1558 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO.

5, MAY 2010
Quantifying an Iterative Clipping and Filtering Technique for
Reducing PAR in OFDM
Kitaek Bae, Student Member, IEEE, Jeffrey G. Andrews, Senior Member, IEEE,
and Edward J. Powers, Life Fellow, IEEE
AbstractIn Orthogonal Frequency Division Multiplexing,
a simple clipping method is widely used in order to reduce
the peak-to-average power ratio since it is easy to implement.
The performance analysis of the clipping approach has been
previously introduced in the literature. Clipping, however, is a
nonlinear process and may cause two major undesirable effects:
(i) spectral regrowth, which causes unacceptable out-of-band
radiation; and (ii) distortion of the desired signal, which increases
bit-error-rate (BER). The out-of-band radiation can and often is
suppressed by ltering, which leads to peak regrowth. Therefore,
iterative clipping and ltering is required until the desired
clipping level is achieved. However, this iterative process makes
BER estimation difcult. This letter provides expressions and
analytical techniques for estimating the attenuation factor, error
vector magnitude, and BER, using a noise enhancement factor
that is obtained by simulation. Simulation results show strong
agreement with our semi-analytical results for 1024 subcarriers.
Index TermsPeak-to-average ratio (PAR), iterative clipping
and ltering, noise enhancement factor, OFDM, BER, AWGN.
I. INTRODUCTION
A
LTHOUGH Orthogonal Frequency Division Multiplex-
ing (OFDM) offers multiple advantages over single-
carrier systems in high data rate wireless communication
systems, its inherently high Peak-to-Average Ratio (PAR) is
a major impediment to current and future OFDM standard
systems including WiMAX, 3GPP LTE, and 802.11n. This
high PAR leads to in-band and out-of-band distortions in
high power RF power ampliers (PA) because of the limited
linear operating range of such PAs. To avoid these nonlinear
distortions, a relatively large output backoff (OBO) is required
at the cost of power efciency. Therefore, PAR reduction of
the baseband OFDM signal prior to the PA will increase power
efciency with minimum distortion.
A number of PAR reduction techniques have been proposed
in the literature [1], among them, clipping based techniques,
such as tone-reservation, active constellation extension, and
clipping and ltering. Clipping, however, introduces two major
undesirable nonlinear effects: out-of-band radiation and in-
band distortion. The out-of-band radiation results in unac-
ceptable interference to users in neighboring RF channels.
However, ltering, which allows one to remove this out-of-
band interference introduced by clipping, leads to the peak
Manuscript received April 8, 2009; revised October 2, 2009 and March 7,
2010; accepted March 11, 2010. The associate editor coordinating the review
of this letter and approving it for publication was D. Dardari.
The authors are with the Wireless Networking and Communications Group,
Dept. of Electrical and Computer Engineering, University of Texas at Austin
(e-mail: {ktbae, jgandrews, ejpowers}@mail.utexas.edu).
Digital Object Identier 10.1109/TWC.2010.05.090508
regrowth. Therefore, iterative clipping and ltering algorithms
have been proposed to both remove the out-of-band inter-
ference and suppress the regrowth of the peak power [2, 3].
Iterative methods of clipping and ltering can also be used to
alleviate the clipping noise at the receiver [4, 5].
To quantify the PAR reduction of iterative clipping and
ltering, the complementary cumulative distribution function
(CCDF) of PAR is widely used [6]. However, the CCDF
for PAR has no direct relationship to the BER performance
degradation of an OFDM system with clipping, because the
statistical distribution of PAR focuses only on the highest peak
in an OFDM symbol. To understand the PAR problem, we
must be able to characterize the clipping noise associated
with clipping. Hence, the performance analysis for a clip-
ping channel in terms of signal-to-noise and distortion ratio
(SNDR) has been investigated in recent years [711]. Based
on the assumption that an OFDM signal may be characterized
as a complex Gaussian process, these approaches treat the
clipping noise as an additive Gaussian process and derive the
variance and the power spectral density of the clipping noise.
On the other hand, Bahai et al. characterized the distortion
as a rare impulse noise [12]. A similar analysis in optical
communications has been carried out at a high clipping level
[13]. However, to the best of our knowledge, no analytic model
has considered both iterative clipping and ltering together.
The main obstacle for such a model is that iterative processing
ensuring PAR reduction violates the Gaussian approximation
of the input signal to the next stage of clipping. Therefore, it is
difcult to characterize clipping noise for iterative processing.
Instead, the effects of iterative clipping and ltering have been
studied by extensive simulations, which is a time-consuming
process [11, 14].
In this letter, using the noise enhancement factor, we
present and derive the semi-analytical results for the output of
asymptotic iterative clipping and ltering. With these semi-
analytical results, we can obtain a relationship between the
BER and the target clipping level for asymptotic iterative
clipping and ltering. This semi-analytical BER performance
result is veried by comparison with simulation results.
II. SYSTEM MODEL
Fig. 1 shows the OFDM system using iterative clipping
and ltering to be considered. First, data bits are mapped
into

, the complex data symbols of the th subcarrier

using quadrature amplitude modulation (QAM). Let

, =
0, 1, , 1 denote an oversampled complex baseband
1536-1276/10$25.00 c 2010 IEEE
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 5, MAY 2010 1559
OFDM
modulator
Iterative Clipping
and Filtering
+
OFDM
demodulator
k
X
k
Y
n
x
( ) L
n
x
n
w
n
y
Fig. 1. Block diagram of an OFDM transmission system with iterative
clipping and ltering.
OFDM signal expressed by

=
1

=0

(1)
where is the number of subcarriers, and the oversampling
factor 4 is used to obtain accurate peaks of the continuous
baseband signals in the discrete time domain [15] and to
investigate the out-of-band nonlinear effects associated with
at least the third intermodulation product (IMP).
We assume that the OFDM signal power is normalized as

=
1
2
[

2
] = 1.
Next, the modulated OFDM signal is transmitted through
the iterative clipping and ltering module to reduce the PAR.
We consider iterative clipping and ltering in the frequency
domain [3].
With respect to the clipping of the amplitudes of the samples

, a complex nonlinear function for the th sample can be

expressed by

) =
_

max

max
e

, otherwise
(2)
where

, and the clipping level

max
and the clipping
ratio are related by


max

2
]
. (3)
Clipping is followed by ltering to remove the undesirable
out-of-band radiation; however, this ltering results in a degree
of peak regrowth at some points. Until the target clipping
ratio is achieved, iterative clipping and ltering is used. As
an example of iterative clipping and ltering, the number
of subcarriers is = 1024, and the modulation is 16-
QAM, and a target clipping ratio = 4dB is used. Fig.
2 shows the envelope power distribution function using the
iterative clipping and ltering with denoting the number of
clipping and ltering operations for different clipping ratios.
The envelope power distribution function is the probability that
the instantaneous normalized power of
()

, the output after

iterations of clipping and lering, is greater than the clipping
ratio . From the gure, it is observed that, at the xed target
clipping ratio = 4dB, the envelope power uctuation is
strongly reduced by increasing the number of iterations, .
In Fig. 2 and subsequent gures, we permit the number of
iterations to be as large as = 40 in order to ensure we
reach an asymptotic result. However, this does not necessarily
imply that = 40 iterations are required to achieve a practical
level of performance. For example, as Fig. 2 suggests, fewer
iterations, say = 8, yields near-asymptotic performance.
This issue will be discussed later in section V.
0 2 4 6 8 10 12
10
4
10
3
10
2
10
1
10
0
Reference Envelope Power Level (dB)
E
n
v
e
l
o
p
e

P
o
w
e
r

D
i
s
t
r
i
b
u
t
i
o
n

F
u
n
c
t
i
o
n
L=0
L=1
L=2
L=40
Fig. 2. The envelope power distribution function (CCDF of instantaneous
power) of
()

vs. different reference envelope power levels for a target

clipping ratio =4dB when the number of iterations ranges from from 1
to 40.
At the receiver, the time-domain signal received after pass-
ing through the additive white Gaussian noise channel can be
expressed as,

=
()

, = 0, 1, , 1, (4)
where
()

is the output after iterations of clipping and

ltering and

is the additive white Gaussian noise (AWGN).

Finally, signal samples

including Gaussian noise are fed to

a conventional DFT-based OFDM demodulator as shown in
Fig. 1.
In the following section, we will provide a semi-analytical
expression for an asymptotic iterative clipping and ltering
scheme and quantify its effect on a signal received over an
AWGN channel.
III. REPRESENTATION OF THE OUTPUT FOR ASYMPTOTIC
ITERATIVE CLIPPING AND FILTERING
The statistical properties of the only-clipped OFDM signal
have recently been analyzed [8, 10]. We rst represent the
output of clipping (2) as a form of peak cancellation as
follows:

) =

, for = 0, 1, ..., 1 (5)

where

is the anti-peak sample after clipping. Based on

the Bussgang theorem for Gaussian inputs [16], the anti-peak
sample can be expressed by

= ( 1)

, (6)
where the rst term on the right-hand side of (6), represents an
attenuated replica of the original signal components,

is a
zero-mean noise process uncorrelated with the signal ([

] =
0 and [

] = 0), and the attenuation factor is given by

[10]
=
[

]
[

]
= 1

2
+

2
erfc(). (7)
1560 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 5, MAY 2010
Note that the nonlinear noise

is spread both inside and

outside the bandwidth of the signal. To eliminate the out-of-
band interference caused by clipping, ltering follows clip-
ping. After applying the DFT to (5), a clipped OFDM symbol
after ltering in the frequency domain can be written as

= (

, for = 0, 1, ..., 1
=

, for = 0, 1, ..., 1,
(8)
where

is the complex anti-peak term in the frequency

domain, i.e., the DFT of (6), and

and

are the anti-

peak term and distortion term after ltering in the frequency
domain, respectively, and

is the frequency domain lter

described as

=
_
1, for = 0, ..., 1
0, otherwise.
(9)
While the rst term on the right hand side of (8),

, is
not affected by the ltering, the uncorrelated distortion

is
bandlimited as a result of removing the out-of-band radiation
of

. However, the band limitation of the clipped signal will

cause peak regrowth of the clipped signal at some points in the
time domain. The amount of peak regrowth equals the peak
difference between the ltered signal and the signal before
clipping. To reduce the peak regrowth, we can express iterative
clipping and ltering as

(+1)

()

()

, for = 0, 1, ..., 1
=

(1)

(2)

+ +

()

. .

()

=1

()

(10)
where

()

is the clipping noise after the -th iterative

clipping and ltering operation, and

()

corresponds to the
envelope power of
()

greater than the xed target clipping

ratio of = 4dB in our example. See Fig. 2 for the envelope
power distribution function of
()

with a target clipping ratio

of = 4dB. The asymptotic iterative clipping and ltering
has nite clipping noise because the envelope power of
()

reaches the target clipping level; therefore,

()

becomes
negligibly small with an increasing number of iterations,
which will be generalized for any value of the target clipping
ratio in Fig. 3.
When 2, the input signal to the clipping process is not
Gaussian. Therefore, it is difcult to derive both its distribution
and a closed form for the BER performance analysis to the
iterative clipping and ltering. When the clipping ratio is very
large, the clipping noise after iterative clipping and ltering is
analyzed as a series of parabolic arc pulses and is simplied
in [17]. Since this approach is established for relatively high
clipping levels, the simplied clipping noise is not accurate
when the clipping ratio is small, the latter case of which is of
interest for our performance analysis. Moreover, it is difcult
to derive a simple closed form for the clipping noise because
the interaction between the clipping pulses increases as the
clipping ratio decreases. In this letter, in order to provide a
general expression for the asymptotic iterative clipping and
ltering for a given clipping ratio, the noise enhancement
factor is introduced as follows:
Denition 1: Noise Enhancement Factor
()
Let the th clipping noise,

()

be calculated using the

sum of anti-peak signals
()

of each iteration as indicated

in (10). We dene the noise enhancement factor
()
as the
normalized cross-correlation between the clipping noise at the
-th iteration and the clipping noise at the rst iteration in the
frequency domain. Therefore,
()
is dened as

()
=
[

()

(1)

]
[

(1)

(1)

]
, (11)
where

(1)

and

()

are the clipping noise at the rst and

-th iterations, respectively.
In accordance with the clipping model in (8), we can
represent the output of the iterative clipping and ltering after
the th iteration in (10) as

()

+
()

(1)

, for = 0, 1, ..., 1
=
()

+
()

(12)
where
()

=
()

is the complex distortion term on

the th subcarrier after the th iteration, and the attenuation
factor
()
at the th iteration can be easily obtained in a
straightforward way as

()
= 1 (1 )
()
. (13)
Note that
(1)
= , since
()
= 1 when = 1; otherwise,

()
is reduced since > 1 as increases, which means that
the signal constellation after the th iteration at the receiver
is smaller than that with = 1.
IV. BER ESTIMATION OF ASYMPTOTIC ITERATIVE
CLIPPING AND FILTERING
We focus on the asymptotic iterative clipping and ltering
represented in the previous section. To assess the receiver
performance, the received signal after the DFT with perfect
synchronization can be represented as

=
()

+
()

, = 0, 1, ..., 1, (14)
where
()

is the in-band distortion after iterations and

is the additive zero-mean white Gaussian noise (AWGN) with

variance
2

=
1
2
[

2
]. Note that

,
()

, and

are
assumed to be mutually independent.
For the asymptotic iterative clipping and ltering, the
clipped pieces of signal higher than the target clip level
max
become in-band distortion without affecting the out-of-band
because of ltering, while the signal power decreases due to
the clipping process. Thus, with the chosen parameters,
()
in
(11) and
()
in (13), the effective SNR for the th subcarrier
can be expressed as [10]
SNR

=
[
()

2
]
[

()

2
]
=
[
()

2
]
[
()

2
]+[

2
]
, (15)
where [
()

2
] is the average reduced signal power and
[
()

2
] is the average power of the distortions, which
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 5, MAY 2010 1561
is obtained by making use of the power spectral density of
distortion for the soft limiter in [9] as follows:
[
()

2
] =(
()
)
2
[

] = (
()
)
2

(),
for = 0, 1, ..., 1,
(16)
where

() is the power spectral density (PSD) of the

distortion noise and can be obtained from the DFT of the
correlation function of the distortion

() = [
+

] =

=1

2+1
_

()

(0)
_
2+1
. (17)
Here, the coefcient
2+1
, which can be found in [9], is a
function of the clipping ratio , and

(), the autocorrela-

tion of the signal

, is given by [11]

() = 2
2

sinc(

)
sinc(

()

, (18)
where sinc() =
sin

and the input signal power

(0) is
2
2

. Note that [
()

2
] is a function of the -th subcarrier
since

() is not constant for all OFDM subcarriers, but, we

assume that the distortion power is spread out equally over the
N-subcarriers. In practice, the error vector magnitude (EVM)
is dened as [18]
EVM=

=0
[
()

2
]
[

2
]
=

(
()
)
2
2
2

=0

(). (19)
Consequently, the theoretical evaluation of the average bit
error rate (BER) can be obtained by the central limit theorem
when the number of subcarriers is sufciently large as
follows:

=
4
log
2
()
_
1
1

3
1
SNR

_
, (20)
where is the constellation order, () =
1

exp(
2
/2) is the normal error integral, and
SNR

is the average effective SNR which is expressed by

SNR

=
[
()

2
]
[
()

2
]+[

2
]
=
_
SDRSNR

1+SDR+SNR

_
, (21)
where SDR =
[
()

2
]
[
()

2
]
=
(
()
)
2
EVM
2
, and SNR

=
[

()

2
[

2
]
= log
2
()

.
Note that Eq. (20) for BER depends on the analytic results
for both the attenuation factor
()
in (13) and the EVM
in (19). Furthermore, both
()
and EVM depend on the
noise enhancement factor
()
dened in (11). As mentioned
previously, it is difcult to derive a closed form expression
for the clipping noise because the interaction between the
clipping pulses increases as the clipping ratio decreases.
Therefore, in the next section the noise enhancement factor

()
is determined by simulation. These values are then used
in the analytical expressions to determine
()
and EVM, and
ultimately BER. For this reason, the result for these latter three
quantities are referred to as semi-analytical. However, the
good agreement between the semi-analytical results and the
simulation results in section V provides condence that the
empirical estimates of the noise enhancement factor
()
are
valid.
0 1 2 3 4 5 6 7 8 9 10
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2.4
2.6
Clipping Ratio (dB)

(
L
)

L=1
L=2
L=4
L=8
L=16
L=32
L=40
Fig. 3. The noise enhancement factor
()
vs. different clipping ratios,
when the number of iterations, =1,2,4,8,16,32, and 40.
V. SIMULATION RESULTS
In this section, we verify the expression for asymptotic
iterative clipping and ltering by comparison with simulation
results. We consider a 4-time oversampled OFDM system with
1024 subcarriers and 16 QAM, where iterative clipping and
ltering is used for PAR reduction.
First, we determine the noise enhancement factor
()
for
different clipping ratios when the number of iterations = 1,
2, 4, 8, 16, 32, and 40, as shown in Fig. 3. Note that
()
is dened in (11) as the normalized cross-correlation between
the clipping noise at the -th iteration and the clipping noise
at the rst iteration in the frequency domain. As indicated
in (10), the clipping noise after the -th iteration equals the
sum of the ltered peak cancel signal from the rst iteration
to the -th iteration. From this gure, it is obvious that

(1)
is constant regardless of clipping ratio , and the value

(1)
= 1 is consistent with (11). On the other hand, for
> 1,
()
increases as the clipping ratio increases.
Moreover, the value of
()
at any xed saturates after
about 8 iterations. This saturation of
()
can be explained
by the observation that the envelope power uctuation is
dramatically reduced as the number of iterations increases for
any given clipping ratio , which implies that a target clipping
ratio is achieved with no out-of-band radiation. Thus, Fig. 3
can be regarded as a generalization of the envelope power
distribution trend shown in Fig. 2. To validate the uniqueness
the noise enhancement factor
()
at any xed clipping ratio,
we have investigated the dependence of
()
on the system
parameters: the number of subcarriers , oversampling factor
, and constellation size . As increases ( 256),
()
at any xed clipping ratio converges to a single value.
()
saturates for any xed clipping ratio when the oversampling
factor 4. On other hand,
()
appears to be insensitive
to constellation size. Thus, this letter relies on the assumption
that the number of subcarriers is sufciently large ( = 1024)
and the oversampling factor 4.
In Fig. 4, we compare simulation results with the semi-
analytic results for the attenuation factor,
()
, in (13) for
iterative clipping and ltering as a function of when = 1
1562 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 5, MAY 2010
0 1 2 3 4 5 6 7 8 9 10
0.6
0.65
0.7
0.75
0.8
0.85
0.9
0.95
1
Clipping Ratio (dB)
A
t
t
e
n
u
a
t
i
o
n

F
a
c
t
o
r

(
L
)


Semianalytic (L=1)
Semianalytic (L=40)
Simulation (L =1)
Simulation (L =40)
Fig. 4. Comparison of the semi-analytic results in (13) and the simulation
results for the attenuation factor
()
as a function of when the number
of iterations is 1 and 40.
0 1 2 3 4 5 6 7 8 9 10
70
60
50
40
30
20
10
Clipping Ratio (dB)
E
V
M
(
d
B
)


Semianalytic (L=1)
Semianalytic (L=40)
Simulation (L =1)
Simulation (L =40)
Fig. 5. Comparison of the semi-analytic results in (19) and the simulation
results for EVM as a function of when the number of iterations is 1 and
40.
and 40. The simulation results are closely matched by the
semi-analytic results, which utilize simulation-based values
for the noise enhancement factor
()
. It is shown that the
attenuation factor with = 40 is smaller than that with
= 1 as the clipping ratio decreases. This means that for a
smaller clipping ratio , the constellation of OFDM using the
asymptotic iterative clipping and ltering is attenuated more
than the case with = 1, which results in a reduction of
signal power.
To investigate the impact of asymptotic iterative clipping
and ltering on modulation accuracy, the EVM is plotted
as a function of clipping ratio in Fig. 5, along with the
corresponding simulation results. The measured EVM with
= 1 and 40 is compared to the semi-analytical EVM in
(19), which is obtained from the average PSD of the clipping
noise and the clipping noise enhancement factor. Also, it
can be seen that the agreement between semi-analytical and
simulated EVM is good, thereby validating our asymptotic
0 2 4 6 8 10 12 14 16 18 20
10
5
10
4
10
3
10
2
10
1
10
0
Eb/No (dB)
B
E
R


Simulation (=0dB)
Simulation (=2dB)
Simulation (=4dB)
Simulation (=6dB)
Simulation (=8dB)
Simulation (=10dB)
Semianalytic
Fig. 6. BER vs.

comparison between the semi-analytic results (solid

lines) in (20) and the simulation results for 16 QAM-OFDM in AWGN when
the clipping ratio ranges from 0dB to 10dB, and =40.
iterative clipping and ltering model. Although the in-band
distortion due to clipping is not constant in practice, EVM is
presented in (19) as a signal quality measurement with the
assumption that in-band distortion is constantly distributed
over the bandwidth of interest. Contrary to the attenuation
factor, it can be observed that EVM with = 40 is greater
than that with = 1 at any xed . This implies that the
peak regrowth of iterative clipping and ltering cumulatively
becomes in-band distortion.
To verify our asymptotic iterative clipping and ltering
model, the analytically calculated BER versus

over
an AWGN channel for various values of the clipping ratio
is compared to simulation results in Fig. 6. These simulation
results show good agreement with the semi-analytic BER
analysis (20) for our asymptotic iterative clipping and ltering
model. We observe that at low , it is hard to achieve low
BER because the in-band distortion is dominant, compared to
the additive Gaussian noise

. On the other hand, the in-

band distortion becomes negligibly small for larger , which
allows us to reach the target BER with minimum

. For
example, at the 10
3
BER, the required

is 18.2dB,
11.5dB, 10.75dB, and 10.7dB for = 4dB, 6dB, 8dB, and
10dB, respectively. However, it is impossible for even higher

to reach a target BER of 10

3
with = 0dB and 2dB
due to the distortion associated with severe clipping. It is noted
that our BER performance of asymptotic iterative clipping and
ltering ( = 40) provides the upper bound (worst case) for
a given target clipping ratio, because the lower attenuation
factor in Fig. 4 and the higher in-band distortion in Fig. 5
with = 40 at a given target clipping ratio can lead to a
lower SDR than those of = 1.
We have considered a maximum of = 40 iterations to
ensure asymptotic results, the principal focus of this paper.
However, motivated by potential transmitter complexity issues,
we have considered the effects of using fewer iterations.
Specically, smaller values of lead to improved BER due to
less signal attenuation and less in-band distortion, the latter of
IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 9, NO. 5, MAY 2010 1563
which is achieved at the expense of an increase of out-of-band
radiation. For example, for = 4dB and = 8 iterations, the
decrease with respect to = 40 in

to maintain a BER
of 10
3
is 1dB at the cost of a 2dB increase of the out-of-band
noise power at the normalized frequency /

= 0.3 (in-band
cutoff is /

= 0.25).
VI. CONCLUSION
In this letter, an asymptotic iterative clipping and ltering
model for PAR reduction is presented and its effects are
analyzed in terms of the clipping ratio . This approach makes
use of the clipping noise which is estimated by simulation
from the power spectral density of the clipping noise at the rst
iteration. The semi-analytical results for the attenuation factor,

()
, EVM, and BER were derived and veried by comparison
with simulation experiment results. Thus, these semi-analytical
results can be used to provide a BER performance benchmark-
ing tool for designing PAR reduction techniques using iterative
clipping and ltering in OFDM systems.
REFERENCES
[1] S. H. Han and J. H. Lee, An overview of peak-to-average power
ratio reduction techniques for multicarrier transmission, IEEE Wireless
Commun. Mag., vol. 12, no. 2, pp. 5665, Apr. 2005.
[2] S. H. Leung, S. M. Ju, and G. G. Bi, Algorithm for repeated clipping
and ltering in peak-to-average power reduction for OFDM, Electron.
Lett., vol. 38, no. 25, pp. 17261727, Dec. 2002.
[3] J. Armstrong, Peak-to-average power reduction for OFDM by repeated
clipping and frequency domain ltering, Electron. Lett., vol. 38, no. 8,
pp. 246247, Feb. 2002.
[4] D. Kim and G. L. Stuber, Clipping noise mitigation for OFDM by
decision-aided reconstruction, IEEE Commun. Lett., vol. 3, no. 4-6,
pp. 14771484, Jan. 1999.
[5] H. Chen and A. Haimovich, Iterative estimation and cancellation of
clipping noise for OFDM signals, IEEE Commun. Lett., vol. 7, no. 7,
pp. 305307, July 2003.
[6] R. van Nee and A. de Wild, Reducing the peak-to-average power ratio
of OFDM, in Proc. IEEE Veh. Technology Conf., May 1998, pp. 2072
2076.
[7] M. Friese, On the degradation of OFDM-signals due to peak-clipping
in optimally predistorted power ampliers, in Proc. IEEE Globecom,
Nov. 1998, pp. 939944.
[8] D. Dardari, V. Tralli, and A. Vaccari, A theoretical characterization of
nonlinear distortion effects in OFDM systems, IEEE Trans. Commun.,
vol. 48, no. 10, pp. 1755 1764, Oct. 2000.
[9] P. Banelli and S. Cacopardi, Theoretical analysis and performance of
OFDM signals in nonlinear AWGN channels, IEEE Trans. Commun.,
vol. 48, no. 3, pp. 430441, Mar. 2000.
[10] H. Ochiai and H. Imai, Performance analysis of deliberately clipped
OFDM signals, IEEE Trans. Commun., vol. 50, no. 1, pp. 89101, Jan.
2002.
[11] R. Dinis and A. Gusmao, A class of nonlinear signal-processing
schemes for bandwidth-efcient OFDM transmission with low envelope
uctuation, IEEE Trans. Commun., vol. 52, no. 11, pp. 20092018,
Nov. 2004.
[12] A. R. S. Bahai, M. Singh, A. J. Goldsmith, and B. R. Saltzberg, A
new approach for evaluating clipping distortion in multicarrier systems,
IEEE J. Sel. Areas Commun., vol. 20, no. 5, pp. 10371046, June 2002.
[13] J. E. Mazo, Asymptotic distortion spectrum of clipped, DC-biased,
Gaussian noise, IEEE Trans. Commun., vol. 40, no. 8, pp. 13391344,
Aug. 1992.
[14] X. Li and L. Cimini, Effects of clipping and ltering on the perfor-
mance of OFDM, IEEE Commun. Lett., vol. 2, no. 5, pp. 131133,
May 1998.
[15] C. Tellambura, Computation of the continuous-time PAR of an OFDM
signal with BPSK subcarriers, IEEE Commun. Lett., vol. 5, no. 5, pp.
185187, May 2001.
[16] H. E. Rowe, Memoryless nonlinearities with Gaussian inputs: elemen-
tary results, Bell Syst. Tech. J., vol. 61, pp. 15191525, Sep. 1982.
[17] L. Wang and C. Tellambura, A simplied clipping and ltering tech-
nique for PAR reduction in OFDM systems, IEEE Signal Process. Lett.,
vol. 12, no. 6, pp. 453456, June 2005.
[18] Wireless LAN medium access control (MAC) and physical layer (PHY)
specications: high-speed physical layer in the 5 GHz Band, IEEE Std.
802.11a, Sep. 1999.