Analyse Performance of Fractional Fourier Transform on Timing and Carrier Frequency Offsets Estimation

International Journal of Wireless & Mobile Networks (IJWMN) Vol. 8, No. 2, April 2016
DOI : 10.5121/ijwmn.2016.8201 1
ANALYSE PERFORMANCE OF FRACTIONAL
FOURIER TRANSFORM ON TIMING AND CARRIER
FREQUENCY OFFSETS ESTIMATION
Cherif Rezgui
Communication Systems Lab, National School of Engineers of Tunis (ENIT), Tunisia
ABSTRACT
This paper deals with the performance of the use of fractional Fourier transform (FRFT) instead of
conventional Fourier transform (FFT) in either symbol timing offset (STO) and carrier frequency offset
(CFO) estimation. Orthogonal frequency division multiplexing is widely used in many systems due to
advantages of theses technique compared with mono-carrier systems. In spite of his advantages, OFDM
presents drawbacks such as sensitivity to timing and frequency offsets. Many techniques are used in the
literature to estimate these two parameters in order to compensate them (synchronization task). These
techniques used conventional Fourier transform. In this paper, we are interested in estimating STO and
CFO using fractional Fourier transform. Monte Carlo simulation demonstrates the performance of the use
of FRFT instead of FFT.
KEYWORDS
FRFT, STO, CFO, CAZAC sequences, MSE, SNR
1. INTRODUCTION
The principle of OFDM is to divide, on a wide number of subcarriers, the digital signal that is to
be transmitted. In wireless systems, transmitted data reaches to receivers after transit through a
radio channel. For conventional coherent receivers, the effect of channel on transmitted signal
must be estimated to recover transmitted data.
The signal is spread in frequency, time, and angle; therefore, the received signal suffers from the
impact of these spreads on transmitted signal. In spite of many advantages of OFDM compared to
single carrier systems, it has major drawbacks such carrier frequency and symbol timing offsets.
So, estimating these two parameters (STO and CFO) is very important. Timing and carrier
frequency synchronizations are the major task of OFDM receiver.
Various researches had been proposed to estimate CFO [2-6]...Other techniques are used in the
literature to estimate STO [7-11].... These techniques, for both STO and CFO estimation, use
conventional Fourier transform (FFT). This article is about estimating CFO and STO for known
method with the use of fractional Fourier transform (FRFT) to illustrate the importance of FRFT
compared to FFT.

2
The remainder of this paper is organized as follows. In section 2, we introduce OFDM signal and
the effect of CFO. Section 3 gives the effect of STO. Then, we present Fractional Fourier
Transform in section 4. Thereafter, Section 5 shows the proposed method and results of
simulations between the use of FFT and FRFT in terms of MSE. Conclusion is given in the last
section.
2. OFDM BASIC AND CFO
OFDM signal is the sum of many independent signals modulated onto sub-channels of equal
bandwidth. Let us define N symbols in OFDM as {Xn, n=0,1,…,N } The complex baseband
representation of a multicarrier signal consisting of N subcarriers is given by :
(1)
Where j = √-1, ∆f is the subcarrier spacing, l the lth
OFDM symbol and NT denotes the useful
data block period. The subcarriers in OFDM system are supposed to be orthogonal to each other.
In order to demodulate an OFDM symbol correctly at the receiver using N point DFT, it is very
much required to take exact samples of transmitted OFDM symbol. The signal x(t) is modulated
by a carrier frequency f0 in the transmitter, then demodulated at the reception at the same
frequency. This is the ideal case where the frequency of modulation and demodulation are
identical. But this case does not happen in practice because these frequencies are derived from
different oscillators (at the transmitter and receiver) which each have defects. Note that the
following assumption has been made implicitly in the baseband model. Let fc
t
and fc
r
denote the
carrier frequencies in the transmitter and receiver, respectively. Let foffset denote their difference
(i.e., foffset = fc
t
- fc
r
). Meanwhile, Doppler frequency fd is determined by the carrier frequency fc
t
and the velocity v of the terminal (receiver) as:
(2)
The normalized carrier frequency offset (CFO), ε, is defined as:
The time-domain received signal, with the presence only of CFO (no Phase Noise), can be written
as:
(3)
where Xl(k) and Hl(k) denote the kth
subcarrier frequency components of the lth
transmitted
symbol and channel frequency response in the frequency domain respectively, and zl(n) noise in
the time domain.
The table 1 presents impact of CFO in time and frequency domain signal. The frequency shift of
–ε in the frequency-domain signal X(k) subjects to the CFO of ε and leads to an inter-carrier
interference (ICI), which means a subcarrier frequency component is affected by other subcarrier
frequency components. It is therefore essential to estimate it in order to resolve problems such as
Inter-Carrier Interference.

3
Table 1: The effect of CFO on the received signal, [1]
Received signal Effect of STO δ on the received signal
Time-domain signal y(n) ej2πkε/N
x(n)
Frequency-domain signal Y (k) X(k-ε)
The most known methods that estimate CFO are Moose [2] and Classen [3] mehods. It was
demonstrated in the literature that Classen method is better than the one done by Moose. That’s
why we are interested in estimating CFO using Classen method but the use of FRFT. The main
idea of Classen is: pilot tones can be inserted in the frequency domain and transmitted in every
OFDM symbol for CFO tracking. First, two OFDM symbols,yl(n) and yl+D(n) are saved in the
memory after synchronization. Then, the signals are transformed into Yl(K)N-1
k=0 and Yl+D(k)N-
1
k=0 via FFT, from which pilot tones are extracted. After estimating CFO from pilot tones in the
frequency domain, the signal is compensated with the estimated CFO in the time domain. In this
process, two different estimation modes for CFO estimation are implemented: acquisition and
tracking modes. In the acquisition mode, a large range of CFO including an integer CFO is
estimated. In the tracking mode, only fine CFO is estimated. The integer CFO is estimated by:
(4)
where L, p(j), and Xl(p(j))denote the number of pilot tones, the location of the jth
pilot tone, and
the pilot tone located at p(j) in the frequency domain at the lth
symbol period, respectively. D is an
integer that represents the ratio of the OFDM symbol length to the length of a repetitive pattern.
3. EFFECT OF STO
In order to correctly demodulate an OFDM symbol at the receiver using N point DFT, it is very
important to take exact samples of transmitted OFDM symbol. The correct starting point of DFT
window is needed to preserve the orthogonality among the sub-carriers. OFDM systems present
many advantages in comparison with single carrier system however all these advantages can be
relevant only when the orthogonality among sub-carriers is kept. If one DFT window takes
sample of two different OFDM symbol generating ICI and ISI. Table 2 shows the effect of timing
offset in the received signal in time and frequency domain the effects of channel and noise are
neglected for simplicity of exposition. The presence of STO in the time domain causes phase
offset of 2πkd/N in the frequency domain. This phase offset is proportional to the subcarrier index
k as well as the STO δ,, Table 2.
Table 2: The effect of STO on the received signal, [1]
Received signal Effect of STO δ on the received signal
Time-domain signal y(n) x(n + δ)
Frequency-domain signal Y (k) ej2πkδ
/N X(k)
Four possible cases may occur, Figure 1. The first case occurs when there is no timing error (i.e.
timing offset δ = 0). That is the case that estimated starting point of the OFDM symbol coincides
with the exact time, preserving the orthogonality among sub-carriers, so the OFDM symbol can
be perfectly left without any type of interference.The second case takes place when the estimated

4
starting point of the OFDM symbol before the exact point, but after the end of the (displaced)
channel response of the precedent OFDM symbol. In that case the ith
symbol not overlapped with
the preceding (i-1) th
OFDM symbol, that is, without incurring ISI by the preceding symbol in
such cases. The third one is happen if the starting point of the OFDM symbol is estimated exist
prior to the end of the (displaced) channel response of the previous OFDM symbol and thus the
symbol synchronization is too soon to avoid the ISI. In that case the orthogonality between the
components of subcarrier is destroyed by the ISI (the previous symbol) and furthermore, ICI
occurs. The fourth case occurs if the starting point of the OFDM symbol is estimated just after the
exact time. In this case, the samples for current FFT operation interval is consists of a part of the
current OFDM symbol xl(n) and a part of the following OFDM symbol xl+ 1 (n).
Figure 1: Four different cases symbol starting point subject to STO.
Many techniques in the literature are implemented using whether cyclic preﬁx or training
sequence. As an example of estimating STO using cyclic preﬁx is done by Tourtier, P.J.,
Monnier, R., and Lopez, P. [7].
Figure 2: STO estimation technique using double sliding windows.
The estimated STO is done by:
(5)
Another technique used in [11]. It consists on minimizing the squared difference between a NG
sample block (seized in window W1) and conjugate of another NG sample block (seized in
window W2).
(6)

5
Estimation techniques using training sequences are presented in Figure 3:
Figure 3: STO estimation using the repetitive training symbol, (period = Tsub/2).
In this case, STO is estimated as, [9] and [10]:
(7)
In this paper, STO is estimated in frequency domain. As implied in second case, the received
signal subject to STO suffers from a phase rotation, Figure 4:
Figure 4 : Effect of STO in frequency domain
STO is estimated as:
(8)
Where is deﬁned as:
(9)
4. FRACTIONAL FOURIER TRANSFORM
Time and Frequency form two orthogonal axes in the time-frequency plane. As the conventional
Fourier transform is to reformulate a defined signal either in the time domain as a function of
frequency, it can be seen as a rotation of π/2 in the trigonometric direction in the time-frequency
plane. The fractional Fourier transform is intended as a generalization of the classic Fourier
transform .It could be rotated to any angle α in the plane time-frequency. For each α value, FRFT
is a different representation of the original signal function of time. So actually, the FRFT
corresponds to a one parameter family of transformation. For α = 0, the FRFT coincides with the
identity transformation which restores the original signal in the time domain. For α = π/2, it yields
to the conventional Fourier transform. Mathematically, we defined the fractional

6
Fourier transform a signal s(t) is defined as follows :
Where p is a real number known as FRFT order, α=pπ/2 is the angle of FRFT, and Kα(t,u)
is the kernel of FRFT:
Fractional Fourier transform is used in many field such in [12].
5. PROPOSED METHOD AND SIMULATION RESULTS
The performances of FRFT are shown in this section. For CFO estimation, we use the same
technique done by Classen [3] with FRFT and CAZAC sequences inserted in comb-type. For
STO estimation, we use Shi & Serpedin method [11] but with the use of FRFT instead of FFT.
Table 3 shows the simulation parameters.
Table 3: The parameters for simulation
Modulation 16 QAM
Number of sub-carrier 128
Cyclic Preﬁx Length (Ng) 32
Pilot sequences CAZAC sequences
Number of Bits per Symbol 4
Pilot Spacing (Nps) 4
CFO 0.5
STO -3
Number of iteration 100
Signal to Noise Ratio (SNR) range 0:30
It is clearly shown in figure 5 that the use of fractional Fourier Transform with Classen method
gives better results in term of MSE compared to Classen method using conventional Fourier
Transform and other known techniques (CP based and Moose technique).
It is also shown in figure 6 that the method done by Shi & Serpedin with FRFT is more efficient
than the one done with FFT. Graphs show the efficiency of using FRFT compared to the most
known method (Schmidt and Cox).

7
Figure 5: MSE of CFO of most known method (CP based, Moose method, Classen with FFT) and one done
with FRFT.
To more explain the effect of fractional Fourier transform of CFO, we draw the graph for
different order of FRFT, figure 6.
Figure 6: MSE of CFO for FFT and FRFT for different order.
Figure 6 shows the effect of FRFT order on MSE of carrier frequency offset. It shows that when
the order becomes small (i.e. it tends to identity) the MSE becomes better and when the order is
great (i.e. it tends to FFT) the results becomes worse. For example, when the order is equal to
0.01 and for SNR equal to 15 dB, we obtain a gain equal to 6 dB.

8
Figure 7: MSE of STO for Scmidt & Cox, Shi & Serpedin and Shi & FFT with CAZAC & FRFT with
CAZAC sequences.
It is also shown in figure 7 that the method done by Shi & Serpedin with FRFT is more efficient
than the one done with FFT. Graphs show the effiency of using FRFT compared to the most
known method (Schmidt and Cox).
Figure 8: MSE of timing offset for FFT and FRFT for different order.
Figure 8 shows the effect of FRFT order on MSE of carrier frequency offset. It shows that when
the order becomes small (i.e. it tends to identity) the MSE becomes better and when the order is
great (i.e. it tends to FFT) the results becomes worse. For example, when the order is equal to
0.01 and for SNR equal to 15 dB, we obtain a gain equal to 9 dB.

9
From these results, we conclude that FRFT is an efficient method for both CFO and STO
estimation. For Further research in timing and frequency synchronization, we could use FRFT
instead of FFT for better results.
6. CONCLUSIONS
This paper proposes a study for performance analysis for symbol timing offset (STO) and carrier
frequency offset (CFO) estimation. This study compares Fractional Fourier transform to the
conventional Fourier transform. The study uses most known methods for estimating CFO and
STO with the use of FRFT and CAZAC sequences as pilot. The study demonstrates the efficiency
of fractional Fourier transform. This study gives a good candidate (FRFT) for OFDM
synchronization with the same complexity of conventional FFT.
REFERENCES
[1] Cho, Y.S., Kim J., Yang, W.Y. and Kang C. G., ”MIMO-OFDM wireless communications with
MATLAB”, John Wiley and Sons, Asia, IEEE Press, 2010, pp.153-161.
[2] MOOSE P H, ”A technique for orthogonal frequency division multiplexing frequency offset
correction,” IEEE Transactions on Communications. 1994, 42 (10), pp. 2908-2914.
[3] Classen, F. and Myer, H. ”Frequency synchronization algorithm for OFDM systems suitable for
communication over frequency selective fading channels,” IEEE VTC94, pp. 1655-1659.
[4] Schmidl T M, Cox D C, ”Robust frequency and timing synchronization for OFDM,” IEEE
Transactions on Communications. 1997, 45 (12), pp.1613-1621.
[5] Chun linYan, Shaoqian Li, Youxi Tang, Xiao Luo, and Jiayi Fang, ”A Novel Frequency Offset
Estimation Method for OFDM Systems with Large Estimation Range,” IEEE Transactions on
Broadcasting. 2006, 52(1), pp. 058-061.
[6] Wen-Jeng Lin; I-Cheng Liu; Jung-Shan Lin ”Novel frequency offset estimation for IEEE 802.11p
standard in V2V channels”, ITS Telecommunications (ITST), 2012 12th International Conference on,
On page(s):264 - 268.
[7] Tourtier, P.J., Monnier, R., and Lopez, P. ”Multicarrier modem for digital HDTV terrestrial
broadcasting” .Signal Process., 5(5), 379403, 1993.
[8] Speth,M., Classen, F., and Meyr,H. ”Frame synchronization of OFDM systems in frequency selective
fading channels”. IEEE VTC97, pp.18071811, 1997.
[9] T. M. Schmidl and D. C. Cox , ”Robust Frequency and Timing Synchronization for OFDM”, IEEE
Trans. on Commun. , pp.1613 -1621 , 1997.
[10] D. C. Schmidl and D. C. Cox , ”Low-overhead, low-complexity [burst] synchronization for OFDM” ,
IEEE Int. Conf. on Commun , vol. 3 ,pp.1301 -1306, 1996.
[11] K. Shi and B. E. Serpedin. ”Coarse frame and carrier synchronization of OFDM systems: A new
metric and comparison”, IEEE Trans. Wireless Commun., vol. 3, no. 4, pp. 12711284, Jul. 2004.
[12] Massimiliano (Max) Martone , ”A Multicarrier System Based on the Fractional Fourier Transform
for Time-Frequency-Selective,” IEEETransactions on Communications, VOL. 49, NO. 6, June 2001.
[13] Yushi Shen and Ed Martinez, ”Channel Estimation in OFDM Systems”, Freescale Semiconductor,
Inc., 2006.
[14] Coleri, S., Ergen, M., Puri, A., and Bahai, A.,”Channel Estimation Techniques Based on Pilot
Arrangement in OFDM Systems,” IEEE Transactions on Broadcasting, vol.48, pp.223229, Sept.
2002.
[15] Sinem Coleri, Mustafa Ergen,Anuj Puri, Ahmad Bahai, ”A Study of Channel Estimation in OFDM
Systems”, IEEE 56th Vehicular technology conference.
[16] Minn, H. and Bhargava, V.K. (1999), An investigation into time-domain approach for OFDM channel
estimation. IEEE Trans. on Broadcasting,45(4), 400409.

10
Author
Cherif Rezgui is graduate astelecommunication Engineer from higher school of
communication (Sup’Com) in 2004, his master degree in 2011 from national school of
engineers of Tunis. He is actually working on Ph.D. He is working in the field of MIMO,
OFDM, Chanel estimation and synchronization.

Analyse Performance of Fractional Fourier Transform on Timing and Carrier Frequency Offsets Estimation

More Related Content

What's hot (20)

Viewers also liked (15)

Similar to Analyse Performance of Fractional Fourier Transform on Timing and Carrier Frequency Offsets Estimation (20)

Recently uploaded (20)

Analyse Performance of Fractional Fourier Transform on Timing and Carrier Frequency Offsets Estimation