same operation for P2 . But in order to keep the
order of null subcarriers and switched data subcarriers, we rst remove all the
data subcarriers positions D1 from the search space. Thus the size of the data
subcarrier search space for N Ci , i 1 is dependent on the outcome of search
with N Ci1 . Step-by-step by descriptions of the original null and data subcarrier
switching and our proposed methods are listed in Figs. 4 and 5 respectively.

Step: 1 Select {N C1 , N C2 , .., N CP }P number of null subcarriers

out of available L for switching with total available N L data
subcarriers and set i = 1.
Step: 2 Select a data subcarrier from data subcarrier index vector.
Step: 3 Switch the the null subcarrier N Ci with the data subcarrier
generated in step 2.
Step: 4 Apply IFFT for OFDM operation and calculate PAPR from
the time domain signal.
Step: 5 Repeat Step 2 to Step 4 for for all other data subcarriers one
at a time, record the lowest PAPR and mark the corresponding
data subcarrier as DCi .
Step: 6 Remove all data subcarrier indices DCi from data subcar-
rier index vector and increase i by 1.
Step: 7 Repeat Step 2 to Step 6 P 1 times.
Step: 8 Apply the data subcarrier combination
{DC1 , DC2 , .., DCP } as switched data subcarriers for sig-
nal transmission.

Fig. 5. Proposed null and data subcarrier switching

The computational complexity reduction is depicted in Table 1. As we can

see, for P = 2, maximum computations required by our method is only about
7.76% of the original one, i.e., a reduction of around 92.24%. Again for the
original method, with higher values of P , one expects more reduction in PAPR.
But at the associated computation cost can be extremely high. For example,
as mentioned earlier, for P = 4, 270725 number of searching is required which
makes it almost unfeasible for practical implementation. As a matter of fact if
signicant PAPR reduction can be achieved with P = 2, higher values of P may
not be required. Thus primarily we compare our method with the original one
for P = 2.

Table 1. Comparative computational complexity

No. of switched null No. of Searching

subcarriers, P Original Proposed (Maximum)
2 1326 103
4 270725 202
6 20358520 207
8 752538150 388
 Reduced Complexity Subcarrier Switching Scheme for PAPR Reduction 73

Our method with P = 4 is also computationally ecient compared to the

original one for higher values of P . But due to reduction of the search space for
every increment in P (Step 6), there remains a possibility of the search space get-
ting very limited for Pi , i 1. This problem becomes very much probable when
P 2. In order to avoid this problem, we propose another searching where we
relax the need for keeping the order of the switched null and data subcarriers
by modifying step 6 as below,
Step: 6 Remove data subcarrier index DCi from data subcarrier index vector and increase i by 1.

Table 2. Simulation parameters

Modulation scheme BPSK

IFFT size 64
Number of data subcarriers 52
Number of switched null subcarriers 2, 4
Total number of OFDM symbols 104
Oversampling factor 4
HPA Model SSPA
HPA level of non-linearity, r 2
Channel model AWGN

This eventually also removes the constraint of switched null subcarriers and
data subcarriers having the same order. As a result the receiver can not perform
the opposite operation of switching unless it has some extra information. Thus
the need for side information comes into eect. We think this trade-o is worthy
of, since some side information (e.g. control signal) is always needed irrespective
of whether any PAPR reduction scheme is applied or not.

4 Performance Evaluation
Our simulation is based on the IEEE802.11a standard specication. However,
without distinguish between user data and pilot data subcarriers, we used all the
52-not null subcarriers for switching with null subcarriers. In contrast, one may
recall, 48 data subcarriers was used in [11]. We considered BPSK modulation
and investigated scenarios for both P = 2 and P = 4. For accurate estimation
of PAPR, we considered an over sampling factor of 4 [14]. All the pertinent
simulation parameters are listed in Table 2.
Figure 6 shows the complementary cumulative distribution function (CCDF)
of PAPR without any PAPR reduction scheme, the original null and data sub-
carrier switching scheme and our proposed scheme, the later two with P = 2.
In this gure, we refer to the original null and data subcarrier switching as SC
Switching and our method as Reduced complexity (No side info required)
respectively. To make a fare comparison, we selected the null subcarriers in-
dexed at 27 (pls. refer to Fig. 2) as was done in [11]. As can be seen from
74 S. Ahmed and M. Kawai

1
No PAPR Reduction
0.9 SC Switching
Reduced Complexity (No side info reqd.)
0.8

0.7
Pr[PAPR >= Abcissa]
0.6

0.5

0.4

0.3

0.2

0.1

2 3 4 5 6 7 8 9 10 11 12
PAPR(dB)

Fig. 6. Comparative PAPR distribution for P = 2

Fig. 6, compared to no PAPR reduction scheme, our method achieves signicant

PAPR reduction which is slightly inferior to the original method. For exam-
ple the probability of PAPR equalling or exceeding 6dB is around 0.6 when no
PAPR reduction scheme is applied. Compared to this, the corresponding prob-
abilities of our proposed method and the original method are approximately 0.2
and 0.18 respectively. Again in other terms, with a 0.2 probability, our method
and the original method shows PAPR of approximately 6dB and 5.9dB respec-
tively. It means our method shows a 1.66% degradation in PAPR reduction. This
very little loss is more than compensated when we consider 92.24% reduction in
computational complexity as was mentioned before.
In Fig. 7, we show the result with P = 4. Here, we selected null subcarriers
positioned at 27, 28 for switching. Our methods of both requiring side-info
and no side-info are shown. If we sacrice some throughput by sending small
amount of side-info, our method can achieve good PAPR reduction with very
little computations compared to the original method. Here, we need to send
the switched data subcarrier index to the receiver with which it will exchange
the values of the null subcarrier. Since log2 64 = 6 bits is required to represent
a subcarrier index, four 6 bit words will be needed for this purpose. In this
gure we can see with a 0.2 probability, our method (side info reqd.) and the
original method shows PAPR of approximately 5dB and 4.7dB. And the associ-
ated number of searchings are 270725 and 202 (maximum) respectively. Which
means our method shows 6% higher PAPR and requires additional side-info but
reduces 99% computational complexity.
Finally, in Fig. 8, the BER result for our proposed system (P = 2) under the
inuence of AWGN channel and non-linear HPA is depicted. We x the non-
linearity level r = 2 and the saturation power level at 3dB. This result shows
the combined eect of the noise contamination by the wireless channel and the
 Reduced Complexity Subcarrier Switching Scheme for PAPR Reduction 75

1
No PAPR Reduction
0.9 SC Switching
Reduced Complexity (No side info reqd.)
0.8 Reduced Complexity (Side info reqd.)

0.7
Pr[PAPR >= Abcissa]
0.6

0.5

0.4

0.3

0.2

0.1

2 3 4 5 6 7 8 9 10 11 12
PAPR(dB)

Fig. 7. Comparative PAPR distribution for P = 4

0
10
No PAPR Reduction
SC Switching
Reduced Complexity (No side info reqd.)

1
10
BER

2
10

3
10

4
10
0 5 10 15
Eb/No(dB)

Fig. 8. Comparative BER performance for P = 2

non-linear amplication due to high PAPR. As we can see from this gure,
our proposed method shows almost similar BER as the original one. And this
performance is better compared to OFDM system with no-reduction strategy
applied.

5 Conclusions
Null and data subcarrier switching scheme is a very recent technique of PAPR
reduction which appears to be an eective solution for IEEE 802.11a OFDM
system. But the associated computational overhead can be very high specially
when higher number of null subcarriers are used for switching. In this paper, we
presented a modied technique for searching the data subcarriers positions to be
76 S. Ahmed and M. Kawai

switched with null subcarrier with reduced complexity. We demonstrated that

when 2 null subcarriers are used for switching, the proposed method achieves
very high computational complexity reduction yet achieves similar PAPR com-
pared to the original method. And the associated BER performance is also very
good. For higher number of null subcarriers, we compromise on the no-side infor-
mation requirement to achieve acceptable PAPR with very low computational
overhead.

References
1. ETSI, Radio broadcasting systems: Digital Audio Broadcasting (DAB) to mobile,
portable and xed receivers, ETS 300 401 ed.2 (1999)
2. ETSI, Digital video broadcasting systems: Framing structure, channel coding, and
modulation for digital terrestrial television. European Telecommunications Stan-
dard, EN 300744 (1997)
3. IEEE, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer
(PHY) specications. High Speed Physical Layer in the 5GHz Band, IEEE Std
802.11a (1999)
4. IEEE, Part 16: Air Interface for Fixed Broadband Wireless Access Systems. IEEE
std. 802.16-2004 (Revision of IEEE Std. 802.16-2001) (2001)
5. http://www.3gpp.org
6. Li, X., Cimini, L.J.: Eects of clipping and ltering on the performance of OFDM.
IEEE Communications Letters 2(5), 131133 (1998)
7. Bauml, R.W., Fischer, R.F.H., Huber, J.B.: Reducing the peak-to-average power
ratio of multicarrier modulation by selected mapping. IEE Electronics Let-
ters 32(22), 20562057 (1996)
8. Cimini, L.J., Sollenberger, N.R.: Peak-to-average power ratio reduction of an
OFDM signal using partial transmit sequences. IEEE Communications Let-
ters 4(3), 8688 (2000)
9. Thompson, S.C., Proakis, J.G., Zeidler, J.R.: The eectiveness of signal clipping
for PAPR and total degradation reduction in OFDM systems. In: Proc.: IEEE
Global Telecommunications Conference, GLOBECOM 2005, St. Louise, Montana,
USA, pp. 28072811 (2006)
10. Chaokuntod, S., Uthansakul, P., Uthansakul, M.: Fast Dummy Sequence Insertion
method for PAPR reduction in WiMAX systems. In: Proc.: Asia Pacic Microwave
Conference, Singapore, pp. 21682171 (2010)
11. Wong, K.T., Wang, B., Chen, J.-C.: OFDM PAPR reduction by switching null
subcarriers and data-subcarriers. Electronics Letters 47(1), 6263 (2011)
12. Hussain, S., Guel, D., Louet, Y., Palicot, J.: Performance comparison of PRC
based PAPR reduction schemes for WiLAN Systems. In: Proc.: European Wireless
Conference, Aalborg, Denmark, pp. 167172 (2009)
13. Rapp, C.: Eects of HPA-nonlinearity on a 4-DPSK/OFDM signal for a digital
sound broadcasting system. In: Proc.: Second European Conference on Satellite
Communications, Liege, Belgium, pp. 179184 (1991)
14. Tellambura, C.: Computation of the Continuous-Time PAR of an OFDM signal
with BPSK subcarrier. IEEE Communications Letters 5(5), 185187 (2001)