Samir Kapoor, Daniel J. Marchok Yih-Fang Huang

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Proceedings of 48th Vehicular Technology Conference (VTC98), Ottawa, Canada, May 1998

PILOT ASSISTED SYNCHRONIZATION FOR WIRELESS OFDMSYSTEMS OVER FAST


TIME VARYINGFADING CHANNELS
Samir Kapoor, Daniel J. Marchok
Tellabs Research Center
Mishawaka, IN 46545
Yih-Fang Huang
University of Notre Dame
Notre Dame, IN 46556
ABSTRACT
This paper presents a frequency synchronization technique
for wireless OFDM communication systems over fast time
varying channels. The overall method, known as loop-timing,
can be used to synchronize multiple mobile transceivers with
a base-station. In this method, the mobiles rst achieve
downlink synchronization and subsequently derive their up-
link timing reference from the recovered downlink clock.
To accomplish the former task, a periodic time-frequency
pattern of pilot tones is embedded in the transmitted down-
link baseband signal. Frequency offsets are estimated and
compensated using an approximate maximumlikelihood fre-
quency offset estimator. The method can be used to correct
for frequency offsets up to half the OFDM symbol rate and
is suitable for real time implementation.
I. INTRODUCTION
Orthogonal Frequency Division Multiplexing (OFDM) has
recently emerged as an attractive alternative to conventional
modulation and multiple access schemes for wireless com-
munications [1]. Potential applications include next genera-
tion wireless local loop, wireless local area networks and
cellular and PCS systems [5]. The development of fast,
accurate, and reliable synchronization algorithms for wire-
less OFDM systems is a critical and challenging problem
with signicant impact on overall system design and per-
formance [2, 5, 6, 8]. This paper presents a pilot-assisted
technique for synchronization of mobile transceivers with a
centralized base-station over fast time varying fading chan-
nels. For instance, this method may be used for synchroniz-
ing multiple mobiles with a single base-station in cellular
orthogonal frequency division multiple access (OFDMA)
systems.
Figure 1 depicts a schematic diagram of the base-station
and mobile OFDM transceivers with emphasis on the syn-
chronization sub-system. The proposed method utilizes pilot-
tones embedded in the baseband OFDM signal which are
used by the mobile receiver to lock-on to the base-stations
timing reference. OFDM tones are also referred to as sub-
symbols, sub-carriers or bins in the literature. As is cus-
tomary, a cyclic-prex (or guard time) is appended to all
symbols to preserve orthogonality between sub-carriers and
simplify equalization [1, 8]. A key distinguishing feature
of a pilot sub-symbol is that it is modulated with a constant
pre-determined sub-symbol at all times. Also, its spectral
location (or bin number) is chosen such that the pilot-tone
waveformhas as close to an integer number of periods in the
duration of the cyclic-prex as possible. This ensures that
the pilot-tone appears as a phase-continuous waveform to
the remote transceiver irrespective of its time domain sym-
bol alignment with the base-station.
A frequency acquisition and tracking procedure com-
prising of all-digital baseband algorithms is developed us-
ing the method of loop timing. The proposed method is
not only robust but also incurs low bandwidth overhead. To
overcome the effect of frequency selective fades across the
entire signal bandwidth, multiple pilot tones are used. The
inter-pilot frequency spacing is chosen commensurate with
the coherence bandwidth of the channel and the allowable
bandwidth overhead. In the loop timing method, each mo-
bile transceiver rst synchronizes itself to the base-station
and then derives its uplink transmitter timing reference from
the recovered downlink clock. Each mobile transceiver has
a local timing reference, usually derived from a Voltage
Controlled Crystal Oscillator (VCXO) which provides the
timing reference for the receiver A/D, transmitter D/A and
all radio frequency (RF) circuitry. Frequency offsets be-
tween the receive and transmit symbol clock occur due to
non-idealities in the remote transceiver VCXOs, possibly
of the order of several parts-per-million (ppm). When time-
invariant channels are encountered, each data modulated tone
in a OFDM signal is exactly orthogonal to all other tones.
However, transmission over fast time varying channels di-
minishes the orthogonality leading to inter-channel inter-
ference (ICI). Other ICI causing factors include phase noise
and jitter.
To achieve downlink frequencysynchronization, the mo-
bile receiver constructs a frequency offset estimator using
the set of embedded pilot-tones in the received baseband
signal. This algorithm also maintains synchronization dur-
Proceedings of 48th Vehicular Technology Conference (VTC98), Ottawa, Canada, May 1998
Base Station
Transceiver
RF to baseband
conversion
RF conversion
Baseband to
Base Station
Timing Reference
RF to baseband
conversion
RF conversion
Baseband to
Loop timed VCXO
Timing Reference
Mobile
Transceiver
D/A
A/D
D/A
A/D
Figure 1: Schematic diagram of synchronization sub-system in a wireless OFDM base-station and mobile transceiver.
ing steady-state operation in tracking mode. Note that the
use of conventional phase locked loops for frequency ac-
quisition and tracking is precluded by the fast time-varying
nature of the channel.
II. FREQUENCY OFFSET ESTIMATION
Consider a discrete time OFDM system model in which
orthogonal sinusoids are transmitted each symbol time.
Thus, the complex baseband transmitted signal at the th
symbol time is given by
(2.1)
where ; is the th frequency domain sub-
symbol during the th symbol time typically selected from a
quadrature amplitude modulation (QAM) constellation; is
the time index and denotes the set of sub-channel indices
being used for data transmission. With a sample rate , as-
sume that each user encounters a time-selective multipath
fading channel impulse response (CIR) spanning a maxi-
mum duration . Also, it is assumed that the
the cyclic prex is chosen to be of length samples such
that , i.e. . For typical wireless chan-
nels, the coherence bandwidth far exceeds the OFDM sub-
channel frequency spacing. Thus, each sub-channel experi-
ences at fading. The received signal after passing through
the channel is given by,
(2.2)
where denotes the CIR at time due to the impulse
. Note that slowly time-varying channels may be
considered to be time-invariant over a symbol period lead-
ing to . Using (2.1) in (2.2), the received
signal is given by
(2.3)
where denotes discrete-time AWGNsamples with vari-
ance . At the receiver, cyclic prex samples are discarded
before demodulation and it is sufcient to consider each
symbol independently. Consider now the demodulation of a
pilot sub-symbol with a frequency offset. Assuming that the
frequency offset between the received and transmitted base-
band signals is Hz, let the normalized frequency offset
be given by
where the inter-channel spacing . It is also
assumed that a total of bins, indexed by , are allocated
to be pilot tone bins. Thus, the th demodulated pilot tone is
given by
(2.4)
for and where denotes a random phase
offset accumulated over the previous ( ) symbols due to
the frequency offset. The time index on is dropped be-
cause the pilot sub-symbols are constant for every symbol.
and denote the channel attenuation and additive
ICI for the th bin during the th symbol respectively as,
Proceedings of 48th Vehicular Technology Conference (VTC98), Ottawa, Canada, May 1998
Each demodulated sub-symbol is nowcorrupted by ICI from
all other sub-symbols. Similarly, for the next symbol,
where for notational convenience,
First consider time-invariant (or slowly time-varying) chan-
nels for which . Thus, in this case,
(2.5)
Using only the th pilot tone, an estimate for is now con-
structed using a block of symbols as,
(2.6)
Similar approaches with differing time-frequency congu-
rations of pilots have also been suggested in [5]-[8]. Using
all the pilots simultaneously,
(2.7)
Other combining methods, such as selection combining, may
also be used. With appropriate assumptions on the additive
noise and ICI, (2.6) is an approximate maximum likelihood
estimate (AMLE) of . The proof follows along the lines
of [4, 7]. The approximation comes from two factors - the
time-varying nature of the CIR and statistical assumptions
on the additive noise and ICI. Considering the th pilot tone,
let
Using (2.5), we have
Denoting the MLE for as
To derive the MLE, conventional assumptions are made on
the post-demodulation noise and ICI. Namely, they are as-
sumed to be instances of stationary, ergodic zero-mean com-
plex valued processes. The mean of is and
can be factored out into a product of partial densities. Thus,
can be obtained by maximizing the log-likelihood
function according to
Differentiating with respect to and setting to zero yields
Observing that the right hand side is merely the complex
conjugate of the left-hand side, the imaginary part may be
set to zero yielding (2.6). It can be also be shown that
the estimator for in (2.6) is unbi-
ased and asymptotically consistent. Consider now the auto-
correlation between the channel attenuation of successive
pilot sub-symbols for a time-varying channel. Assume a
wide-sense stationary uncorrelated scattering (WSSUS) fad-
ing model and a Rayleigh fading Doppler spectrum of the
form [3],
where is the maximum Doppler frequency of the re-
ceived signal. Thus,
where denotes the zeroth order Bessel function;
and for . The frequency-offset estimate
at each symbol time is suitably scaled to provide control
signals to the local VCXO. In order to uniquely ascertain the
signed frequency offset (in Hz) from the normalized offset
, it is necessary that
For example, with a symbol rate of 8 KHz, baseband fre-
quency offsets up to 4 KHz can be unambiguously esti-
mated and corrected. For systems with large-scale initial
frequency offsets, search procedures are typically deployed
prior to acquisition.
Instead of block updates, the AMLE can also be readily
updated recursively to provide a control signal to the VCXO
each symbol time for steady state frequency tracking. The
argument of the inverse tangent function in (2.6) is updated
each symbol time over a sliding window of previous sym-
bols while is computed each symbol time using (2.6) and
(2.7).
Proceedings of 48th Vehicular Technology Conference (VTC98), Ottawa, Canada, May 1998
0 20 40 60 80 100 120 140 160 180 200
0.95
0.955
0.96
0.965
0.97
0.975
0.98
0.985
0.99
0.995
1
Doppler frequency (Hz)
N = 128
N = 256
N = 512
Figure 2: Auto-correlation of channel fading coefcient be-
tween successive OFDM symbols.
III. SIMULATION RESULTS
Figure 2 plots the auto-correlation between the fading co-
efcients for successive OFDM pilot sub-symbols ( ) and
for = 128, 256 and 512 and symbol rates of =
16 KHz, 8 KHz and 4 KHz respectively. This results in the
same data rate and occupied bandwidth for all three cases.
For example, with QPSK sub-symbols, this yields a maxi-
mum aggregate data rate of 4.096 Mb/s at the base-station.
The cyclic prex is taken to be 10 long in each case. The
performance of the AMLE algorithm is depicted in Figure
3. The mean error magnitude and the frequency offset esti-
mation error standard deviation is plotted versus for SNR
= 20 dB; = 10; Doppler spreads of 0, 50 and 100 Hz; =
256 and = 8 KHz. It is seen that frequency offsets can
be reduced to a very small fraction of the symbol clock.
IV. CONCLUSIONS
This paper has described a frequencysynchronization method
for wireless OFDMcommunication systems suitable for op-
eration over fast time-varying channels. The method may
be used to carry out downlink synchronization of mobile
transceivers with a base-station. For this purpose, a time-
frequency pattern of pilot tones is embedded in the down-
link signal. An approximate maximumlikelihood algorithm
is used to estimate and compensate for frequency offsets.
The overall method, called loop timing, is suitable for real-
time implementation for operation over fast time varying
and frequency selective fading channels.
20 30 40 50 60 70 80 90 100
0
1
2
3
4
5
M
e
a
n

e
r
r
o
r

m
a
g
n
i
t
u
d
e

(
H
z
)
fm = 100 Hz
50 Hz
0 Hz
20 30 40 50 60 70 80 90 100
0
1
2
3
4
5
6
L (symbols)
E
r
r
o
r

s
t
a
n
d
a
r
d

d
e
v
i
a
t
i
o
n

(
H
z
)
0 Hz
50 Hz
100 Hz
Figure 3: Frequency offset estimate using the AMLE algo-
rithm.
V. REFERENCES
[1] J. A. C. Bingham, Multicarrier Modulation: An idea
whose time has come, IEEE Comm. Mag., pp. 5-14,
May 1990.
[2] J.K. Cavers, An Analysis of Pilot Symbol Assisted
Modulation for Rayleigh Fading Channels, IEEE
Trans. on Veh. Tech., Vol. 40, No. 4, pp. 686-693, Nov.
1991.
[3] W. C. Jakes, Microwave Mobile Communication,
IEEE Press, 1994.
[4] F. Classen, H. Meyr and P. Sehier, Maximum likeli-
hood open loop carrier synchronizer for digital radio,
Proc. ICC93, pp. 493-497, 1993.
[5] F. Classen and H. Meyr, Frequency synchronization
algorithms for OFDM systems suitable for commu-
nication over frequency selective fading channels,
Proc. VTC94, pp. 1655-1659, Stockholm, June 1994.
[6] M. Luise and R. Reggiannini, Carrier frequency ac-
quisition and tracking for OFDM systems, IEEE
Trans. Comm., pp. 1590-1598, November 1996.
[7] P. Moose, A technique for OFDM frequency offset
correction, IEEE Trans. Comm., pp. 2908-2914, Oc-
tober 1994.
[8] H. Sari, G. Karam and I. Jeanclaude, Channel equal-
ization and carrier synchronization in OFDM sys-
tems, in Audio and Video Digital Radio Broadcasting
Systems and Techniques, R. Gaudenzi and M. Luise,
Eds, Amsterdam: Elsevier, 1994.

You might also like