(1)
where N is the number of IFFT points which includes M
modulated information data and (NM) zeropadding in the
frequency domain. Here R=N/M represents the over sampling
ratio. The PAPR for the transmitted OFDM signal is defined as
the ratio of the maximum peak to average power as given by[9],
2
10 2
10
[ ]
k
k
max x
PAPR log
E x
= (2)
where
2
k
[ ] E x represents the averaged signal power.
From (1), it can be observed that the maximum peak power
of OFDM time domain signal would be relatively larger than
the conventional single carrier modulated signal because the
time domain signal is obtained by the summation of M random
information data in the processing of IFFT. The OFDM signal
with higher PAPR would produce the larger intermodulation
noise incurred at the nonlinear amplifier which falls into both
inside and outside of allocated frequency bandwidth. The
former noise causes the degradation of BER performance and
the latter noise causes the interference to the adjacent channel
in the FDMA system.
The simple solution for PAPR problem is to operate the
nonlinear amplifier in the linear region by taking enough
lower input backoff. Although the level of intermodulation
noise can be reduced sufficiently when the nonlinear amplifier
is operated in the linear region, the usage of power amplifier
becomes inefficient. The inefficient usage of power amplifier
leads the serious problem especially for the wireless portable
terminal, which requires the long operation by the battery.
Another solution to solve the PAPR problem is to employ
the PTS method. Fig.1 shows the structure of OFDM
transmitter with PTS technique when the number of clusters V
is 4. In the PTS method, M data information ] [
M 1
X X = X in
the frequency domain is partitioned into V clusters, which is
expressed by the following equation.
=
=
V
1 v
v X X (3)
where
v X represents the information data at the v th cluster
which consists of M/V subcarriers. The time domain signal
v
x which is converted from the frequency data information at
the v th cluster is given by,
{ }
0 1 1
X [ ]
v k N
IFFT x , x , x , , x
v v v v
v
= = x (4)
The time domain signal for each cluster is multiplied by the
certain weighting factor so as to minimize the PAPR
performance. The combined time domain signal with the
certain weighting factors for all clusters can be given by,
 
'
1
' '
1
'
0
1
, , , , ,
~
=
= =
N k v
V
v
v
x x x x b x x (5)
where b
v
is the phase coefficient for the v th cluster. The
phase coefficient will be selected from among the following,
when the number of predetermined phases is 4.
3 1 2 4
, , ,
j j j j
b e e e e
u u u u
v
( e
(6)
By using (5), the optimum phase coefficient for each cluster,
which could achieve the best PAPR performance can be
obtained by the following equation.
{ }
1 2 3 4
'
1 2 3 4
0 1
[ , , , ]
[ , , , ] arg min max
k
k N
b b b b x
u u u u
s s
=
(7)
Actually, the optimum phase coefficients for all clusters are
decided by the exhausted search for all combinations of phase
coefficients to achieve the best PAPR performance which leads
the higher computational complexity.
S
/
P
a
n
d
p
a
r
t
i
t
i
o
n
i
n
t
o
c
l
u
s
t
e
r
s
IFFT
IFFT
IFFT
X
1
X
2
X
3
X
4
X
1
x
Optimization of
weighting factor
2
x
3
x
4
x
x
b
1
b
2
b
3
b
4
IFFT
+
Fig.1 Structure of PTS Method when V=4.
III. PROPOSAL OF NEW PTS METHOD WITH CYCLE
SAMPLE SHIFTING
In the PTS technique, the computational complexity to
achieve the better PAPR performance increases exponentially
as increasing the number of clusters and phase coefficients. In
the optimization for the phase coefficients for all clusters given
in (7) includes both processing for addition and multiplication
of phase coefficients.
To reduce the computational complexity required in the
PTS method, this paper proposes a new PTS method in which
the time sample for each cluster is shifted so that the combined
time domain signal for all clusters could achieve the better
PAPR performance. Fig.2 shows the structure of proposed PTS
method with cycle time sample shifting.
X
1
X
2
X
3
X
4
X
1
x
2
x
3
x
4
x
+
x
d
1
d
2
d
3
d
4
cyclic
shift
IFFT
S
/
P
a
n
d
p
a
r
t
i
t
i
o
n
i
n
t
o
c
l
u
s
t
e
r
s
IFFT
IFFT
IFFT
cyclic
shift
cyclic
shift
cyclic
shift
Optimization of
time sample shifting
Fig.2. Structure of proposed PTS method.
In the proposed PTS method, as the same manner as the
conventional PTS method, the data information in the
frequency domain is partitioned into V clusters as given in (3),
and converted them to the time domain signal by IFFT as given
in (4). The difference of proposed PTS method from the
conventional PTS method is to shift the certain time sample for
each cluster instead of multiplying the phase coefficient to the
time domain signal. Since the time domain OFDM signal is
obtained from the frequency domain data information by using
IFFT, adding the time delay to the time domain OFDM signal
is equivalent to shift the time sample cyclically. When d
samples is delayed to the original time domain OFDM signal
for thev th cluster as given in (4), the delayed time domain
signal can be obtained by shifting d samples cyclically as
shown in Fig.3. The delayed signal for the v th cluster is given
by the following equation.
] , , , , , , , , , [ ) (
1 1 0 1 1
v
d
v v v
N
v
k d
v
d
v
d v v
v v v v
x x x x x x x d
+ +
= x
(8)
In (8), if the sample number ( ) d k
v
+ in the time domain signal
is larger than (N1), the sample number becomes modulo N of
( ) d k
v
+ from the fact that the time domain signal after IFFT is
repeated cyclically by N. Here it should be noted that the time
delay in the time domain signal corresponds to the phase
shifting in the frequency domain signal as given in the
following equation.
2 nd
j
N
k d n
x X e
v
v
t
v v
+
(9)
In the proposed PTS method, the PAPR could be reduced by
summations of all clusters time domain signal, which have the
different time sample shifting. The combined time domain
signal for all clusters with the certain time sample shifting can
be given by the following equation.
 
"
1
" "
1
"
0
1
, , , , , ) (
~
=
= =
N k v
V
v
v d
x x x x d x x
(10)
In the proposed method, the time sample shifting for each
cluster is selected from among the following.
 
1 2 /
, , ,
N D
d m m m
v
e (11)
where D represents the interval of time sample shifting and
N/D is the total number of predetermined candidate sample
shifting. By using (10), the optimum sample shifting for each
cluster, which can achieve the best PAPR performance, can be
given by the following equation when the number of cluster V
is 4.
{ }
1 2 /
''
1 2 3 4
0 1
[ , , , ]
[ , , , ] arg min max
N D
k
k N
m m m
d d d d x
s s
=
(12)
In the proposed PTS method, the best PAPR performance
could be achieved by the exhausted search for all
combinations of time sample shifting for all clusters. However,
the computation complexity becomes exponentially larger as
increasing the number of clusters and the number of
predetermined time sample shifting N/D as given in (11).
When the number of clusters is V and the number of time
sample shifting is N/D, the order of computational complexity
by using the exhausted search is V
(N/D)
, which becomes larger
computational complexity.
To solve the above problem, this paper proposes the
algorithm, which could achieve the better PAPR with small
computational complexity, which is almost the same as the
conventional PTS method. Fig.4 shows the procedures for the
proposed algorithm. In the proposed algorithm, the
optimization of time sample shifting for each cluster is carried
out independently from other clusters.
d Sample Shift
Original Time Domain Signal
Cyclic Sample Shifting Signal
0
x
1
x
2
x
2 d
x
1 d
x
d
x
1 d
x
+ 2 d
x
+ 2 N
x
1 N
x
0
x
1
x
2
x
2 d
x
1 d
x
d
x
1 d
x
+ 2 d
x
+ 2 N
x
1 N
x
Fig.3 Cyclic sample shifting of OFDM time domain signal.
 
  s
Change Cluster
Repeat all procedures
v
(
m
d Change Time Sample
m
( d )
v
v
x
1
V
s s
d
x ( d )
v
v
=
=
=
x
s
m d
( d ) x
v
v
+ x
if
s
m m
( d ) x ( d )
v v
v v
= x
m N / D >
if
V v >
if
s S >
if
1 m m = +
1 v v = +
1 s s = +
1
V
s s
d
x ( d )
v
=
=
x
Lower PAPR Signal with Optimum Time Sample Shifting
YES
NO
YES
NO
YES
NO
YES
NO
Calculate PAPR ]
v
v
m
, ,d [s
m m1
[s, , d ] [s, , d ] PAPR PAPR
v v
s
0
1
V
x ( d )
v
v =
=
x
Fig.4 Procedures for proposed PTS method.
As shown in Fig.4, the time sample shifting is changed [m]
from m=1 to m=N/D for the v th cluster with keeping the
same time sample shifting for other clusters and the same
procedures are repeated by changing the cluster [v] one by one
so as to achieve the better PAPR. Then all procedures are
repeated by changing [s] up to obtain the better PAPR
performance. The proposed procedures could achieve the
better PAPR performance with small computation complexity
because the exhausted search is unemployed as the
conventional PTS method. The proposed PTS method also
enables the reduction of computational complexity, because
the required processing is the cyclic time sample shifting and
summation. The required computation complexity for the
proposed PTS method is evaluated in Section IV as compared
with the conventional PTS method.
IV. PERFORMANCE EVALUATION
This section presents the various computer simulation
results to verify the effectiveness of proposed PTS method.
Table.1 shows the simulation parameters to be used in the
following evaluations. The following simulations assume that
the allocated frequency bandwidths is 5MHz, the modulation
method is 16QAM, the number of subcarriers M is 64, the
number of FFT points N is 256, and the over sampling ratio is
4 to achieve the accurate evaluation of PAPR performance [9].
Table 1 Simulation parameters.
Allocated Bandwidth 5 MHz
Modulation method 16QAM
Number of subcarriers (M) 64
Number of FFT points (N) 256
Guard interval 15 samples
Frame length 50 symbols
Number of cluster (V) 4
Interval of time delay (D) for NPTS 2  8 samples
Number of phase coefficients for CPTS 4
Nonlinear amplifier SSPA
Rap coefficient for nonlinear (r) 1
Input backoff 3dB
Multipath fading model
Power delay profile exponential
Delay constant 1dB
Number of delay paths 12
The nonlinear amplifier assumed in this paper is the Solid
State Power Amplifier (SSPA), of which AM/AM conversion
characteristics is modeled by [9],
1 2
1 2
r / r
F[ ]
{ ( / A) }
=
+
(13)
where
is the amplitude of input signal, A is the saturated
output level and r is the parameter to decide the nonlinear
level which calls the rap coefficient. In the simulation, the rap
coefficient r is 1 and the input back off (IBO) is 3dB. The
multipath fading is modeled by the exponential power delay
profile of 12 delay paths with 1dB decay constant. In the
evaluation of PAPR performance, the complementary
cumulative distribution function (CCDF) is used which shows
the probability of exceeding a given threshold PAPR
0
.
Fig.5 shows the improvement of PAPR performance for
the New PTS (NPTS) with proposed algorithm as compared
with the conventional PTS (CPTS) at CCDF=10
1
when
changing the interval of sample shifting D. Here it should be
noted that the BER performance in the nonlinear channel is
decided dominantly by the PAPR performance at around
CCDF=10
1
. From the figure, it can be observed that the
PAPR performance is converged when the number of
repetitions S is 2 regardless of the interval of time sample
shifting D. Although the improvement of NPTS as compared
with CPTS at CCDF=10
1
increases as decreasing the interval
of time sample shifting D, the computational complexity
would increase accordingly.
Fig.5 PAPR improvement of NPTS as compared with CPTS when
changing the number of repetitions.
The computational complexity for the conventional CPTS
method, when the number of clusters is V and the number of
phase coefficients is P, can be estimated by the following
equation.
( )
1
2
1
V
C PTS
L NV log N 2NP V
= +
(14)
where the computational complexity of CPTS method L
CPTS
consists of V times IFFT processing
2
( NV log N ) and
1
1
V
N(V )P
times summation and multiplication of phase
coefficients.
The computational complexity for the NPTS method when
the number of clusters is V, the number of predetermined
sample shifting is N/D and the number of repetitions for the
proposed algorithm S is evaluated by the following equation.
( ) { }
2
2
2 3
N PTS
L NV log N VN / D N V S
= + + (15)
where the computational complexity of proposed NPTS
method L
NPTS
consists of N NV 2 log IFFT processing and
{ }
2
2 3 VN / D N( V ) S +
times processing of summation. From
(15), it can be seen that the required processing for the
proposed method is only for the summation except V times
IFFT processing. The ratio of computational complexity
between the CPTS and NPTS can be defined by,
L N PTS C PTS
R L / L
= (16)
Table.2 shows the comparison of computational complexity
between the CPTS and NPTS methods when the number of
clusters V is 4. From Table.2 and Fig. 5, the proposed NPTS
method with D=6 and D=8 show the better PAPR performance
by 0.91dB and 0.79dB with the smaller computational
1 2 3 4
0.7
0.8
0.9
1
1.1
Number of Repetitions for NPTS (S)
P
A
P
R
I
m
p
r
o
v
e
m
e
n
t
o
f
N

P
T
S
f
r
o
m
C

P
T
S
[
d
B
]
D=8
D=6
D=4
D=2
complexity by 0.92 and 0.72, respectively than those for the
conventional CPTS method.
Fig.6 shows the PAPR performance based on CCDF for
the proposed NPTS method with V=4, D=6 and S=2. In the
figure, the PAPR performances for the conventional OFDM
signal without PAPR reduction method and the conventional
CPTS method with V=4 and P=4 are also shown as for the
comparison. From the figure, it can be observed that the
proposed NPTS method shows better PAPR performance than
the conventional OFDM by 3dB and the CPTS by 0.91dB at
CCDF=10
1
.
Table.2 Computational complexity and PAPR Improvement.
CPTS
(V=4, P=4)
NPTS
(V=4, S=2) Computational
Complexity
Ratio R
L
Eq.(16)
PAPR
Improvement
of NPTS
from C PTS
at CCDF=10
1
Computational
Complexity
L
CPTS
Eq.(14)
Computational
Complexity
L
NPTS
Eq.(15)
106,496
D=2 272,896 2.56 1.08dB
D=4 141,824 1.33 0.94dB
D=6 98,133 0.92 0.91dB
D=8 76,288 0.72 0.79dB
Fig.6. CCDF PAPR performance for proposed NPTS.
Fig.7 shows the BER performance for the proposed NPTS
and the conventional CPTS methods in the nonlinear and
multipath fading channel. The input backoff (IBO) of SSPA
nonlinear amplifier is operated at 3dB. From the figure, the
proposed NPTS method shows better BER performance than
the CPTS and almost the same BER performance as that in the
linear channel.
V. CONCLUSIONS
This paper proposed the new PTS method with cyclic
sample time shifting, which can achieve the better PAPR
performance and lower computational complexity than those
for the conventional PTS method. The salient feature of
proposed method is to employ the characteristics of IFFT,
which enables the addition of time delay to the time domain
signal by shifting the time sample cyclically. From this fact, the
proposed method can achieve the better PAPR performance
only by the processing of sample shifting and summation,
which is different from the conventional PTS methods required
for the processing of summation and multiplications. This
paper also proposed the computational complexity reduction
algorithm for the new PTS method.
The computer simulation results showed that the proposed
PTS method can achieve the better PAPR performance with
lower computational complexity than the conventional PTS
and shows the better BER performance in the nonlinear
channel, which is almost the same as that in the linear channel.
The detection method for the time sample shifting of each
cluster, which enables the correct demodulation without side
information at the receiver, is subject to the further study.
Fig.7. BER performance of proposed NPTS versus C/N.
References
[1] J. Chuang and N. Sollenberger, Beyond 3G : Wideband Wireless Data
Access Based on OFDM and Dynamic Packet Assignment, IEEE
Commun. Mag., vol.38, no.7, pp78pp87, July.2001.
[2] ARIB STDB31, Transmission System for Digital Terrestrial
Television Broadcasting, May 2001.
[3] IEEE Std.802.11a, Highspeed Physical Layer in the 5GHz Band,
1999.
[4] A. Jamalipour, T. Wada and T. Yamazato, A Tutorial on Multiple
Access Technologies for Beyond 3G Mobile Networks, IEEE
Commun., Mag., Vol.43, pp. 110117, Feb. 2005.
[5] J. Armstrong, Peaktoaverage power reduction for OFDM by repeated
clipping and frequency domain filtering, Electron. Lett., vol. 38,
pp.246247, Feb. 2002.
[6] L. Wang and C. Tellambura, A Simplified Clipping and Filtering
Technique for PAR reduction in OFDM System, IEEE SIGNAL
PROCESSING LETTERS, vol. 12, no. 6, Jun 2005.
[7] R.W. Bauml, R. F. H. Fisher, and J. B. Huber, Reducing the Peakto
Average Power Ratio of Multicarrer Modulation by Selected Mapping,
Elect. Lett., vol. 32, no. 22, pp. 20562057, Oct. 1996.
[8] S. H. Mller and J. B. Hurber, OFDM with reduced peaktoaverage
power ratio by optimum combination of partial transmit sequence,
Elect. Lett., vol. 33, no. 5, pp. 368369, Feb. 1997.
[9] A.R.S. Bahai and B.R.Saltzberg, MultiCarrier Digital
Communications Theory and Applications of OFDM, Kluwer
Academic/Plenum Publishers, 1999.
0 1 2 3 4 5 6 7 8 9 10 11 12
10
4
10
3
10
2
10
1
10
0
PAPR[dB]
C
C
D
F
CPTS
(V=4, P=4)
OFDM
NPTS
(V=4, D=6, S=2)
20 25 30 35 40
10
5
10
4
10
3
10
2
10
1
C/N[dB]
B
E
R
OFDM
CPTS
(V=4, P=4)
NPTS
(V=4, D=6, S=2)
OFDM without SSPA
IBO=3dB
with SSPA