You are on page 1of 4

2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

A Simplified Training Pattern for MIMO-OFDM Channel Estimation Technology

Wang Zhe/Bai Fan Jiang Zhanjun


Institution of Electronic and Information Engineering National Mobil Communications Research Lab
Lanzhou liaotong University South East University
Lanzhou, China Nanjing, China
e-mail: wangzhe@mail.lzjtu.cn e-mail: jzjang@sina.com

Abstract-The tracking of fading wireless channel is needed in overlap of the signals from each transmitting antenna,
MIMO-OFDM system. Training sequences can be used for channel estimators based on OFDM system can not be used
synchronization and channel estimation. However, if too much in the MIMO-OFDM system directly. Then many channel
training blocks are used, the efficiency of system will be estimation methods for MIMO-OFDM systems were
decreased. A simplified training pattern which can reduce the developed in [4]-[10). In [4], the LS estimator for OFDM
waste of resources is proposed in this paper. Computer system with transmitting diversity based on the robust
simulations prove that performance of the system with
estimation presented in [3] was proposed. Because a large
simplified pattern is almost equivalent to the system doesn't
matrix inversion is needed in this method, the simplified
use the simplified training pattern under low SNR, with SNR
method for that estimator was presented in [5] by avoiding
rising, the performance of the system with simplified training
large matrix inversion. And many researches on blind and
block pattern will become better and better, and the simplified
pattern can save four OFDM periods for transmitting training
semi-blind channel estimator was proposed in [6], [7].
blocks. However, the complexity of the blind and semi-blind
estimators is substantially high. In recent years, there are
Keywords-MIMO; OFDM; channel estimation; STBC; some researches on improving estimator performance by
training sequence improving the training pattern. Some design methods of
optimal training sequences proposed in [4], [5], [8]. And the
transmitting pattern of training sequence can also be changed
1. INTRODUCTION
to improve the performance of channel estimator. In [9], a
Future mobile wireless communication systems will method to transmit all the training symbols cyclically at each
require high-bit-rate technologies. Orthogonal frequency antenna can avoid the limitation caused by transmitting same
division multiplexing (OFDM ) is a novel technique that training sequence at same antenna. In [10], the low
divides entire channel into many narrow parallel sub­ complexity channel estimation for MIMO-OFDM system
channels, so that it improves data rate and avoids inter­ based on STBC training sequences was proposed. Although
symbol interference (ISI ) caused by multi-path propagation. the low complexity channel estimator presented in [10] is
Multi-input multi-output (MIMO ) system uses multiple reduced the complexity for the system with two transmitting
antennas at both sides to increase capacity of wireless antennas, when the method used in some systems with more
channel without the need of extra bandwidth. These two transmitting antennas, more OFDM periods will be needed
techniques can be used together to both improve capacity for transmitting training blocks. In this paper, we propose a
and quality of wireless communication systems. simplified transmitting pattern of training blocks for system
In MIMO-OFDM systems, channel state information with four transmitting antennas, which can save four OFDM
(CSI ) is required for Space-Time decoding. Various channel periods and ensure the estimator performance.
estimation methods for OFDM systems have been reported
in [I ]-[3]. In [I ], least squares (LS ) estimator and minimum II. SYSTEM MODEL
mean squared error (MMSE ) channel estimator only based Consider the fading multi-path channel model, which is
on time-domain channel statistics have been described. The constituted by L paths. And the time-domain response of the
complexity of MMSE estimator is higher than the LS channel is
estimator, but the performance of that is better as the use of
(1 )
channel statistics. In [2], linear minimum mean squared error
(LMMSE ) channel estimator which only based on frequency
where 7, is the delay of the Ith path, a, (t) is the
correlation of channel has been proposed, which use LS
estimation as the rough estimation. And the optimal low-rank corresponding complex amplitude, and due to motion of the
estimator based on singular-value decomposition (SVD ) was vehicle, it is modeled to be wide-sense stationary
also derived in [2] to reduce the complexity of LMMSE narrowband complex Gaussian processes, and it is
estimator. The robust channel estimation for OFDM systems independent for different paths. For allis, a, (t) has the same
based on both time-domain and frequency-domain normalized time correlation function and different average
correlation of channel is proposed in [3]. However, in the powers a}.
multi-antenna systems the signal on the receiver is the

978-1-4244-7237-6/10/$26.00102010 IEEE V5-385


2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

Consider the MIMO-OFOM system shown in Fig. I, where r[ n ]== (r[n,0],·. ·,r[ n,N _I] ) T ; S[n]== (Sl [n], s 2 [n] ) ,
where a data block is encoded by ST encoder and transmitted
into two parallel OFDM blocks of baseband QPSK where si[n]== diag{s i[n, O], . ··, s i[n, N -In; H== [Hl'H2f ,
constellation symbols. A cyclic prefix is used for each where Hi== (Hi[0],·. ·,Hi[N _I] ) T
OFOM branch to preserve orthogonality of the sub-carriers
l1[n]== (1][n,O]"",1][n,N _1])7.
and eliminate the inter-symbol interference between
consecutive OFDM symbols. The number of sub-carriers is
N. The length of the cyclic prefix must be larger than the
III. SIMPLIFIED TRANSMITTING PATTERN BASED ON
delay spread of channel.
STBC TRAINING BLOCK
At data transmitting time n, data block
b[n, k]:k== 0, I,"', N -I is encoded into two signals A. Traditional LS and LMMSE Estimator for OFDM
s i[n, k]:k== O, I"", N-I fori==I, 2. The two signals are system
transmitted by two different antennas simultaneously after For OFDM system, the LS estimator can be expressed as
modulating of OFDM, and the OFDM demodulation of l
HI,S == X[nr Y[n] (5)
received signal at each receiving antenna is the superposition
of the two distorted transmitted signals. The frequency where X==diag{x[n,O],x[n,I]"",x[n,N-In denotes the
response of channel for OFDM system can be expressed as transmitted trammg block,
L-l Y== ( y [n, O] , y [n, I],. ", y [n,N I]
_ T ) denotes the received
H[n, k]== Lh[l]F:; (2)
'=0 signal.
where h[l]== h(r== DJJ, t, ==11 NN , and N is the spacing The LMMSE estimator for OFDM system can be
expressed as
of the sub-carrier. D, is the delay spread of the lth path
sampled by the rate of I / t, , and L is the number of paths.
And FN == exp( -j2Jr / N) . (6)
Hence, the received signal at the jth receiver can be
expressed as where
2 R == E
HH
{ HHH }
r)n, k]== LH,)n,k]s,[n,k]+1])n,k] (3)
i=l
where Hij [n,k] is the channel frequency response between
R ;s
HH
== E {HH:� }
theith transmitting antenna and the jth receiving antenna for
the kth sub-carrier at time n. Assuming that signals where HLS is the LS estimate of channel frequency response,
transmitted through different antennas and received by
different antennas undergo independent fades. 1])n,k] is the and X==diag{x[n,O],x[n,I],. ··,x[n,N-In is the training

additive complex Gaussian noise on thejth receiving antenna block. (j"� is the variance of the additive complex Gaussian
with the zero-mean and the variance is (j"� . For different n's, noise. R
HH
is the auto-covariance matrix of channel
k's andj's 1])n,k] is uncorrelated. frequency response H.
The index j denotes different receiving antennas will be B. Simplified Transmitting Pattern of training blocksfor
omitted from rj [n,k] , Hij [n,k] and 1])n, k] in the following system with four transmitting antennas
for brief. The matrix form of received signal is as When the STBC training pattern applied in systems with
r [n]== S[n]H +11[n] (4) four transmitting antennas, the training block is given by
T; T2 7; T4
-T2 T; -T4 7;
-7; T4 -T; T2
-T4 -7; T2 T;
T== (7)
T,'1 T:2 T*3 T*4
-T'2 T,*1 -T4* T,*3
-7;' T*4 T,1* -T;
-T'4 7;* T2* T;*
Figure I. MIMO-OFDM system with two transmitting antennas Hence, eight OFOM periods will be needed to transmit
training sequences, and assume that the channel is quasi­
fixed over eight OFDM periods. More resources are wasted

V5-386
2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

for transmitting training sequences, and the assumption is not


reasonable. We can transmit training blocks according to the
following pattern shown in fig. 2.
At training time n, transmitting antennas tl and t3 ( 12)
transmit the first group of training sequence T" and �, the
transmitting antennas t2 and t4 do not transmit any data. At IV. SIMULATION RESULTS

the time n+1 , tz and t4 transmit the first group of training In the simulation we consider a system with four
sequence Tz and T4, tl and t3 do not transmit any data. transmitting antennas and four receiving antennas. The
channels between different transmitting antennas and
Then at time n+2 , tl and t3 transmit the second group of
receiving antennas are independent, but have the same
training sequence -T3' and T,,' , tz and t4 do not transmit statistic characteristic. The simulation results are acquired
any data. At the last training time n+3 , t2 and t4 transmit under typical urban (TU ) and hilly terrain (HT ) delay
profiles respectively. The bandwidth for OFOM system is
the second group of training sequence -T4* and T; , tl and 500 kHz, divided into 64 sub-carriers and the total symbol
t3 do not transmit any data. The received signal can be period is 138Jis , where 10 Jis is the cyclic prefix. Four-state

( ) ( )( )
expressed as ST trellis code is used to encode the data block.
r[n] HI lJ[n] Figs. 3 and Fig. 4show the MSE of channel estimator
T3 + (8) with the old pattern and the simplified training pattern
r[n+2] = 1 H3 lJ[n+2]
( r[n+1]
r[n+3]
) ( )(
=TZ4
H2
H4
+
IJ[n+I]
lJ[n+3]
) (9)
proposed here under TU and HT delay profiles respectively.
From the simulation results we can see that the MSE of
estimator with simplified transmitting patter of training
blocks is almost equal to that of estimator with old pattern
where
under low SNR. With the rise of SNR., the performance of
estimator with simplified pattern gets better than the old one.
And note that the estimator with simplified pattern can save
Then we can see that the transmission of training blocks four OFOM periods for transmitting training blocks, and the
is completed only in four OFOM periods, and the estimation complexity of estimator can also be reduced. The efficiency
method used in systems with two transmitting antennas can of the estimator with simplified pattern is higher than the old

( ) )( )
one.

( ( )
be applied as following
C 1 r[n] HI 1 lJ[n]
1 = �3 H = + �3 H (10) V. CONCLUSIONS
� � �n+� � � �n+�
( ) ( )( ) ( )
In this paper OFOM system with multi-antennas is
Cz r[n+1] Hz
..!. lJ[n+1] studied. The estimator for channel parameter based on STBC
=..!. TZ4H = + TZ4H (11) training block is discussed. Through theoretical analysis, the
C4 2 r[n+3] H4 2 lJ[n+3]
STBC estimator plays a good performance in systems with
where Ci=(cJO],ci[]I ,.··,cJN -]I )T for i=I,2,3,4. It is two transmitting antennas. However when it is used in
obviously that C , Cz, C3 and C4 is the LS estimation of systems with four transmitting antennas, the training blocks
1 take up more bandwidth resources, the efficient of system is
HI' Hz, H3 and H4 respectively. decreased. We propose a simplified transmitting pattern of
training blocks for the system with four transmitting
antennas. The simplified pattern saves four OFOM periods to
transmitting training blocks, and improves the efficiency of
the system. Finally, computer simulation results prove that
the system with the simplified pattern has almost equal
performance with the old one under low SNR, and has better
performance under high SNR.

ACKNOWLEDGMENT

This work is supported by the open research fund of


national mobile communications research laboratory of
Figure 2. simplified transmitting pattern of training blocks Southeast University.

From (6) we acquired the LMMSE estimation of Hi for


i=1,2,3,4as following

V5-387
2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

-
� � � � � �: � � � � -�smC:-MMsEcru)- -
------- , - - - ---"1- NEW STBC-MMSE
- - - - - - - - - - (TU)
I
-----�------r-----l----- REFERENCES
-----�------�-----------

[I] 1.-1. Van de Beek, O. Edfors, M. Sandel1, S. K. Wilson and P. O.


Borjesson, "On channel estimation in OFDM systems," Proc. IEEE,
vol. 2, July 1995, pp. 815-819.
[2] O. Edfors, M. Sandell, I.-J. Van de Beek, S. K. Wilson and P. O.
Borjesson, "OFDM channel estimation by singular value
decomposition," IEEE Trans. Commun., vol. 46, July 1996, pp. 923-
927.
_____ � _____ � ______ L _____ � ____ [3] Y. Li, L. J. Cimini Jr and N. R. Sollenberger, "Robust channel
_____ � _____ �______ L _____ � _____ estimation for OFDM systems with rapid dispersive fading channels,"
I I I I
----- T ----- �------ r ----- ' ----- IEEE Trans. Commun., vol. 46, July 1998, pp. 902-915.

1 04 L------- � ------�L------- � ------� L-------" [4] Y. Li and N. Seshadri, "Channel estimation for OFDM systems with
5 1O 1 5 2O 2 5 W transmitter diversity in mobile wireless channels," IEEE J. Select.
SNR in DB Areas Commun., vol. 17, Mar. 1999, pp. 461-471.

Figure 3. MSE of channel estimators under TU delay profile [5] Y. Li, "Simplified channel estimation for OFDM systems with
multiple transmit antennas," IEEE Trans. Wireless Commun. , vol. I,
Jan. 2002, pp. 67-75.
[6] Z Liu, G. B. Giannakis, S. Barbarossa and A. Scaglione, "Transmit­
antennae space-time block coding for generalized OFDM in the
:::::::::: � = :::: = = ::::::::::: -e- STBC-MMSE(HT) presence of unknown multipath," IEEE J. Select. Areas Commun.,
- - - - -
T
- - - - - -I - - - - -"V- NEW STBC-MMSE(HT) vol. 19, July 2001, pp. 1352-1364.
[7] Gaoping Hu and Dong Li, "A low complexity algorithm for channel
estimation of MIMO OFDM system," International conference on
electronic computer technology, IEEE Press, 2009, pp. 113-116,
doi:lO.lI09IICECT.2009.8.
[8] Eugene Golovins and Neco Ventura, "Optimal training for the SM­
MIMO-OFDM systems with MMSE channel estimation,"
Communication Networks and Services Research Conference, IEEE
_____ ! _____ � ______ L ______ Press, 2008, pp. 470-477, doi:10.1109/CNSR.2008.57.
I J I I
----- � ------ r ----- ' --
[9] Wu Jianhua, Li Zhihong and He Lihong, "Channel estimation
-----
T
_ � ___ � ______ L _____ � ___
I algorithms for broadband MIMO-OFDM systems," International
_____ ! _____ �______ L _____ j _____ Conference on Computer Science and Information Technology, IEEE
1 "
Press, 2008, pp. 139-142, doi:10.1109/ICCSIT.2008.115.

4
[10] Gong Y and LetaiefKB, "Low complexity channel estimation for
1 0 5�------� 5------�2�O------�2�
1 O�----�1� 5------�W· space-time coded wideband OFDM systems," IEEE Transactions on
SNR in DB Wireless Communications, Sept. 2003, vol. 2, pp. 876-882,
doi:10.1109ITWC.2003.816797.
Figure 4. MSE of channel estimators under HT delay profile

V5-388

You might also like