You are on page 1of 4

A New Symbol Timing and Frequency

Synchronization Design for


OFDM-based WLAN Systems
Taekyu Kim and Sin-Chong Park
School of Engineering,
Information and Communications University
119, Munjiro, Yuseong-gu, Daejeon, 305-732, Korea

Abstract - In this paper, we present a new symbol timing long training symbol, T1 in Fig. 1. Therefore, the timing
and frequency synchronization method for OFDM WLAN synchronization methods which use the long training symbol
system. A new method for synchronization using preamble or need IFFT operation for CIR are excluded from the
structure of the IEEE 802.11a is proposed. The consecutive discussion. The proposed method for timing synchronization
preamble structure is exploited to find the symbol starting- is accomplished sequentially by packet detection, fine and
point and frequency offset accurately even under low SNR. The coarse time synchronization using only cross-correlation. In
proposed timing synchronizer performs the fine timing this paper, the performances of conventional and proposed
synchronization first, and recognizes the end of short training synchronizer are compared in terms of synchronization failure
symbol by coarse timing synchronization. The new frequency rate.
synchronization which means the estimation of frequency offset
is more accurate than conventional method by considering the The frequency synchronization, which means the
amplitude of preamble samples. Simulations show that it has frequency offset estimation in this paper, is critical since any
distinct advantage over the conventional methods under multi- frequency offset causes a loss of inter-subcarrier
path wireless channel condition even though it uses the fixed orthogonality which results in inter-carrier interference (ICI)
threshold for decision. [6]. In this paper, an efficient frequency offset estimation
scheme for IEEE 802.1 la WLAN is proposed. The estimator
Keywords OFDM, timing offset, frequency offset exploits the consecutive preamble pattern as the conventional
estimation, synchronization method [7][8], but proposed estimator measures the
frequency offset considering the amplitudes of each preamble
1. Introduction sample. The larger amplitude's sample get the more weight
than the smaller one in the frequency offset estimation
process since smaller amplitude's samples of preamble are
Orthogonal frequency-division multiplexing (OFDM) is more unreliable even though they are under the same channel
an effective modulation technique to mitigate the inter- condition.
symbol interference (ISI) caused by multi-path propagation. It
has been adopted for high-bit rate wireless local area The rest of the paper is organized as follow: Section II
networks (WLANs) standards such as IEEE802.1 la [1] due gives a brief description of OFDM WLAN system and the
to its high data rate transmission and its robustness to multi- frame structure use for synchronization. The conventional
path propagation channels. and proposed timing and frequency synchronization
algorithms are described in Section III and IV, respectively
However, OFDM systems are extremely sensitive to and our simulation results are discussed in Section IV.
receiver synchronization imperfection, which can cause the Finally, Section V gives the conclusion.
degradation of system performance. Since these properties of
OFDM systems, synchronization at the receiver side is one of
the most critical issues. A number of symbol timing 2. OFDM-WLAN Preamble Structure
synchronization methods using auto-correlation or cross-
correlation algorithms of the received preamble have been
developed and proposed. The conventional symbol timing The 802.11 a standard defines the preamble as shown in
synchronization schemes for IEEE802.1 la WLAN systems Fig. 1. Every frame has the preamble. The preamble consists
use short training symbols to estimate a coarse symbol time of ten identical short training symbols (t1,t2,. ..,t10) and two
acquisition via auto-correlation, and use long or short . S+ti-...~~16,
symbols to find a fine symbol time acquisition via cross-
correlation as in [2-4]. There is also another scheme using 4
, --I ,4

CIR (channel impulse response) for fine symbol time [5].


Since the 802.1 la protocol allows long training symbols
to be utilized for channel estimation [1], not only coarse but Figure. 1. Downlink long preamble structure.
fine time synchronization should be completed before the first

ISBN 978-89-5519-131-8 93560 - 1669 - Feb. 12-14, 2007 ICACT2007


Authorized licensed use limited to: UNIVERSITY OF WESTERN ONTARIO. Downloaded on March 08,2010 at 13:12:48 EST from IEEE Xplore. Restrictions apply.
identical long training symbols (T1, T2). Each short and long 2) Fine time synchronization Cross-correlation
training symbol consists of 16 and 64 samples, respectively technique is used for fine time synchronization. It utilizes
[1]. In addition, the GI field represents a guard interval for the end of the short training symbol for correlation from the
the long symbol. It is a replicated filed consisting of the last position obtained from the coarse time synchronizer. The
32 samples of the long symbol. start position of the next OFDM symbol, or long training,
There are in the literature several algorithms for time and can be estimated by:
frequency synchronization based on the correlation of known Lf
sequences and the repetition of short training symbol. On the
other hand, the long training symbol is used for channel Trine = arg max mmLf +Lf (5)
estimation and fine frequency offset correction. Last symbol
in the header is a special symbol called SIGNAL which
carries information about length of the packet as well as the
modulation type of the encoded data. (coare + 160 -Lf < n<Tcoarse+160)
where Trcoarse iS the resultant sample position estimated by the
3. Symbol Timing Synchronization coarse time synchronizer and Lf is the window size for cross-
correlation. And, Sk denotes one short training symbol.
Conventional Method
1) Coarse time synchronization Based on algorithm in Proposed Method
[2], if the first half part of the training symbol is identical in Proposed time synchronization method is accomplished
time-domain to the second part except for the phase shift sequentially by packet detection, fine and coarse time
caused by the carrier frequency offset, we are able to synchronization. These all operations are accomplished by
determine the first sample of the packet by using the one correlator which performs the cross-correlation between
following metric: the received samples and known transmitted short training
symbol. It can be described by the following equation:
M(n) = R(n)12 (1) 15 2
(P(n))2 E (rn+MSM ) (6)
where C(n) = m=O15
LI
L rn+m 12
m=O
R(n) = Z (rfmrmL) and (2)
m=O C(n), the normalized cross-correlation metric, has ten
peak values since the preamble has ten short training symbol.
LI 2
P(n) = E |rn+m+L (3) 3) Packet detection : In a packet detection process, the
m=O
receiver decides the existence of the packet by comparing the
Here, r, is the time-domain n-th received sample, * C(n) with a threshold, qp. Let the instant at which the
denotes complex conjugate and LC is the number of sample in packet detector recognizes be called Tpd.
the first half part of the symbol. LC can be fixed to 16 or 32 4) Fine time synchronization Because the peak values
depending on how long the correlation window is desired. of C(n) exist every 16 samples as shown in Fig. 6, the peak
Since this LC means the auto correlation window size, the position indicates the start point of each short training
larger window, the more stable the estimation. However, too symbol. Fine time synchronization process search this time-
large window shortens the time duration for AGC operation, offset of the peak positions which can be estimated using
and it causes the unstable estimation of coarse frequency following equations:
offset and fine timing synchronization. (7)
Tfine offset = arg max [ f (k)]
The metric can be interpreted as a normalized auto-
correlation of the training symbol. It can be observed that the where
metric follows a curve starting with a ramp that finishes in a LT,,,,,, /1 6
(8)
flat region and at the end the metric falls down. The coarse
timing synchronization, or packet detection, operation is to
f(k)= I
i=l T,, / 16
C(16i+k) O<k<15-
find out the instant at which this metric, M(n), start to be
lager then the threshold, 7coarse as Tcoarse denotes the instance founded by the following
coarse time synchronizer. LAi means the largest integer
Tcoarse = arg min{ M(n) > coarse} * (4) number smaller than n. Equation (8) gives the non-coherent
summation of correlation values, which can mitigate the
distortion of peak positions due to channel noise and

ISBN 978-89-5519-131-8 93560 - 1670 - Feb. 12-14, 2007 ICACT2007


Authorized licensed use limited to: UNIVERSITY OF WESTERN ONTARIO. Downloaded on March 08,2010 at 13:12:48 EST from IEEE Xplore. Restrictions apply.
frequency offset. Tfi, o.ffset is a kind of phase offset number
from 0 to 15 within a short preamble symbol.
cor. detect
5) Coarse time synchronization: In coarse time 1Pd
synchronization process, receiver obtains the information
Fine time TI
whether the short training symbol is finished or not. This * --4
sync.
information can be acquired by comparing the correlation
energy in the fine timing, phase offset among the 16 samples, Tfileqoff,;et,
which is founded by the fine time synchronizer, with the _ oa rse
average of correlation energy for the other 15 positions. The
ratio of this two correlation values will become one, ideally
time sync. _ crx"e
T

if there is no peak in the 16 energy value. Coarse time Figure. 3. Transaction of Proposed time synchronization.
synchronizer recognizes this position as the starting-point of
the G12. This ratio and decision rules can be represented
by: Step4. After three short training symbols from starting-
position of fine time synchronizer, coarse time
gVK)::::::
J 1 15
C(k) (9) synchronizer is activated using the temporary Tfine offset
which is updated continuously by the fine time
15 15,i
i=0,7i,Tfin, nffs,t
C(k - i) synchronizer.
and Step5. After the acquisition of Tcoarse, finally, synchronizer
calculates the starting point of G12 using (11).
lcoarse arg min [g(k) < Ycoarse ] ('10)
4. Frequency Synchronization
where Ycoarse denotes a threshold value and k is integer The frequency offset estimation is performed by the
number of multiple of 16 which is larger then Tpd Finally measuring the phase rotation in the Euclidian two
the starting point of G12, TG12, is computed using: dimensional space between two complex vectors which are
separated in specific time interval. There are two kind of
frequency offset estimation in the OFDM-based wireless
TGI 2 =Tcoarse Tfine offset (11) LAN system. One is coarse and the other is fine frequency
Based on the description in subsection 1), 2) and 3), a offset estimation. These are identical operation except of
computational procedure of timing synchronization is time interval and the number of samples to be calculation. In
summarized as follow (see Fig. 2 and Fig. 3). this paper, only the coarse frequency estimation which
operates in the short training symbol period is concerned.
StepO. Calculate the cross correlation C(n) until the coarse The fine frequency offset estimation can be understood in a
time synchronization is completed. same way.
Stepl. Packet detector counts the number of peak position at We assume that the coarse frequency offset estimator uses
which correlation value C(n) is larger then Ypd to Npeak. the last three short preamble symbols, i.e. t8, t9 and tlO.

Step2. If there is no second peak within n NSTS, go to stepO Conventional Method


Step3. After the acquisition of Tpd, fine time synchronizer The conventional method is average of phase of rotation
starts to accumulate the C(n) for each time-offset phase [7][8]. This is given by
from 0 to 15.

AO 1 (12)
TI t g tI
t t t Ls
C(n)A where Ls is the length of short preamble symbol, 16. From
this operation, system knows the frequency offset between
transmitter and receiver in forms of amount of phase rotation
Tpd
for the 16 samples, 0.8[ts. Since the each sample of short
preamble symbol, however, has its own amplitude, it is unfair
T

Figure. 2. Process sequences of the proposed the time that all samples have the same weight for averaging of
synchronization. arctangent evaluation for frequency offset.

ISBN 978-89-5519-131-8 93560 - 1671 - Feb. 12-14, 2007 ICACT2007


Authorized licensed use limited to: UNIVERSITY OF WESTERN ONTARIO. Downloaded on March 08,2010 at 13:12:48 EST from IEEE Xplore. Restrictions apply.
Proposed Method
As stated above introduction subsection, the smaller
amplitude of sample is, the more sensitive, and unreliable it is,
for evaluating its phase in the same channel condition. The
proposed scheme, therefore, has the consideration of these
amplitudes as weight factor. This operation can be expressed
by the following (13).

AO = arctan r2LO
(13)
s ERe{ Re {rn r+Ls }

where Im {r rL } Re{+L } are weighting


and factor for
Figure. 4. Failure rate of the timing synchronization.
imaginary and real component, respectively.

5. Simulation Results
In this section, the proposed synchronization method is
simulated and compared with the conventional method in
terns of the synchronization failure rate. The proposed
method is simulated under multi-path Rayleigh fading
channel with Doppler shift, 150Hz and frequency offset
(Af If, = 0.3). The thresholds used for simulation are fixed
as licoarse =0.4, pd 00 9coarse = 3 The parameters of Npeak
for step 1 and n for step2 are three and five, respectively.
Simulation results exhibit the effectiveness of the
proposed method in AWGN and fading channels. Fig. 4
shows probability of timing synchronization. The Figure. 5. MSE of the frequency estimation.
synchronization failure includes either false detection or miss
detection. The curve shows that propose method is better
than conventional one by about 2dB at the low SNR condition. REFERENCES
Although the conventional method produced an irreducible
probability of synchronization failure at the high SNR, the [1] IEEE, "IEEE802.1la-1999 part 11:Wireless LAN Medium Access
Control (MAC) and Physical Layer(PHY) specification," IEEE 1999.
failure probability of the proposed method decreased [2] T.M. Schmidl and D.C.Cox, "Robust frequency and timing
depending on the SNR value, continuously. synchronization for OFDM," IEEE Trans. Commun., vol. 45, pp.
Fig. 5 presents the result of frequency offset estimation 1613-1621, 1997.
performances of conventional and proposed method [3] M. Wu, W. Zhu, "A preamble-aided symbol and frequency
synchronization scheme for OFDM systems," IEEE International
according to the SNR. Normalized MSE means the mean of Symposium on Circuits and Systems, Vol. 3, pp2627 - 2630, May
squared phase error normalized by 221. We can see that the 2005.
proposed method is quite better than conventional one. The [4] J. Xu, B Guan, "An efficient timing synchronization scheme for
slope of our estimator is much smaller than the other. OFDM systems in IEEE 802.16d," Asia-Pacific Conference
Proceedings, vol. 1 Dec., 2005.
[5] E Zhou, X Zhang, H Zhao, W Wang, "Synchronization algorithms for
6. Conclusions MIMO OFDM systems," IEEE Wireless Communications and
In this paper a low complex and accurate adaptive Networking Conference, vol. 1, ppl8-22 Mar. 2005.
symbol timing synchronization and robust frequency offset [6] R.V. Nee and R. Prasad, OFDM for Wireless Multimedia
Communications: Artech House, 2000.
estimation for the IEEE 802.1 la WLAN are proposed. The
[7] E. Sourour, H. El-Ghoroury, D. McNesil, "Freqeucny offset
proposed timing synchronization method use only cross- Esitmation and Correction in the IEE 802.1 la WLAN," IEEE VTC,
correlation where the threshold is fixed for practical vol. 3, ppl509-1513 Sept. 2005.
implementation and frequency offset estimation is more [8] B. Ai, J. Ge, Y. Wang, S. Yang, P. Liu, G. Liu, "Frequency Offset
accurate since the sample's amplitude is considered. Esitmation for OFDM in Wireless Communications", IEEE Trans.
Simulation results show that the proposed methods can Consumer electornics. Vol. 50, no. 1 Feb. 2004.
provide robust perfornance under not only low but also high
SNR, frequency offset and multi-path fading channels.

ISBN 978-89-5519-131-8 93560 - 1672 - Feb. 12-14, 2007 ICACT2007


Authorized licensed use limited to: UNIVERSITY OF WESTERN ONTARIO. Downloaded on March 08,2010 at 13:12:48 EST from IEEE Xplore. Restrictions apply.

You might also like