Professional Documents
Culture Documents
-point
channel estimate. The question that arises is how to choose
and optimally, so that the minimum channel estimation
Mean Square Error (MSE) is attained with a given preamble
length.
0090-6778/10$25.00 c 2010 IEEE
Authorized licensed use limited to: Jeppiaar Engineering College. Downloaded on May 01,2010 at 04:03:03 UTC from IEEE Xplore. Restrictions apply.
400 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 58, NO. 2, FEBRUARY 2010
Our approach does not seek nor claim optimality in a global
sense. We rather seek to nd optimum performance among
the family of training preambles that consist of repeatedly
transmitting a unique training symbol, as it is common in
current commercial systems and likely in future ones.
This letter is organized as follows. Section II describes
our signal model and the estimator used on the probed
frequencies. Section III presents the interpolation procedure
and derivation of the estimation MSE. The result is then used
in Section IV for deriving optimal (, ) pairs for a preamble
of given length. Examples based on the IEEE 802.11a OFDM
standard [20] are given for illustration. Finally, Section V
summarizes the main conclusions.
II. OFDM SIGNAL MODEL AND CHANNEL ESTIMATION
Consider sounding a frequency selective channel of band-
width
() +(), 0, (1)
where () is the channel impulse response,
indicates linear
discrete convolution, and () represents AWGN of power
spectral density
0
.
The maximum excess delay of the channel (
samples)
may be larger or smaller than the period of the preamble
(). In either case, () is -periodic for
1.
contiguous periods of samples are extracted from the
periodic portion of (). Perfect knowledge of symbol timing
and frequency offset are assumed at the receiver. The
resulting sequences are dened as (cf. Figure 1b)
() = ( + [ 1]), (2)
with = 1, . . . , and = 0, . . . , 1. Because each cycle
of () is a cyclic prex to the following one, it is simple to
show that
() = ()
() +
(), (3)
where
() = (+[
1]), with = 1, . . . , and = 0, . . . , 1. Taking an
-point forward FFT on
() yields
() =
()
() +
(), (4)
where
() =
1
()
=1
() (5)
is the maximum likelihood channel estimator for subcarrier
, when no information of other subcarriers is known [21].
It is to be noted that when = 1 and equals the number
of subcarriers of the OFDM payload transmission (
), then
(5) is van de Beeks LS estimator [1].
III. INTERPOLATION AND MSE DERIVATION
Given the channel estimates
() at equi-spaced
frequencies, we seek to obtain
channel estimates
(),
= 0, . . . ,
zeros
and, nally, taking a forward
() =
1
=0
(
1
=0
()
+2/
)
2/
.
(6)
Using (4) in (5) and the result in (6), we nd
() =
1
=0
(
1
=0
()
+2/
)
2/
(7)
+
1
=0
(
1
=0
[
=1
()
]
+2/
)
2/
,
with
() =
1
()
(). (8)
Considering an ensemble of frequency selective channels
with Rayleigh fading exponentially decaying power proles
and RMS delay spread
RMS
(cf. Appendix A), we show
in Appendix B that the expected MSE of the estimated and
interpolated channel with respect to the exact channel is
MSE
{
}
=
{
2
(
RMS
RMS
)
+
0
, <
.
(9)
The MSE of (9) is shown with solid lines in Figure 2 as
a function of the preamble sequence length for a
Signal-to-Noise Ratio (SNR) of 7 dB. For each curve,
is xed and increases. The markers on top of the curves
are simulation results. Each point collects statistics from 3000
channel realizations whose expected
RMS
, normalized to the
sampling rate, is one (E{
RMS
} = 1). This represents, for
instance, indoor ofce environments for OFDM transmissions
congured with IEEE 802.11a parameters [20].
The channel impulse response sequences used in the curves
of Figure 2 are
/
RMS
/
RMS
)
depends solely on
and on the channel parameters
RMS
and
, and not on
the number of observations . As increases, the MSE
decreases until reaches the value of
= 4
RMS
2
2
RMS
2
RMS
, (10)
which does not depend on . Thus, given a desired channel
estimation MSE, (10) can be solved for and the result used
in (9) for nding . The lower envelope and optimal (, )
pairs obtained this way are plotted in Figure 3 for various SNR
(environment as in Figure 2 with expected
RMS
equal one).
Figure 3 also shows the intervals of MSE that lead to different
choices of , as well as the MSE attained when estimating
the channel using the two long symbols of the IEEE 802.11a
preamble (square markers, equal
0
2
, from (9); to be fair, the
standard does not mandate to estimate the channel using only
the two long symbols.) In practice, it is clear that and
cannot be adjusted as a function of SNR. The pair = 8 and
= 4 attains or surpasses the estimation MSE obtained with
the two long symbols for all SNRs considered, but uses a
preamble four times shorter.
V. CONCLUSIONS
The problem of attaining initial channel estimates in OFDM
with minimum transmission overhead was considered. The
Authorized licensed use limited to: Jeppiaar Engineering College. Downloaded on May 01,2010 at 04:03:03 UTC from IEEE Xplore. Restrictions apply.
402 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 58, NO. 2, FEBRUARY 2010
relationship between the estimation mean square error and
the preamble length was established for an estimator based
on observing the channel spectrum a given number of times
at a given number of equi-spaced frequencies, followed by
interpolation. The result was used for nding the shortest
training sequence that attains a given channel estimation mean
square error. An example based on IEEE 802.11a parameters
illustrates that the optimal preamble is at least four times
shorter than the one used by 802.11a.
APPENDIX A
CHANNEL MODEL
The channel model used in our mathematical analysis and
simulations is the exponentially decaying Rayleigh fading
channel model. Its discrete impulse response sequence is
() =
1
2
(1
2
)
=
16
RMS
. This corresponds to eight time-constants of the
decaying exponential of (11), and ensures that no multipath
components with signicant power are neglected.
APPENDIX B
DERIVATION OF THE MSE (9)
In order to simplify the notation in (7), we dene
() =
1
=0
()
2
(12)
() =
1
=0
[
=1
()
]
, (13)
and denote the exact channel by
()
}
= E
{
()
()
2
}
=
1
1
=0
1
2
=0
E{(
1
)
(
2
)}
2
(
1
2
)
(14)
Re
{
1
=0
E
{
()
()
}
}
(15)
+E
{
()
()
}
(16)
+
1
1
=0
1
2
=0
E{(
1
)
(
2
)}
2
(
1
2
)
. (17)
For deriving the expectations in (14) to (17), we generically
denote the -point frequency response of the channel impulse
response () by
(). Then
E{
1
(
1
)
2
(
2
)} = (18)
1
1
=0
2
=0
E{(
1
)
(
2
)}
2
1
2
2
,
where E{(
1
)
(
2
)} can be obtained using (11). Thus,
E
{
1
(
1
)
2
(
2
)
}
= (19)
1
=0
(
1
2
)
1
2
2
.
Using (19), the expected values in (14) to (17) are given by
E{(
1
)
(
2
)} =
1
2
1
1
=0
1
2
=0
=0
(1
2
)
2
(
1
2
)
, (20)
E
{
()
()
}
=
1
(21)
=0
=0
(
1
2
)
,
E
{
()
()
}
=
1
=0
(
1
2
)
2
, (22)
and
E{(
1
)
(
2
)} = (23)
1
1
=0
1
2
=0
(
1
=1
2
=1
E
{
1
(
1
)
2
(
2
)
}
)
+2
.
Using (8) and recalling that ()
2
= 1, we nd
E
{
1
(
1
)
2
(
2
)
}
=
{
0
2
,
1
=
2
,
1
=
2
0, otherwise
,
(24)
where
0
is the variance of the noise. Then, (23) becomes
E{(
1
)
(
2
)} =
0
2
1
=0
+2
(
1
2
)
. (25)
By replacing (20), (21), (22) and (25) into (14) to (17),
respectively, and averaging over all subcarrier frequencies
we obtain
Authorized licensed use limited to: Jeppiaar Engineering College. Downloaded on May 01,2010 at 04:03:03 UTC from IEEE Xplore. Restrictions apply.
OBERLI et al.: ON USING TRANSMISSION OVERHEAD EFFICIENTLY FOR CHANNEL ESTIMATION IN OFDM 403
MSE
{
}
=
1
=0
1
1
=0
1
2
=0
[
1
1
=0
1
2
=0
(26)
(
=0
(
1
2
)
2
2(
1
2
)/
)
+2(
1
2
)/
]
2(
1
2
)/
Re
{
=0
1
=0
[
1
=0
(27)
(
=0
(
1
2
)
2
2/
+2/
+2/
]
2/
}
+
1
=0
=0
(
1
2
)
2
(28)
+
0
=0
1
1
=0
1
2
=0
(29)
(
1
=0
+2(
1
2
)/
)
2(
1
2
)/
.
The MSE (26) to (29) can be simplied signicantly. We
do this in the sequel and thus obtain the components of (9).
Simplication of (26): If
1
=
2
, then the only part that
depends on is
2(
1
2
)/
1
=0
2(
1
2
)/
evaluates to zero.
If
1
=
2
= and
1
=
2
, then the
1
=0
only
operates on
2(
1
2
)/
. This is also a geometric series
that evaluates to zero.
If
1
=
2
and
1
=
2
, then (26) is a geometric series
in and can be reduced to 1
2
.
Simplication of (27): The second line in (27) (round brack-
ets), is a geometric series. After solving it, the expression in
square brackets can be broken up into the following geometric
series of
terms:
(
1
2
)
1
=0
(
1 +
(22
+. . . (30)
+
(22
)(
1)
2
(
1)
.
The
1
=0
is then applied to each summand in (30), and
the corresponding geometric series is evaluated. The result,
denoted (30)-bis, is the argument of the
1
=0
(rst sum
outside square brackets in (27)). Using (30)-bis, it is simple
to show that for each value of <
). This
allows to re-write the sum on as
min{,
}1
=0
(30)
2
= (31)
(
1
2
)
min{,
}1
=0
,
which is a straighforward geometric series that yields
(
1
2 min{,
}
)
. Finally, the
1
=0
evaluates to
(
1
2 min{,
}
)
.
Simplication of (28): This is a straightforward geometric
series.
Simplication of (29): If
1
=
2
, then the sum on in
round brackets is a geometric series that evaluates to zero.
When
1
=
2
the sums are trivial and yield
0
.
APPENDIX C
DERIVATION OF (10)
Eq. (10) is obtained by a system of 3 equations. The rst
equation restricts two preambles to have the same MSE. This
is imposed by evaluating (9) once with pair (
1
,
1
), another
time with pair (
2
,
2
), and then by setting both expressions
equal. The second equation restricts both preambles to have
the same length,
1
1
=
2
2
, and the third equation
forces
2
= 2
1
. Solving for
1
and replacing the result in
(9) yields (10).
ACKNOWLEDGMENTS
This work was supported by grants FONDECYT 1060718
and ADI-32 2006 from CONICYT Chile.
REFERENCES
[1] J.-J. van de Beek, O. Edfors, M. Sandell, S. K. Wilson, and P. O.
Brjesson, On channel estimation in OFDM systems," in Proc. IEEE
VTC95, vol. 2, Sept. 1995, pp. 815-819.
[2] Y. G. Li, N. Seshadri, and S. Ariyavisitakul, Channel estimation for
OFDM systems with transmitter diversity in mobile wireless channels,"
IEEE J. Sel. Areas Commun., vol. 17, no. 3, pp. 461-471, Mar. 1999.
[3] Y. G. Li, Pilot-symbol-aided channel estimation for OFDM in wireless
systems," IEEE Trans. Veh. Technol., vol. 49, no. 4, pp. 1207-1215, July
2000.
[4] B. Yang, K. B. Letaief, R. S. Cheng, and Z. Cao, Channel estimation for
OFDM transmission in multipath fading channels based on parametric
channel modeling," IEEE Trans. Commun., vol. 49, no. 3, pp. 467-479,
Mar. 2001.
[5] O. Edfors, M. Sandell, J.-J. van de Beek, S. K. Wilson, and P. O. Br-
jesson, OFDM channel estimation by singular value decomposition,"
IEEE Trans. Commun., vol. 46, no. 7, pp. 931-939, July 1998.
[6] , Analysis of DFT-based channel estimation for OFDM," Wireless
Personal Commun., vol. 12, no. 1, pp. 55-70, Jan. 2000.
[7] Y. G. Li, Simplied channel estimation for OFDM systems with
multiple transmit antennas," IEEE Trans. Wireless Commun., vol. 1,
no. 1, pp. 67-75, Jan. 2002.
[8] H. Minn, D. I. Kim, and V. K. Bhargava, A reduced complexity channel
estimation for OFDM systems with transmit diversity in mobile wireless
channels," IEEE Trans. Commun., vol. 50, no. 5, pp. 799-807, May
2002.
[9] H. Tang, K. Y. Lau, and R. W. Brodersen, Interpolation-based max-
imum likelihood channel estimation using OFDM pilot symbols," in
Proc. IEEE GLOBECOM02, vol. 2, Nov. 2002, pp. 1860-1864.
[10] D. Shen, Z. Diao, K.-K. Wong, and V. O. K. Li, Analysis of pilot-
assisted channel estimators for OFDM systems with transmit diversity,"
IEEE Trans. Broadcast., vol. 52, no. 2, pp. 193-202, June 2006.
[11] I. Barhumi, G. Leus, and M. Moonen, Optimal training design for
MIMO OFDM systems in mobile wireless channels," IEEE Trans.
Signal Process., vol. 51, no. 6, pp. 1615-1624, June 2003.
[12] H. Minn and N. Al-Dhahir, Optimal training signals for MIMO OFDM
channel estimation," in Proc. IEEE GLOBECOM04, vol. 1, Nov./Dec.
2004, pp. 219-224.
[13] J.-W. Choi and Y.-H. Lee, Optimum pilot pattern for channel estimation
in OFDM systems," IEEE Trans. Wireless Commun., vol. 4, no. 5, pp.
2083-2088, Sept. 2005.
[14] X. Ma, L. Yang, and G. B. Giannakis, Optimal training for MIMO
frequency-selective fading channels," IEEE Trans. Wireless Commun.,
vol. 4, no. 2, pp. 453-466, Mar. 2005.
Authorized licensed use limited to: Jeppiaar Engineering College. Downloaded on May 01,2010 at 04:03:03 UTC from IEEE Xplore. Restrictions apply.
404 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 58, NO. 2, FEBRUARY 2010
[15] D. Hu, L. Yang, Y. Shi, and L. He, Optimal pilot sequence design for
channel estimation in MIMO OFDM systems," IEEE Commun. Lett.,
vol. 10, no. 1, pp. 1-3, Jan. 2006.
[16] D. Hu, L. He, and L. Yang, Joint pilot tone and channel estimator
design for OFDM systems," IEEE Commun. Lett., vol. 10, no. 10, pp.
695-697, Oct. 2006.
[17] S. Coleri, M. Ergen, A. Puri, and B. Ahmad, Channel estimation
techniques based on pilot arrangement in OFDM systems," IEEE Trans.
Broadcast., vol. 48, no. 3, pp. 223-229, Sept. 2002.
[18] M. Dong and L. Tong, Optimal design and placement of pilot symbols
for channel estimation," IEEE Trans. Signal Process., vol. 50, no. 12,
pp. 3055-3069, Dec. 2002.
[19] B. Hassibi and B. M. Hochwald, How much training is needed in
multiple-antenna wireless links? IEEE Trans. Inf. Theory, vol. 49, no. 4,
pp. 951-963, Apr. 2003.
[20] Part 11: Wireless LAN Medium Access Control (MAC) and
Physical Layer (PHY) specications High-speed Physical Layer
in the 5 GHz Band, IEEE Std. 802.11a-1999. [Online]. Available:
http://standards.ieee.org/getieee802
[21] C. Oberli and B. Daneshrad, Channel estimation for MIMO-OFDM
with training overhead trade-off," in Proc. IEEE PIMRC04, Sept. 2004.
Authorized licensed use limited to: Jeppiaar Engineering College. Downloaded on May 01,2010 at 04:03:03 UTC from IEEE Xplore. Restrictions apply.