978-1-4799-5835-1/14/$31.00 @2014 IEEE 1170
The 2014 7th International Congress on Image and Signal Processing
Carrier Frequency Offset Estimation of OFDM
Systems Based on Complementary Sequence
Zhenyu Zhang
1,2
, Lijia Ge
2
, Fengchun Tian
1
, Fanxin Zeng
2
, Guixin Xuan
2
1. College of Communication Engineering, Chongqing University, Chongqing, China;
2. Chongqing Key Laboratory of Emergency Communications, Chongqing Communication Institute, Chongqing, China
Abstract—As a key point of orthogonal frequency division
multiplexing (OFDM) technology, carrier frequency offset (CFO)
has a significant effect on OFDM system performance. This
paper proposes a CFO estimation method by using
complementary sequence (CS) as training sequence. The
proposed CS-based method can operate integer CFO and
fractional CFO estimations on the basis of the same training
sequence. The fractional CFO can be effectively estimated by
exploiting repetition properties of training symbol in time
domain, which is similar to traditional methods. In order to
increase the acquisition range and estimation accuracy of integer
CFO, we design aperiodic zero correlation zone for training
sequence in frequency domain. Combining the ideal auto-
correlation property of CS with zero-inserting operation, integer
CFO estimation can be perfectly performed. Simulation results
show that mean square error of CFO estimation is close to
Cramer-Rao lower bound for the fixed detection threshold and
better performance can be obtained for flexible detection
thresholds at different signal-to-noise ratios.
Keywords-OFDM; carrier frequency offset estimation;
complementary sequence; zero correlation zone
I. INTRODUCTION
Orthogonal frequency division multiplexing (OFDM) is a
multi-carrier modulation scheme to effectively resist frequency
selective fading channel and has been widely used to high-
data-rate wireless transmission, e.g., WLAN, WiMAX and
LTE [1]. One of the main shortcomings of OFDM modulation
is that it is quite sensitive to carrier frequency offset (CFO)
caused by Doppler shifts or oscillator inaccuracies. In terms of
sub-carrier spacing scale, the CFO can be divided into two
parts, namely integer CFO and fractional CFO. The integer
CFO will generate cyclic shift of the modulated data and lead
to detection error at receiver while the fractional CFO will
cause inter-carrier interference (ICI) and signal-to-noise ratio
(SNR) degradation. Then, suitable methods should be used to
correctly estimate and compensate CFO.
Based on repetition properties of training sequence in time
domain, Schmidl and Cox (S&C) presented a robust estimation
algorithm on CFO for OFDM systems [2], where the number
of the identical parts in one training symbol is equal to 2.
Morelli and Mengali (M&M) extended the S&C algorithm by
dividing one training symbol into more identical parts [3]. In
addition to similar mean square error (MSE) performance,
M&M algorithm can obtain larger estimation range on the
basis of just one training symbol. Combining the repetition
property of training symbol with its symmetric property, two
modified structures of training symbol and corresponding
estimation methods were presented [4-5], where only one
training symbol was needed for each method. To further
enlarge estimation range of CFO, a novel method based on
envelop equalized processing was proposed [6] and the
acquisition range can be as large as the bandwidth of the
OFDM signal. In addition to the above joint estimations of
fractional CFO and integer CFO, some separate estimations in
the presence of fractional CFO were researched, e.g., residual
CFO estimations [7-9] and integer CFO estimations [10-13].
For CFO estimation of OFDM systems, a training sequence
in frequency domain plays an important role. Generally, most
estimation methods employed pseudo-noise (PN) sequence as
training sequence, such as the cases in [2-5, 14]. Besides, some
researches focused on special sequences with better correlation
properties, such as constant amplitude zero auto-correlation
(CAZAC) sequence [15-16], almost-perfect auto-correlation
sequence [17] and extended m-sequence [18]. Based on
systematic design methods, these special training sequences
possess ideal or almost ideal periodic correlation property
which is helpful to improve integer CFO estimation
performance in frequency domain. However, the crucial factor
for frequency-domain integer CFO estimation is the aperiodic
auto-correlation function (ACF) of training sequence, not its
periodic ACF. Training sequences with ideal periodic ACF
property can not assure that their aperiodic ACF properties are
also ideal. Aiming at this problem, we construct frequency-
domain training sequence with excellent aperiodic ACF
property in this paper. Based on complementary sequence (CS),
an aperiodic zero correlation zone (ZCZ) is designed and
aperiodic ACF sidelobes in ZCZ are always equal to zero,
which will greatly increase the frequency-domain estimation
performance of integer CFO.
After analyzing the effect of CFO on OFDM systems, we
present the CS-based CFO estimation scheme in Section III. In
order to show the performance of the proposed method, the
corresponding simulation and analyses are provided in Section
IV. Finally, Section V summarizes the results.
II. OFDM
MODULATION AND ITS PERFORMANCE
DEGRADATION CAUSED BY CFO
The baseband OFDM modulation/demodulation can be
realized by means of inverse discrete Fourier transform
(IDFT)/DFT which decreases operation complexity on the
This work was supported in part by National Natural Science Foundation of
China (NSFC) Grant #61471366, #61002034, #61271251 and #61271003,
China Postdoctoral Science Foundation Grant #2014M552318, Natural
Science Foundation Project of CQ Grant #cstc2014jcyjA40050, Chongqing
Postdoctoral Science Special Foundation Grant #Xm2014031, Open Fund of
Chongqing Key Lab of Mobile Communications Technology, Program for
Innovative Research Team in University of Chongqing #KJTD 201343.