The 17th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC'06)

PEAK-TO-AVERAGE POWER RATIO OF SINGLE CARRIER FDMA SIGNALS

WITH PULSE SHAPING
Hyung G. Myung Junsung Lim David J. Goodman
Department of Electrical and Computer Engineering, Polytechnic University
5 MetroTech Center, Brooklyn, NY 11201
USA
{hmyung01; jlim01} @utopia.poly.edu, [email protected]

ABSTRACT
Single carrier frequency division multiple access (SC- Encoder Decoder

FDMA), which utilizes single carrier modulation and

frequency domain equalization is a technique that has similar Serial-to-Parallel Parallel-to-Serial
performance and essentially the same overall complexity as
those of OFDM, in which high peak-to-average power ratio DFT IDFT
(PAPR) is a major drawback. An outstanding advantage of
SC-FDMA is its lower PAPR due to its single carrier Subcarrier Subcarrier De-
structure. In this paper, we analyze the PAPR of SC-FDMA Mapping mapping / *FDE
signals with pulse shaping. We analytically derive the time
domain SC-FDMA signals and numerically compare PAPR IDFT DFT
characteristics using the complementary cumulative
distribution function (CCDF) of PAPR. The results show that Pulse Shaping /
Channel ADC
DAC
SC-FDMA signals indeed have lower PAPR compared to
those of OFDMA. Comparing the two forms of SC-FDMA, *FDE: Frequency Domain Equalization
we find that localized FDMA (LFDMA) has higher PAPR
than interleaved FDMA (IFDMA) but somewhat lower PAPR Fig. 1 Block diagram of SC-FDMA.
than OFDMA. Also noticeable is the fact that pulse shaping
increases PAPR.
DFT
{X k } Subcarrier
{ X }
l
IDFT
I. INTRODUCTION { xn } (N-point) Mapping (M-point) {xm }
Demands for media-rich wireless data services have brought N N M >N M
much attention to high speed broadband mobile wireless T T
T = T ⋅
N T
techniques in recent years. Orthogonal frequency division M
multiplexing (OFDM), which is a multicarrier communication N , M : number of data symbols
technique, has become widely accepted primarily because of T , T : symbol durations
its robustness against frequency selective fading channels Fig. 2 Generation of SC-FDMA transmit symbols. There
which are common in broadband mobile wireless are M total number of carriers, among which N (< M)
communications [1]. Orthogonal frequency division multiple subcarriers are occupied by the input data.
access (OFDMA) is a multiple access scheme which is an
extension of OFDM to accommodate multiple simultaneous
users. X0 X 0
Zeros
X 0
Despite the benefits of OFDM and OFDMA, they suffer a Zeros X0
X1 X1
number of drawbacks including: high peak-to-average power Zeros
ratio (PAPR); a need for an adaptive or coded scheme to X2
overcome spectral nulls in the channel; and high sensitivity to X N −1
frequency offset [2], [3].
Single carrier frequency division multiple access (SC- X N −1 Zeros
Zeros
FDMA), which utilizes single carrier modulation and X M −1 X M −1
frequency domain equalization is a technique that has similar
performance and essentially the same overall complexity as Distributed Mode Localized Mode
those of OFDMA system. One prominent advantage over Fig. 3 Subcarrier mapping modes; distributed and
OFDMA is that the SC-FDMA signal has lower PAPR localized modes.
because of its inherent single carrier structure [4]. SC-FDMA
has drawn great attention as an attractive alternative to
OFDMA, especially in the uplink communications where
lower PAPR greatly benefits the mobile terminal in terms of

1-4244-0330-8/06/$20.002006 IEEE
The 17th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’06)

{ xn } : x0 x1 x2 x3 { xn } x0 x1 x2 x3

 N −1 −j
2π

DFT  X k = ∑ xn e N
nk
, N = 4
 n= 0 

{Xk }: X0 X1 X 2 X3 { x m ,IFDMA } x 0 x1 x2 x3 x 0 x 1 x2 x3 x0 x 1 x 2 x3 x0 x1 x 2 x 3

{ X l , Distributed }: X 0 0 0 0 X1 0 0 0 X 2 0 0 0 X3 0 0 0 { xm , LFDMA } x0 ? ? ? x1 ? ? ? x2 ? ? ? x3 ? ? ?

time
{ X l , Localized }: X0 X1 X 2 X3 0 0 0 0 0 0 0 0 0 0 0 0
Fig. 5 An example of SC-FDMA transmit symbols in the
frequency time domain for N = 4, Q = 4 and M = 16.
Fig. 4 An example of SC-FDMA transmit symbols in the
frequency domain for N = 4, Q = 4 and M = 16.
the DFT outputs of the input data in the localized subcarrier

X l , Distributed denotes transmit symbols for distributed mapping mode. We will refer to the localized subcarrier
subcarrier mapping mode and X l , Localized denotes mapping mode of SC-FDMA as localized FDMA (LFDMA).
The case of M = Q⋅N for the distributed mode with
transmit symbols for localized subcarrier mapping mode. equidistance between occupied subcarriers is called
Interleaved FDMA (IFDMA) [6]. An example of SC-FDMA
transmit power efficiency. It is currently a strong candidate transmit symbols in the frequency domain for N = 4, Q = 4
for uplink multiple access scheme in 3G Long Term and M = 16 is illustrated in Fig. 4. After subcarrier mapping,
Evolution of 3GPP [5]. the frequency data is transformed back to the time domain by
In this paper, we analyze the PAPR of SC-FDMA signals applying inverse DFT (IDFT).
and compare it with that of OFDMA. We analytically derive
the time domain SC-FDMA signals and numerically compare III. PAPR OF SC-FDMA SIGNALS
PAPR characteristics using complementary cumulative In this section, we analyze the PAPR of the SC-FDMA signal
distribution function (CCDF) of PAPR. for each subcarrier mapping mode. For distributed subcarrier
The remainder of this paper is organized as follows: Section mapping mode, we will consider the case of IFDMA. In the
II gives an overview of SC-FDMA. Section III derives the subsequent derivations, we will assume M = Q⋅N and follow
time domain signals for each subcarrier mapping mode of SC- the notations in Fig. 2.
FDMA and it presents the CCDF of PAPR obtained from
Monte-Carlo simulations. Let { xn : n = 0,1," , N − 1} be data symbols to be
modulated. Then, { X k : k = 0,1," , N − 1} are frequency
II. OVERVIEW OF SC-FDMA SYSTEM
domain samples after DFT of { xn : n = 0,1," , N − 1} ,
A block diagram of a SC-FDMA system is shown in Fig. 1.
SC-FDMA can be regarded as discrete Fourier transform { X l : l = 0,1,", M − 1} are frequency domain samples after
(DFT)-spread OFDMA, where time domain data symbols are subcarrier mapping, and { xm : m = 0,1," , M − 1} are time
transformed to frequency domain by DFT before going
through OFDMA modulation. The orthogonality of the users {
symbols after IDFT of X l : l = 0,1, ", M − 1 . The complex }
stems from the fact that each user occupies different passband transmit signal of SC-FDMA x(t) for a block of data
subcarriers in the frequency domain, similar to the case of is represented as
OFDMA. Because the overall transmit signal is a single
carrier signal, PAPR is inherently low compared to the case M −1

of OFDMA which produces a multicarrier signal. x (t ) = e

jω c t
∑ x r (t − mT )
m

Fig. 2 details the generation of SC-FDMA transmit m=0 , (1)

symbols. There are M subcarriers, among which N (< M)
where ωc is the carrier frequency of the system and r(t) is
subcarriers are occupied by the input data. In the time
domain, the input data symbol has symbol duration of T the baseband pulse. In our research, we use a raised-cosine
seconds and the symbol duration is compressed to pulse, which is a widely used pulse shape in wireless
communications, defined as follows in the time domain [7].
T = ( N ) ⋅ T after going through SC-FDMA modulation.
M
There are two methods to choose the subcarriers for  πα t 
cos  
transmission as shown in Fig. 3. In the distributed subcarrier  t
r (t ) = sinc  π 
 T  (2)
mapping mode, DFT outputs of the input data are allocated
 T  22
4α t
over the entire bandwidth with zeros occupying in unused 1−
subcarriers, whereas consecutive subcarriers are occupied by 2
T
The 17th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’06)

where α is the rolloff factor which ranges between 0 and 1. B. Time domain symbols of LFDMA
The PAPR is defined as follows for transmit signal x(t). For LFDMA, the frequency samples after subcarrier mapping
max x (t )
2 { X l } can be described as follows.
peak power of x(t ) 
PAPR = = 0≤t ≤ MT (3)
average power of x (t ) 1 MT  Xl , 0 ≤ l ≤ N −1
∫
2
x(t ) dt X l =  (7)
MT 0 0 , N ≤ l ≤ M − 1
Without pulse shaping, that is, using rectangular pulse Let m = Q ⋅ n + q , where 0 ≤ n ≤ N − 1 , and 0 ≤ q ≤ Q − 1 .
shaping, symbol rate sampling will give the same PAPR as
the continuous case since SC-FDMA signal is modulated over Then,
a single carrier. Thus, PAPR without pulse shaping with m
1 M −1 j 2π

∑ X e
l
symbol rate sampling can be expresses as follows. x m = xQn + q = M
l

2
M l =0
max xm Qn + q
. (8)
m = 0,1,", M −1 1 1 N −1 j 2π l
PAPR =
1
M −1
(4) = ⋅ ∑Xe QN

∑ x
l
2
m
Q N l =0
M m= 0
If q = 0, then,
We first examine the PAPR of transmit symbols for each Qn
block without pulse shaping analytically, and then, investigate 1 1 N −1 j 2π l

the PAPR of transmit symbols with pulse shaping

x m = xQn =
Q
⋅
N
∑X e l
QN

l =0
numerically.
n
1 1 N −1 j 2π

∑
l

A. Time domain symbols of IFDMA = ⋅ X le N

. (9)
Q N l =0
For IFDMA, the frequency samples after subcarrier mapping
{ l}
X can be described as follows. =
1
xn
Q
 Xl/Q , l = Q⋅k (0 ≤ k ≤ N − 1)
X l =  (5) N −1 − j 2π
p
l
0 , otherwise If q ≠ 0, since X l = ∑x e
p= 0
p
N
, then (8) can be

We derive the time symbols { xm } which are obtained by expressed as follows after derivation.
taking inverse DFT of X l . { } 1 j 2π
q
 1 N −1
xp
x m = xQn + q =  1 − e Q
⋅ ∑ (10)
Let m = N ⋅ q + n , where 0 ≤ q ≤ Q − 1 and 0 ≤ n ≤ N − 1 .
Then,
Q  N p=0
1− e
j 2π {
(n− p)
N
+
q
QN
}
1 M −1 j 2π
m
1 1 N −1 j 2π
m As can be seen from (9) and (10), in the time domain,
x m ( = x Nq + n ) = ∑ X l e ∑ X ke
l k
M
= ⋅ N
LFDMA signal has exact copies of input time symbols in the
M l =0 Q N k =0 N-multiple sample positions. In-between values are sum of all
Nq + n the time input symbols in the input block with different
1 1 N −1 j 2π

∑X e
k
= ⋅ N complex-weighting, which would increase the PAPR. An
k
Q N k =0 example of an LFDMA signal is shown in Fig. 5.
. (6)
1 
n
1 N −1 j 2π
C. Numerical results
∑X e
k
= ⋅ k
N

Q N k =0  CCDF (Complementary Cumulative Distribution Function) of
PAPR, which is the probability that PAPR is higher than a
1 certain PAPR value PAPR0 (Pr{PAPR>PAPR0}), is
= xn
Q calculated by Monte Carlo simulation. CCDFs of PAPR for
IFDMA, LFDMA, and OFDMA are evaluated and compared.
The resulting time symbols { xm } are simply a repetition of 105 uniformly random data points were generated to acquire
the CCDF of PAPR. Consecutive chunks were used for
the original input symbols { xn } in the time domain [6].
LFDMA, and in the case of OFDMA, transmission bandwidth
Therefore, the PAPR of IFDMA signal is the same as in the of 5 MHz was assumed and 8 times oversampling was used
case of conventional single carrier signal. An example of an when calculating PAPR [8]. No pulse shaping was applied in
IFDMA signal is shown in Fig. 5. the case OFDMA. In the simulations, the total number of
subcarriers M were set to 256, input data block size N to 64,
and Q to 4. QPSK, 8-PSK, 16-QAM, 32-QAM, and 64-QAM
The 17th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’06)

(a) (a)

(b) (b)

Fig. 6 Comparison of CCDF of PAPR for IFDMA, Fig. 7 Comparison of CCDF of PAPR for IFDMA and
LFDMA, and OFDMA with M = 256, N = 64, and α = LFDMA with M = 256, N = 64, and α of 0, 0.2, 0.4, and
0.5. (a) QPSK. (b) 16-QAM. 0.6, 0.8, and 1. (a) QPSK. (b) 16-QAM.

symbol constellations were considered. Raised cosine pulse Fig. 7 shows the impact of rolloff factor α on the PAPR
r(t) was truncated from -6 T to 6 T time period and it was when using raised cosine pulse shaping, where this impact is
oversampled by 8 times [7]. more obvious in IFDMA. As the rolloff factor increases from
In Fig. 6, plots of CCDF of PAPR for IFDMA, LFDMA, 0 to 1, PAPR reduces significantly for IFDMA. This implies
and OFDMA are shown. We compare the PAPR value that is that there is a tradeoff between PAPR performance and out-
exceeded with probability less than 0.1% (Pr{PAPR>PAPR0} of-band radiation since out-of-band radiation increases with
= 10-3), or 99.9-percentile PAPR. First, in the case of no pulse increasing rolloff factor.
shaping, IFDMA has lower PAPR than the case of OFDMA
by 10.5 dB for QPSK and by 7 dB for 16-QAM, while PAPR Table 1. Comparison of 99.9-percentile PAPR
of LFDMA is lower than that of OFDMA by 3 dB for QPSK
and by 2 dB for 16-QAM but higher than that of IFDMA by Mod. IFDMA LFDMA OFDMA
format No Pulse Pulse No Pulse Pulse
7.5 for QPSK and by 5 dB for 16-QAM. With raised-cosine pulse shaping shaping pulse shaping shaping
pulse shaping with rolloff factor of 0.5, it can be seen that shaping (rolloff (rolloff shaping (rolloff (rolloff
PAPR increases significantly for IFDMA whereas PAPR of 0.5) 0.22) 0.5) 0.22)
LFDMA hardly increases. Full comparison of 99.9-percentile QPSK 0 dB 4.3 dB 6.1 dB 7.5 dB 7.6 dB 7.6 dB 10.7 dB
8PSK 0 dB 4.2 dB 5.9 dB 7.4 dB 7.5 dB 7.5 dB 10.6 dB
PAPR for different modulation formats is in Table I. We can 16QAM 3.5 dB 6.6 dB 7.7 dB 8.4 dB 8.4 dB 8.5 dB 10.5 dB
see that IFDMA and LFDMA have lower PAPR than 32QAM 3.4 dB 6.4 dB 7.5 dB 8.2 dB 8.3 dB 8.4 dB 10.6 dB
OFDMA consistently. 64QAM 4.8 dB 7.1 dB 8.0 dB 8.6 dB 8.7 dB 8.7 dB 10.5 dB
The 17th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’06)

IV. CONCLUSION
In this paper, we analyze the PAPR of SC-FDMA signals and
compare it with the case of OFDMA. Specifically, we derive
the time domain signals of IFDMA and LFDMA, and
numerically compare PAPR characteristics using CCDF of
PAPR. It is shown that SC-FDMA signals indeed have lower
PAPR compared OFDMA. Also, we have shown that
LFDMA incurs higher PAPR compared to IFDMA, but
compared to OFDMA, it is lower, though not significantly.
Another noticeable fact is that pulse shaping increases PAPR
and that rolloff factor in the case of raised-cosine pulse
shaping has a significant impact on PAPR of IFDMA. A
pulse shaping filter should be designed carefully in order to
reduce the PAPR without degrading the system performance.
In conclusion, to fully exploit the low PAPR advantage of
SC-FDMA, IFDMA is more desirable than LFDMA when
choosing subcarrier mapping method.

REFERENCES
[1] L. J. Cimini, Jr., “Analysis and Simulation of a Digital Mobile Channel
Using Orthogonal Frequency Division Multipexing,” IEEE Trans.
Commun., vol. 33, no. 7, July 1985, pp. 665–675.
[2] H. Sari, G. Karam and I. Jeanclaude, “An Analysis of Orthogonal
Frequency-Division Multiplexing for Mobile Radio Applications”, Proc.
IEEE VTC ’94, pp. 1635–1639, Stockholm, Sweden, June 1994.
[3] H. Sari, G. Karam and I.Jeanclaude, “Frequency-Domain Equalization of
Mobile Radio and Terrestrial Broadcast Channels,” Proc. IEEE
GLOBECOM ’94, pp. 1–5, San Francisco, CA, Nov. 1994.
[4] D. Falconer, S. L. Ariyavisitakul, A. Benyamin-Seeyar and B. Eidson,
“Frequency Domain Equalization for Single-Carrier Broadband Wireless
Systems,” IEEE Commun. Mag., vol. 40, no. 4, April 2002, pp. 58–66.
[5] 3rd Generation Partnership Project (3GPP); Technical Specification
Group Radio Access Network; Physical Layer Aspects for Evolved
UTRA, http://www.3gpp.org/ftp/Specs/html-info/25814.htm
[6] U. Sorger, I. De Broeck and M. Schnell, “Interleaved FDMA - A New
Spread-Spectrum Multiple-Access Scheme,” Proc. IEEE ICC ’98, pp.
1013–1017, Atlanta, GA, June 1998.
[7] T. S. Rappaport, Wireless Communications: Principles and Practice,
Second Edition. Prentice Hall, 2002
[8] S. H. Han and J. H. Lee, “An Overview of Peak-to-Average Power Ratio
Reduction Techniques for Multicarrier Transmission,” IEEE Wireless
Commun., vol. 12, no. 2, April 2005, pp. 56–65.