Professional Documents
Culture Documents
discussions, stats, and author profiles for this publication at: http://www.researchgate.net/publication/3156360
CITATIONS
DOWNLOADS
VIEWS
69
33
154
1 AUTHOR:
Slimane Ben Slimane
KTH Royal Institute of Technology
101 PUBLICATIONS 691 CITATIONS
SEE PROFILE
686
AbstractOrthogonal-frequency-division-multiplexing (OFDM)
techniques allow the transmission of high data rates over broadband radio channels subject to multipath fading without the need
for powerful channel equalization. However, they are very sensitive to nonlinear effects due to the high peak-to-average power
ratio (PAPR) owned by their transmitted signals. This paper
proposes an efficient technique for reducing the PAPR of OFDM
signals. The proposed technique is data-independent and, thus,
does not require new processing and optimization for each transmitted OFDM block. The reduction in PAPR of the OFDM signal
is obtained through a proper selection of a precoding scheme that
distributes the power of each modulated symbol over the OFDM
block. The obtained results show that this precoding scheme is
an attractive solution to the PAPR problem of OFDM signals.
It is shown, through computer simulations, that the PAPR of
precoded OFDM signals approaches that of single-carrier signals.
The good improvement in PAPR given by the present technique
permits the reduction of the complexity and cost of the transmitter
significantly. The precoding schemes also take advantage of the
frequency variations of the communication channel and can provide considerable performance gain in fading-multipath channels.
Index TermsBandwidth efficiency, diversity gain, fading
multipath, low complexity, orthogonal-frequency-divisionmultiplexing (OFDM), peak-to-average power ratio (PAPR),
power efficiency, precoding.
I. I NTRODUCTION
Manuscript received November 23, 2004; revised July 4, 2005, October 26,
2005, and February 21, 2006. The review of this paper was coordinated by
Prof. G. Saulnier.
The author is with the Radio Communication Systems Group, Department
of Communication Systems, the Royal Institute of Technology (KTH), 100 44
Stockholm, Sweden (e-mail: slimane@radio.kth.se).
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.2007.891409
SLIMANE: REDUCING THE PEAK-TO-AVERAGE POWER RATIO OF OFDM SIGNALS THROUGH PRECODING
687
0.0
p1.0
P =
..
.
pL1,0
p0,1
p1,1
..
.
..
.
p0,N 1
p1,N 1
..
.
pL1,1
pL1,N 1
(1)
(2)
Es j (0 +d 2
M ),
e
T
d {0, 1, . . . , M 1}
(3)
(4)
where
Yi =
N
1
i = 0, 1, . . . , L 1.
pi,m Xm ,
(5)
m=0
L1
i=0
Yi ej2i T ,
Tg t < T
(6)
688
s(t) = x(t)ej2fc t
(7)
max |x(t)|2
.
E |x(t)|2
(8)
L1
Yi ej2i T
i=0
N
1
m=0
Xm
L1
j2i Tt
pi,m e
0 t < T (9)
i=0
2
N 1 L1
j2i Tt
pi,m e
m=0 i=0
(10)
1
max
=
N 0t<T
2
N 1 L1
j2i Tt
pi,m e
(11)
m=0 i=0
L1
j2i Tt
, 0t<T
pm (t) = i=0 pi,m e
(12)
0,
otherwise
for m = 0, 1, . . . , N 1.
With the above definition, the PAPR of the precoded OFDM
signal can be rewritten in terms of pm (t) as
1
PAPR(t)
N
N 1
2
|pm (t)|
(13)
m=0
L1
t
j2 im
N ej2i T ,
0t<T
pm (t) = i=0 pi,0 e
(15)
0,
otherwise
where pi,m are the entries of the precoding matrix P in (1).
SLIMANE: REDUCING THE PEAK-TO-AVERAGE POWER RATIO OF OFDM SIGNALS THROUGH PRECODING
It is clear that if p0 (t) has only one amplitude peak, then all
the other functions will also have one amplitude peak and all the
amplitude peaks will not occur at the same time instant. Hence,
this format will certainly reduce the peak-to-average ratio of the
precoded OFDM scheme.
We also notice from the above expression that the entries
of the precoding matrix are related to each other. Furthermore,
all the entries of the different columns of the precoding matrix
are directly obtained from the entries of the first column. This
indicates that we only need to find the first column of the matrix
P , which will of course simplify the design of the precoder
considerably. In fact, from (15), we notice that having a mother
function p(t), we can extract all the entries of the precoding
matrix from p(t) as
pi,m = pi,0 ej2 N
im
j2 im
N
=e
1
T
T
p(t)e2i T dt
t
(16)
PAPRmax
1
max
=
N 0tT
N 1
2
|pm (t)|
(18)
m=0
PAPRmax
N 1
L1
pi,m pi,k =
i=0
1,
0,
m=k
m = k.
L1 2
i j2(mk) i
1
T,
N =
P
e
0,
T i=0
T
T
|p ( (t, m))|
(19)
p(t)ej2f t dt
(24)
(1+)/T
s
where
(t, m) =
(23)
m=0
m=k
m = k
where
P (f ) =
(22)
2
2
T
1 1
|p(t)| dt
N Ts
(21)
(17)
689
|P (f )|2 ej2(mk)Ts f df =
(20)
T,
0,
m=k
m = k
(25)
Ts
|P (f )|2 ej2f t df
q(t) =
0
(26)
690
im
j2 im
N
=e
1
T
T
p(t)e2i T dt
t
with
(1+)/T
s
Psrc (f )ej2f t dt
psrc (t) =
0
and
=
LN
Np
=
N
N
|P (f )|2 ej2(mk)Ts f df =
T,
0,
For large number of subcarriers N , the entries of the precoding matrix can be written as
j2 im
N
pi,m = pi,0 e
m=k
m = k
j2 im
N
=e
j2 im
N
can reduce the PAPR of the OFDM signal without altering the
symbol-separability property of the OFDM scheme.
Note that by selecting a precoding matrix satisfying (27),
we basically obtain an OFDM signal similar to a single-carrier
signal due the wide bandwidth of the set of functions {pm (t)}
and the cyclic shifting in time. However, this precoding scheme
does not remove any of the properties of OFDM, where as
discussed in the following sections, simple (single tap) equalization is still possible. It should also be mentioned that this
scheme is also different from single-carrier signals with cycle
prefix and frequency equalization [24]. In the latter, a cyclic
prefix is added at the transmitter and both fast Fourier transform
and inverse fast Fourier transform (IFFT) operators are done
at the receiver. The proposed precoded scheme in this paper is
an OFDM scheme and does not alter any of the structure and
properties of OFDM systems.
In the following example, we illustrate a possible precoding
scheme designed based on the technique just described above
and evaluate the PAPR of the corresponding precoded OFDM
signal.
A. Example
A well-known function that satisfies the Nyquist criterion is
the raised cosine function [23, p. 547]. Consider the square root
of a raised cosine function, which is denoted psrc (t), and having
a Fourier transform [23, p. 547]
Ts sin f2Ts ,
0 < f Ts
1
(28)
Psrc (f ) = Ts ,
Ts f Ts
T sin (f Ts 1) + , 1 < f 1+
s
Ts
Ts
(30)
2)
(1+)/T
s
(29)
= (1)i e
1
Psrc
T
T
1
T
i
N Ts
p(t)e2i T dt
t
(31)
im
with
pi,0
(1)i
i
sin
2Np ,
N
(1)i
=
,
N
(1)
cos (iN ) ,
2Np
N
(32)
0 i < Np
Np < i N
N iL1
Ts sin2 fTs ,
0 < f Ts
1
(33)
Prc (f ) = Ts ,
Ts f Ts
T sin2 (f Ts 1) + , 1 < f 1+
s
instead of (28).
Ts
Ts
SLIMANE: REDUCING THE PEAK-TO-AVERAGE POWER RATIO OF OFDM SIGNALS THROUGH PRECODING
Fig. 2. Maximum PAPR ratio of the precoded OFDM signal for different
number of subcarriers. With no precoding, we have PAPRmax = 18 dB for
N = 64 and PAPRmax = 21 dB for N = 128.
(34)
691
and PAPRmax . A better measure of the PAPR of communication signals is then to consider the complementary cumulative
distribution function defined as
PPAPR = Pr(PAPR PAPR0 )
where PAPR0 is the PAPR threshold.
Fig. 3 illustrates the complementary cumulative distribution
function of the PAPR of the precoded OFDM signal for the
case of N = 64 subcarriers and Precoder 1. It is observed that
the proposed precoding scheme provides considerable gain in
PAPR for the OFDM signal when compared to that of conventional OFDM. Fig. 4 shows the complementary cumulative
distribution of the PAPR of the precoded OFDM signal for the
case of Precoder 2. We also notice here similar results where the
precoded OFDM scheme outperforms conventional OFDM in
terms of PAPR. It is also observed that Precoder 2 is a bit worse
than Precoder 1, which is consistent with the results of Fig. 2.
In both figures, the results were obtained with an oversampling
equal to four. Compared to the case of the DFT matrix [19], we
notice that with little overhead, the PAPR of the OFDM signal
can be reduced with a reduction that depends on the amount
of overhead. For instance, the PAPR can be reduced by about
1.5 dB with an overhead of 10% and by about 3 dB with an
overhead of 20% at a complementary CDF value of 103 .
To assess the effects of the number of subcarriers on the
signal variations of the precoded OFDM signal, we have looked
at the complementary cumulative distribution function of the
precoded OFDM signal for different number of subcarriers.
This is illustrated in Fig. 5 as a function of the PAPR threshold.
Similar to the results of the maximum PAPR, we notice that
the complementary distribution function of the PAPR of the
precoded scheme is not very sensitive to the number of subcarriers and especially at high values of the PAPR threshold. Fig. 6
compares the performance of this precoded scheme to that of
the TR method [7] for the case of N = 256 subcarriers and
an overhead of = 5%. It is observed that precoding provides
lower PAPR values for OFDM signals as compared to the TR
692
this case in this section and see what kind of reliable lowcomplexity detector(s) can be used.
A. Channel Model
The fading channel considered in this paper is a fadingmultipath channel with coherence bandwidth smaller than the
total bandwidth of the OFDM system and, thus, seen as
frequency-selective fading. The fading process is assumed to be
stationary and slowly varying compared with the block duration
of the OFDM signal, such that it is approximately constant
during at least one block length.
The complex baseband representation of the fadingmultipath channel impulse response can be described by [23]
h( ) =
hl ( l )
(35)
l=0
method, 1-dB better than optimum TR, and 1.5-dB better than
TR with 40 iterations at a complementary CDF of 103 .
Notice that the precoding scheme discussed above is just an
example and that several other schemes may exist!
P
1
where hl is a complex random-variable tap weight with variance pl , l is the time delay of the lth path, and P is the
total number of received paths. The tap weights are assumed
independent for different paths.
B. Receiver Structure
Passed through the channel, the equivalent lowpass of the
received precoded OFDM signal during the zeroth block interval is given by
r(t) =
P
1
hl x(t l ) + z(t),
Tg t < T
(36)
l=0
where z(t) is complex Gaussian random process with zeromean and power spectral density N0 , and Tg is the time guard
interval.
SLIMANE: REDUCING THE PEAK-TO-AVERAGE POWER RATIO OF OFDM SIGNALS THROUGH PRECODING
693
T
r(t)ej2 T t dt
i
= T Hi Yi + Zi ,
i = 0, 1, . . . , L 1
(37)
P
1
hl ej2ii /T
Fig. 7. Discrete representation of the precoded OFDM scheme and its receiver
structure.
(38)
l=0
P
1
pl .
l=0
i = 0, 1, . . . , L 1 (39)
where
Yi =
N
1
pi,m Cm ,
i = 0, 1, . . . , L 1
m=0
D = T HP X + Z
(40)
where X is as given in (2), Y is as given in (4)
Z = [Z0 , Z1 , . . . , ZL1 ]T
(41)
H
i
i |2 + 2 / 2
|H
z
s
i = 0, 1, . . . , L 1
= T P GHP X + Z
H = diag{H0 , H1 , . . . , HL1 }.
(42)
(44)
is the noise vector, and H is an L L diagonal matrix representing the channel coefficients of the different subcarriers with
(43)
(45)
694
TABLE I
CHANNEL MODELS FOR HIPERLAN/2 IN DIFFERENT INDOOR SCENARIOS
Fig. 8.
SLIMANE: REDUCING THE PEAK-TO-AVERAGE POWER RATIO OF OFDM SIGNALS THROUGH PRECODING
695
[28] K. Fazel, S. Kaiser, and M. Schnell, A flexible and high performance mobile communications system based on orthogonal multi-carrier SSMA,
Wirel. Pers. Commun., vol. 2, no. 1/2, pp. 121144, 1995.
[29] K. Fazel, Performance of CDMA/OFDM for mobile communication
systems, in Proc. ICUPC, 1993, pp. 975979.
[30] N. Yee and J.-P. Linnartz, Controlled equalization of multi-carrier
CDMA in an indoor Rician fading channel, in Proc. VTC, Stockholm,
Sweden, May 1994, pp. 16651669.
[31] S. Kaiser and J. Hagenauer, Multi-carrier CDMA with iterative decoding
and soft-interference cancellation, in Proc. Globecom, Nov. 38, 1997,
pp. 610.
[32] D. N. Kalofonos, M. Stojanovic, and J. G. Proakis, On the performance
of adaptive MMSE detectors for a MC-CDMA system in fast fading
Rayleigh channels, in Proc. PIMRC, Sep. 811, 1998, pp. 13091313.
[33] T. Mueller, K. Brueninghaus, and H. Rohling, Performance of coherent OFDM-CDMA for Broadband mobile communications, Wirel. Pers.
Commun., vol. 2, no. 4, pp. 295305, 1996.
[34] S. B. Slimane, Partial equalization of multi-carrier CDMA in frequency selective fading channels, in Proc. IEEE ICC, New Orleans, LA,
Jun. 1820, 2000, pp. 2630.
[35] J. Medbo, H. Andersson, P. Schramm, H. Asplund, and J.-E. Berg,
Channel models for HIPERLAN/2 in different indoor scenarios,
EURO-COST, Bradford, U.K., COST 259 TD(98)70, Apr. 1998.