Hindawi Publishing CorporationInternational Journal of Antennas and PropagationVolume 2010, Article ID 213576, 10 pagesdoi:10.1155/2010/213576
Research Article
Optimization of Training Signal Transmission for EstimatingMIMO Channel under Antenna Mutual Coupling Conditions
Xia Liu and Marek E. Bialkowski
School of ITEE, The University of Queensland, Brisbane QLD 4072, Australia
Correspondence should be addressed to Xia Liu, [email protected]
Received 2 December 2009; Revised 26 February 2010; Accepted 17 March 2010
Academic Editor: Hon Tat Hui
Copyright © 2010 X. Liu and M. E. Bialkowski. This is an open access article distributed under the Creative Commons AttributionLicense, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properlycited.
This paper reports investigations on the effect of antenna mutual coupling on performance of training-based Multiple-InputMultiple-Output (MIMO) channel estimation. The influence of mutual coupling is assessed for two training-based channelestimation methods, Scaled Least Square (SLS) and Minimum Mean Square Error (MMSE). It is shown that the accuracy ofMIMO channel estimation is governed by the sum of eigenvalues of channel correlation matrix which in turn is influenced bythe mutual coupling in transmitting and receiving array antennas. A water-filling-based procedure is proposed to optimize thetraining signal transmission to minimize the MIMO channel estimation errors.
1. Introduction
In recent years, there has been a growing interest inmultiple-input multiple-output (MIMO) wireless commu-nication systems as they can significantly increase datathroughput (capacity) without the need for extra operationalfrequency bandwidth. In order to explore the advantages ofMIMO technique, precise channel state information (CSI)is required at the receiver. The reason is that without CSI,decoding of the received signal is impossible [1–5]. Inturn, an inaccurate CSI leads to an increased bit error rate(BER) that translates into a degraded capacity of the system[6–8].
Obtaining accurate CSI can be accomplished usingsuitable channel estimation methods. The methods basedon the use of training sequences, known as the training-based channel estimation methods, are the most popular. In[9, 10], several training-based methods including the leastsquare (LS) method, the scaled least square (SLS) methodand the minimum mean square error (MMSE) method havebeen investigated. It has been shown that the accuracy ofthese training-based estimation methods is influenced bythe transmit signal power to noise ratio (SPNR) in thetraining mode, and the number of antenna elements at thetransmitter and receiver. In particular, it has been pointed
out that when the transmitted SPNR and the number ofantenna elements are fixed, the SLS and MMSE methodsoffer better performance than the LS method. This can beexplained by the fact that the SLS and MMSE methods utilizethe channel correlation in the estimator cost function whilethe LS estimator does not take into account the channelproperties.
It is worthwhile to note that the channel propertiesare governed by a signal propagation environment, whichin turn affects the spatial correlation (SC) observed atthe input/output ports of the MIMO system. The spatialcorrelation is dependent on an antenna configuration anda distribution of scattering objects that are present in thepath between transmitter and receiver. In particular it isinfluenced by the finite antenna spacing in array antennas.This finite antenna spacing is also responsible for mutualcoupling which adversely affects signal power transmissionand reception. The mutual coupling effect is especiallypronounced in tightly spaced arrays. Because there is aconsiderable demand for compact size Mobile Station (MS)terminals, the effect of mutual coupling cannot be neglectedand thus has to be taken into account while assessing theMIMO link performance. The problem of mutual couplingin MIMO systems for the case of peer-to-peer communica-tion has been addressed via simulations and measurements
2 International Journal of Antennas and Propagation
in [10–18]. It has been shown that the mutual coupling mayimprove the channel capacity when the spacing of antennaelements in the array is between 0.2λ to 0.4λ (where λ is thecarrier wavelength) [15–18] .
In this paper, the focus is on the effect of mutual couplingon MIMO channel estimation. The problem of mutualcoupling and its effect on the MMSE method of channelestimation have been investigated in [19, 20]. The presentedresults have been limited to simulations providing generaltrends without any further mathematical insight.
In this paper, we present a mathematical analysis explain-ing the mutual coupling effects on the SLS and MMSEmethods for channel estimation. It is shown that for afixed transmit SPNR and for a given number of transceiverantenna elements the accuracy of SLS and MMSE methods isdetermined by the sum of eigenvalues of channel correlationmatrix characterizing the signal propagation conditions.The presence of mutual coupling in array antennas causesvariations of the sum of these eigenvalues, which in turnaffects the accuracy of estimation. In order to reduce thechannel estimation errors, a water-filling based optimizationmethod is proposed to find an optimal training matrixminimizing the channel estimation errors when the MIMOsystem operates under the condition in which the mutualcoupling cannot be neglected.
The rest of the paper is organized as follows. In Section 2,a MIMO system model is introduced. Also, LS, SLS, andMMSE channel estimation methods are described followedby an analysis of channel estimation accuracy. Section 3describes the method for modeling mutual coupling effects.Section 4 presents the optimization of a training signaltransmission when the MIMO system operates under mutualcoupling conditions. Section 5 describes computer simula-tion results. Section 6 concludes the paper.
2. System Description and Training-BasedMIMO Channel Estimation
In this paper, a narrow band block fading MIMO channelis assumed. The number of transmitting and receivingantennas is denoted as Mt and Mr , respectively. The channelis described by theMr×Mt complex matrix H with entries hi jrepresenting the response between the ith receiving antennaand the jth transmitting antenna.
2.1. Training-Based MIMO Channel Estimation. For thetraining-based channel estimation method, the relationshipbetween the received signal and the training sequence is givenby
Y = HP +V , (1)
where P represents the Mt × L complex training matrix andL is the length of the training sequence.
The goal is to estimate the complex channel matrix Hfrom the knowledge of Y and P. The transmitted power
over L time slots in the training mode is constrained by thefollowing expression
‖P‖2F = P, (2)
where P is a given constant representing the total power and‖ · ‖2
F stands for the Frobenius norm.The training matrices are assumed to be orthogonal [9,
10] and the transmitted signal power to noise ratio (SPNR)in the training mode is set to ρ equal to P/σ2
n .Using the LS method, the estimated channel matrix can
be written as [19]
HLS = YP†, (3)
where {·}† stands for the pseudo-inverse operation and P† =PH(PPH)
−1. The mean square error (MSE) of the estimated
channel matrix in the LS method is given as
MSELS = E{∥
∥
∥H − HLS
∥
∥
∥
2
F
}
subject to‖P‖2F = P. (4)
According to [9, 10], the optimal training sequence satisfying(4) is given as
PPH = P
MtI. (5)
The minimum value of MSE for the LS method is
MSELS = M2t Mr
ρ. (6)
From (6) one can see that the optimal performance of the LSestimator is influenced by the number of antenna elements atthe transmitter and the receiver. However, the channel matrixhas no effect on the minimum value of MSE.
The SLS method further reduces the estimation error thatis obtained in the LS method. The improvement is given by ascaling factor γ which uses the channel estimator as given by
HSLS = γ HLS. (7)
The resulting estimation error is given by (8)
MSESLS = E{∥
∥
∥H − γ HLS
∥
∥
∥
2
F
}
subject to ‖P‖2F = P. (8)
Expression (8) can be rewritten as (9)
MSESLS = E{
tr{
(H − γ HLS)H(
H − γ HLS
)
}}
= E{
(1− γ)2 tr{RH} + γ2MSELS
}
= E
{
(MSELS + tr{RH})(
γ − tr{RH}MSELS + tr{RH}
)2
+MSELS tr{RH}
MSELS + tr{RH}}
,
(9)
International Journal of Antennas and Propagation 3
in which RH is the channel correlation matrix defined asRH = {HHH}.
From (9), it is apparent that in order to minimize theMSE of SLS, an optimal value of γ as given by (10) has tobe chosen
γ = tr{RH}MSELS + tr{RH} . (10)
The minimized value of MSE can be written as [9, 10]
MSESLS = E{
MSELS tr{RH}MSELS + tr{RH}
}
< MSELS. (11)
By substituting (6) into (11), the minimum value of MSE isexpressed as,
MSESLS = E
{
11/ tr{RH} + ρ/M2
t Mr
}
. (12)
The estimated channel matrix as given by SLS method can beobtained using (7) and is given as
HSLS = γ HLS
={
tr{RH}MSELS + tr{RH}YpP
†}
.(13)
In practice, RH can be obtained using the channel matrixestimated by the LS method. Therefore
tr{
RH}
= tr{
HHLSHLS
}
. (14)
In the MMSE method, the estimated channel matrix is givenas [21],
HMMSE = Yψ. (15)
The objective is to find the optimal ψ to minimize the MSE.This task can be expressed as
ψopt = arg min∥
∥
∥H − HMMSE
∥
∥
∥
2
F
= argψ
min∥
∥H − Yψ∥∥2F .
(16)
The mean-square error (MSE) for the MMSE method isgiven as,
MSEMMSE = E{∥
∥H − Yψ∥∥2F
}
,
= E{
tr{RH} − tr{
RHPψ}− tr
{
ψHPHRH}
+ tr{
ψH(
PHRHP + σ2nMrI
)
ψ}}
.
(17)
The optimal ψ minimizing MSE satisfies (18),
∂MSEMMSE
∂ψ= 0. (18)
Thus the optimal ψ can be represented by
ψopt = (PHRHP + σ2nMrI)
−1PHRH. (19)
The estimated channel matrix can be derived as,
HMMSE = Yψopt = Y(PHRHP + σ2nMrI)
−1PHRH. (20)
By defining the channel estimation error as
e = H − HMMSE, (21)
the MSE for the MMSE method can be expressed as
MSEMMSE = E{tr{RE}}
= E{
tr{
E{
eeH}}}
= E
{
tr
{
(R−1H +
1σ2nMr
PPH)−1}}
= E{
tr{
(
Λ−1 + σ−2n M−1
r QHPPHQ)−1
}}
.
(22)
In (22), Q is a unitary eigenvector matrix of RH and Λ isa diagonal matrix with eigenvalues of RH , which are giventhrough the eigenvalue decomposition of RH as
RH = HHH = QΛQH. (23)
Here, to construct the channel matrix H, the Kroneckerchannel model is postulated [22, 23]. In this model, thetransmitter and receiver correlation matrices are assumed tobe separable and the channel matrix is represented as:
H = R1/2R GHR
1/2T , (24)
where GH is the matrix including identical independentdistributed (i.i.d) Gaussian entries with a zero mean anda unit variance, and RR and RT are the spatial correlationmatrices at the receiver and transmitter, respectively.
The superiority of SLS and MMSE methods over LSmethod is due to the fact that SLS and MMSE methodsutilize information of the channel correlation RH . Here, itis assumed that the RH is perfectly obtained before channelestimation.
In further considerations, it is assumed that the trans-mitting and receiving array antennas are formed by verticallypolarized wire dipole antennas which are surrounded byscattering objects. Assuming that these scattering objects areuniformly distributed within circles surrounding the arrayantennas, the spatial correlation matrix elements can beobtained using the Clark’s model as given by.
ρR(T)i, j = J0
(
κdi, j)
. (25)
where J0 stands for the zero-order Bessel function, κ is a wavenumber and di j is the distance between elements i and j of
4 International Journal of Antennas and Propagation
the uniform array antenna. The correlation matrices RT andRR can be generated using (25) as
RR(T) =
⎡
⎢
⎢
⎢
⎣
ρR(T)1,1 · · · ρR(T)
1,Mr(t)
.... . .
...
ρR(T)Mr(t), j · · · ρR(T)
Mr(t),Mr(t)
⎤
⎥
⎥
⎥
⎦
. (26)
Having determined RT and RR, the channel matrix can becalculated using (24).
2.2. Estimation Accuracy Analysis. By taking into accountexpression (23), the minimized MSE of the SLS method (12)can be rewritten as
MSESLS = E
⎧
⎨
⎩
[
(tr{RH})−1 +ρ
M2t Mr
]−1⎫
⎬
⎭
= E
⎧
⎨
⎩
[
(tr{Λ})−1 +ρ
M2t Mr
]−1⎫
⎬
⎭
= E
⎧
⎪
⎪
⎨
⎪
⎪
⎩
⎡
⎢
⎣
⎛
⎝
n∑
i
λi
⎞
⎠
−1
+ρ
M2t Mr
⎤
⎥
⎦
−1⎫
⎪
⎪
⎬
⎪
⎪
⎭
,
(27)
where n = min(Mr ,Mt) and λi is the ith eigenvalue of thechannel correlation matrix RH . ρ is the transmit SPNR andequal to P/σ2
n . P is the transmitted training sequence totalpower. If power P and the number of transmit and receiveantennas are fixed then the following relationship holds
MSESLS = E
⎧
⎪
⎪
⎨
⎪
⎪
⎩
⎡
⎢
⎣
⎛
⎝
n∑
i
λi
⎞
⎠
−1
+ρ
M2t Mr
⎤
⎥
⎦
−1⎫
⎪
⎪
⎬
⎪
⎪
⎭
< E
⎧
⎨
⎩
n∑
i
λi
⎫
⎬
⎭
. (28)
It can be seen from (28) that MSESLS is smaller than thesum of the eigenvalues of the channel correlation matrixRH . Therefore, the sum of the eigenvalues of the channelcorrelation matrix RH is the upper bound of MSE in the SLSmethod.
As observed from expression (28), MSE decreases whenthe sum of eignvalues of RH decreases. The same expressionshows that the MSE in the SLS method is influenced bythe number of antenna elements at the transmitter andreceiver. When the number of antenna elements on thetwo sides of communication link drops to one, the systembecomes the conventional SISO system. In this case, thechannel estimation using a fixed-length training sequencebecomes most accurate. This confirms the expectation that itis easier to estimate the SISO channel which is characterizedby a single transfer coefficient between single transmittingand receiving antennas than the MIMO channel which isdescribed by a matrix of transfer coefficients between manyantenna elements.
The derived expression also shows that when the numberof transmit and receive antennas and the total transmittedpower P is fixed, MSE can be minimized by minimizing thesum of eigenvalues of RH .
The expression (22) for the minimized value of MSE canbe rewritten using the orthogonality properties of a trainingsequence P and the unitary matrix Q, as shown by
MSEMMSE
= E{
tr{
(Λ−1 + ρM−1r I)
−1}}
= E
⎧
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎩
tr
⎧
⎪
⎪
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎪
⎪
⎩
⎡
⎢
⎢
⎢
⎢
⎢
⎣
(
λ−11 + ρM−1
r
)−10 · · · 0
0(
λ−12 + ρM−1
r
)−1 . . ....
.
.
.. . .
. . . 00 0 · · · (
λ−1n + ρM−1
r
)−1
⎤
⎥
⎥
⎥
⎥
⎥
⎦
⎫
⎪
⎪
⎪
⎪
⎪
⎬
⎪
⎪
⎪
⎪
⎪
⎭
⎫
⎪
⎪
⎪
⎪
⎪
⎬
⎪
⎪
⎪
⎪
⎪
⎭
= E
⎧
⎨
⎩
n∑
i
(
λ−1i + ρM−1
r
)−1
⎫
⎬
⎭
.
(29)
Similar derivations for MSE apply for the MMSE method.By using the steps analogous to those used for the SLSmethod, the MSE upper bound for the MMSE method canbe expressed as
MSEMMSE = E
⎧
⎨
⎩
n∑
i
(
λ−1i + ρM−1
r
)−1<
n∑
i
λi
⎫
⎬
⎭
. (30)
The expression (30) shows that similarly as in the SLSmethod, a smaller sum of eigenvalues of the channelcorrelation matrix RH leads to a smaller estimation error inthe MMSE method. Similarly as in the SLS method, the MSEof MMSE is influenced by the number of transmitting andreceiving antennas.
Through the above mathematical analysis, one can seethat if the total transmitted power P as well as the numberof antenna elements at the transmitter and the receiver isfixed, the accuracy of the training-based MIMO channelestimation is governed by the sum of eigenvalues of thechannel correlation matrix RH . The smaller is the sum, themore accurate channel estimation is achieved.
Decreasing the sum of eigenvalues of channel correlationis equivalent to decreasing the effective degree of freedom(EDOF) of the MIMO channel [24]. This is caused by anincreased spatial correlation [25]. Hence, it is apparent thatan increased spatial correlation helps improving the channelestimation accuracy.
3. Mutual Coupling Effects
Mutual coupling in an array of collinear side-by-side wiredipoles can be modeled using the approach described in [26].Assuming the array is formed byMrt wire dipoles, the mutualmatrix can be calculated using the following relationshipwith the impedance matrix
C = (ZA + ZT)(
Z + ZTIMrt
)−1, (31)
where ZA is the element impedance in isolation, for example,when the wire dipole is λ/2, its value is ZA = 73 + j42.5[Ω];ZT is impedance of the receiver at each element chosen as
International Journal of Antennas and Propagation 5
the complex conjugate of ZA to obtain the impedance match.The mutual impedance matrix Z is given by
Z =
⎡
⎢
⎢
⎢
⎢
⎣
ZA + ZT Z12 · · · Z1Mrt
Z21 ZA + ZT · · · Z2Mrt
......
. . ....
ZMrt1 ZMrt2 · · · ZA + ZT
⎤
⎥
⎥
⎥
⎥
⎦
. (32)
Note that this expression provides the circuit representationfor the mutual coupling in array antennas. It is valid forantennas operating in a single mode. Wire dipoles fall intothis category.
For a side-by-side array configuration of dipoles havinglength l equal to 0.5λ, the expressions for {Zmn} can beadapted from [18, 26] and are rewritten here as
Zmn
=
⎧
⎪
⎪
⎨
⎪
⎪
⎩
30[0.5772+ln(2κl)−Ci(2κl)]+ j[30Si(2κl)], m = n,
30[2Ci(u0)− Ci(u1)− Ci(u2)]
− j[30(2Si(u0)− Si(u1)− Si(u2))], m /=n,(33)
where κ is the wave number equal to 2π/λ,
u0 = κdh,
u1 = κ(√
dh2 + l2 + l
)
,
u2 = κ(√
dh2 + l2 − l
)
,
(34)
dh is the horizontal distance between the two dipole antennaelements. Ci(u) and Si(u) are the cosine and sine integrals,respectively. They are given as,
Ci(u) =∫ u
∞
(
cos(x)x
)
dx,
Si(u) =∫∞
0
(
sin(x)x
)
dx.
(35)
Expressions (33) provide quite accurate values for {Zmn} forthe dipole spacing of not less than 0.15λ. If one wishes tohave a more accurate model for the mutual coupling effect,the theory and expressions given in [27] are recommended.
Under the presence of mutual coupling, the channelsmatrix H appearing in expressions (23) and (24) has to bereplaced by the new channel matrix H′ = CRHCT whichtakes into account the mutual coupling. By taking intoaccount the mutual coupling, the expression for the channelmatrix (24) is modified to
Hmc = CRR1/2R GHR
1/2T CT. (36)
The new transmit and receiving correlation matrices can beobtained by introducing,
RRmu = CRR1/2R ,
RTmu = R1/2T CT.
(37)
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 116
17
18
19
20
Sum
ofch
ann
elco
rrel
atio
nei
genv
alu
e
dt/λ
(a) Sum of Channel Correlation Eigenvalues versus Transmitter AntennaSpacing
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 116
17
18
19
20
Sum
ofch
ann
elco
rrel
atio
nei
genv
alu
eTaking MC into accountWithout taking MC into account
dr/λ
(b) Sum of Channel Correlation Eigenvalues versus Receiver AntennaSpacing
Figure 1: Mutual coupling effects on the sum of channel correlationmatrix eigenvalues. (a) Sum of eigenvalues versus transmitterantenna element spacing; (b) Sum of eigenvalues versus receiverantenna element spacing.
By using (23) and (24) they can be rewritten as (39) and (38),respectively
Hmc = RRmuGHRTmc, (38)
RmcH = HH
mcHmc = QmcΛmcQHmc. (39)
Under the mutual coupling conditions, (22) is changedto
MSEmcMMSE = E
{
tr{
(
Λ−1mc + σ−2
n M−1r QH
mcPPHQmc
)−1}}
.
(40)
The inclusion of mutual coupling affects the matrix Λ.Therefore, the sum of the eigenvalues of channel correlationmatrix is changed accordingly. Figure 1 shows the simulationresults illustrating the effect of mutual coupling on the sumof the eigenvalues. For this simulation, a uniform linear arraywith 4 antenna elements is assumed at both transmitter andreceiver sides. The elements are wire dipoles having lengthof 0.5λ. The following normalization of channel matrix H isapplied ‖H‖2
F =MrMt, where ‖ · ‖2F is the Frobenius norm.
In Figure 1(a), the antenna element spacing at thereceiver is fixed to 1λ while the spacing at the transmittervaries from 0.1λ to 1λ. In Figure 1(b), the spacing at thetransmitter is fixed to 1.λ while the spacing at the receivervaries from 0.1λ to 1λ. From both Figures 1(a) and 1(b), one
6 International Journal of Antennas and Propagation
can see that when the antenna elements spacing is between0.1λ to 0.5λ, there is an apparent difference between the sumof channel correlation eigenvalues with and without takingmutual coupling effects into account. The ones with mutualcoupling effects are higher than the ones without. When thespacing is increased and exceeds 0.5λ, the difference becomessmaller and the blue and red curves overlap. As a result, at theantenna spacing between 0.1λ and 0.5λ, the mutual couplingincreases the sum of eigenvalues of the channel correlationmatrix. This is because within this spacing range the mutualcoupling decreases the spatial correlation level. Therefore thechannel estimation accuracy is adversely affected.
4. MMSE Method Performance OptimizationUnder Mutual Coupling Conditions
In order to optimize the performance of MMSE training-based estimation method, we have to find the optimaltraining matrix. This optimization problem can be expressedas,
MSEmcMMSE = min
‖P‖2F=P
tr{
(Λ−1mc + σ−2
n M−1r QH
mcPPHQmc)
−1}
,
(41)
Popt = argp
min tr{
(
Λ−1mc + σ−2
n M−1r QH
mcPPHQmc
)−1}
,
subject to ‖P‖2F = P.
(42)
By denoting the combined training matrix as
P = σ−1n M−1/2
r QHmcP, (43)
expression (41) can be rewritten as
MSEmcMMSE = min
‖P‖2F=P
tr{
(
Λ−1mc + P P
H)−1
}
. (44)
Based on properties of the unitary eigenvector matrix Qmc,
P PH
has a diagonal structure as given by
P PH =
⎡
⎢
⎢
⎢
⎢
⎣
p21
p22
. . .p2Mt
⎤
⎥
⎥
⎥
⎥
⎦
, (45)
where p2i is the power of each combined training symbol.
When no optimization is applied, the transmit power isequally distributed into the training symbols as given by
p2i = σ−2
n M−1r M−1
t P. (46)
Next (44) can be rewritten as,
MSEmcMMSE = min
‖P‖2F=P
Mt∑
i=1
(λ−1i,mc + p2
i )−1
assume Mt =Mr ,
(47)
in which λi,mc is the ith eigenvalue in Λ.
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.02
0.025
0.03
0.035
0.04
0.045
0.05
MSE
dt/λ
(a) MMSE MSE versus Transmitter Antenna Spacing
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.767
0.768
0.769
0.77
0.771
0.772
0.773
0.774
MSE
MSE taking MC into accountMSE without taking MC into account
dt/λ
(b) SLS MSE versus Transmitter Antenna Spacing
Figure 2: MMSE and SLS MSE performance versus transmitterantenna element spacing.
To minimize MSE in (44), a water-filling based algorithmcan be applied to optimize the power allocation to trainingsymbols. The symbols in the optimal training matrix aregiven as [9, 10]
pi =√
(μ− λ−1i,mc)
+, (48)
in which (x)+ = max(x, 0) and μ needs to be found to satisfy‖P‖2
F = P.Finally, the optimal training matrix is given as
Popt = σ−2n M−1
r Qmc([μI −Λ−1mc]
+)
1/2. (49)
5. Simulation Results
This section reports the computer simulations aiming at theassessment of effects of mutual coupling on the training-based MIMO channel estimation methods. In the under-taken investigation, the transmitter and receiver are assumedto be equipped with 4-element uniform array antennas. Theelements are wire dipoles having length of 0.5λ. The antennaelement spacing at the receiver is fixed to 1λwhile the spacingat the transmitter varies from 0.1λ to 1λ. Equations (38)and (24) are used to construct the MIMO channel with andwithout taking mutual coupling effects into account. Thetransmit SPNR is fixed at 25 dB.
Figure 2 presents the MSE performance of MMSE andSLS methods versus transmitter antenna elements spacing. Inthis figure, the blue and red curves indicate with and withouttaking into account mutual coupling cases, respectively. For
International Journal of Antennas and Propagation 7
0 0 . 2 0 . 4 0 . 6 0 . 8 1
0 0 . 5
1 0 . 0 2
0 . 0 3
0 . 0 4
0 . 0 5
M S E
0 . 0 3
0 . 0 3 5
0 . 0 4
0 . 0 4 5
0 . 0 5
M M S E M S E v e r s u s a n t e n n a e l e m e n t s p a c i n g t a k i n g M C i n t o a c c o u n t
dt/λdr/λ
(a) 3D view
0 . 0 3
0 . 0 3 5
0 . 0 4
0 . 0 4 5
0 . 0 5
0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1
0 . 2
0 . 4
0 . 6
0 . 8
1
dt/λ
dr/λ
(b) 2D view
Figure 3: MMSE MSE versus transmitter and receiver antennaelement spacing taking MC into account.
both MMSE and SLS methods, an apparent difference occursbetween the MSE performance with and without takingmutual coupling into account when the transmitter antennaelement spacing varies from 0.1λ to 0.5λ. The MSE obtainedfor taking into account mutual coupling effects is worsethan the MSE obtained without taking into account themutual coupling. When the antenna element spacing is largerthan 0.5λ, the difference disappears. This finding agrees withthe prediction based on the sum of channel correlationeigenvalues that shows that for the spacing between 0.1λto 0.5λ the mutual coupling undermines the estimationaccuracy.
Further investigations are done by varying the antennaelement spacing in both transmitter and receiver arrays. Theresults are given in Figures 3, 4, 5 and 6. In each figure, thereare two subfigures showing three dimensional (3D) and twodimensional (2D) views. Figure 3 shows the MMSE MSEperformance under the impact of mutual coupling. One cansee that the worst estimation error occurs when the antennaelement spacing in the transmitter and receiver is between0.3λ and 0.4λ. Figure 3 shows the MMSE MSE performancewithout taking into account the mutual coupling. When theantenna element spacing in the transmitter and receiver isbetween 0.5λ and 0.7λ, MSE achieves its best performance.Similar results are also obtained for the SLS method, aspresented in Figures 5 and 6.
Figure 7 shows the results for the MMSE estimationmethod when the training sequence is optimized. Forcomparison purposes, the results for nonoptimized MMSEare also plotted. One can see that with the optimized trainingsymbols, the channel estimation error decreases. However,
M S E v e r s u s a n t e n n a e l e m e n t s p a c i n g w i t h o u t t a k i n g M C i n t o a c c o u n t
0 0 . 2 0 . 4 0 . 6 0 . 8 1
0 0 . 5
1
0 . 0 2
0 . 0 3
0 . 0 4
0 . 0 5
M S E
0 . 0 2
0 . 0 2 5
0 . 0 3
0 . 0 3 5
0 . 0 4
0 . 0 4 5
0 . 0 5
dt/λdr/λ
(a) 3D view
0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1 0 . 0 2
0 . 0 2 5
0 . 0 3
0 . 0 3 5
0 . 0 4
0 . 0 4 5
0 . 0 5
0 . 2
0 . 4
0 . 6
0 . 8
1
dt/λ
dr/λ
(b) 2D view
Figure 4: MMSE MSE versus transmitter and receiver antennaelement spacing without taking MC into account.
the optimization does not narrow the difference between thetwo sets of results with and without taking mutual couplinginto account when the antenna element spacing is between0.1λ to 1.0λ. In the case of a compact multiple antennamobile communication device whose antenna element spac-ing is between 0.1λ to 0.5λ, the antenna decoupling is arequired to achieve a better channel estimation performance.
Figure 8 further demonstrates the advantages of theoptimization method. For the undertaken simulation, theantenna element spacing at the receiver is fixed to 1λ whilethe spacing at the transmitter is made 0.2λ. From the pre-sented results one can see that the performances of MSE withoptimization are always better than the ones without it. At alower transmit SPNR, the results for MSE with optimizationare significantly improved. The results also show that whenthe transmitter antenna element spacing is within 0.1λ to0.5λ, the MSE taking into account mutual coupling is worsethan the one without taking into account mutual coupling.This is irrespective from whether optimization is used or not.
6. Conclusion
This paper has reported the investigations on the training-based channel estimation methods for a narrowband MIMOsystem, in which both the transmitter and the receiver areequipped with multiple element antennas in the form ofwire dipoles. The investigations have included the effect ofmutual coupling in addition to spatial correlation that ispresent at the transmitting and receiving array antennas dueto surrounding scattering objects. The spatial correlationhas been taken into account using the Kronecker model, in
8 International Journal of Antennas and Propagation
0 0 . 2 0 . 4 0 . 6 0 . 8 1
0 0 . 5
1 0 . 7 6 5
0 . 7 7
0 . 7 7 5
0 . 7 8
0 . 7 8 5
S L S M S E v e r s u s a n t e n n a e l e m e n t s p a c i n g t a k i n g M C i n t o a c c o u n t
M S E
0 . 7 6 8
0 . 7 7
0 . 7 7 2
0 . 7 7 4
0 . 7 7 6
0 . 7 7 8
dt/λdr/λ
(a) 3D view
0 . 2
0 . 4
0 . 6
0 . 8
1
0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1 0 . 7 6 8
0 . 7 7
0 . 7 7 2
0 . 7 7 4
0 . 7 7 6
0 . 7 7 8
dt/λ
dr/λ
(b) 2D view
Figure 5: SLS MSE versus transmitter and receiver antenna elementspacing taking MC into account.
0 . 7 6 8
0 . 7 7
0 . 7 7 2
0 . 7 7 4
0 . 7 7 6
0 . 7 7 8
S L S M S E v e r s u s a n t e n n a e l e m e n t s p a c i n g w i t h o u t t a k i n g M C i n t o a c c o u n t
0 0 . 2 0 . 4 0 . 6 0 . 8 1
0
0 . 5
1 0 . 7 6 5
0 . 7 7
0 . 7 7 5
0 . 7 8
0 . 7 8 5
M S E
dt/λdr/λ
(a) 3D view
0 . 2
0 . 4
0 . 6
0 . 8
1
0 . 1 0 . 2 0 . 3 0 . 4 0 . 5 0 . 6 0 . 7 0 . 8 0 . 9 1 0 . 7 6 8
0 . 7 7
0 . 7 7 2
0 . 7 7 4
0 . 7 7 6
0 . 7 7 8
dt/λ
dr/λ
(b) 2D view
Figure 6: SLS MSE versus transmitter and receiver antenna elementspacing without taking MC into account.
which correlation matrices for the transmitting and receivingsides are separated. The mutual coupling has been includedusing the closed-form expressions for impedance matrices ofparallel side-to-side wire dipoles. Two cases of nonoptimizedand optimized training sequence transmission schemes havebeen considered. The computer simulations have been
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10.01
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05MMSE MSE versus transmitter antenna spacing
MSE
Non-optimized MSE taking MC into accountOptimized MSE taking MC into accountNon-optimized MSE without taking MC into accountOptimized MSE without taking MC into account
dt/λ
Figure 7: Optimized and nonoptimized MMSE MSE versustransmitter antenna element spacing with and without takingmutual coupling effects into account.
0 5 10 150
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
2MSE versus transmit SPNR
Transmit signal power to noise ratio (dB)
MSE
Optimized MSE taking MC into account
Optimized MSE without taking MC into account
Nonoptimized MSE taking MC into account
Nonoptimized MSE without taking MC into account
Figure 8: MMSE MSE for versus transmit SPNR with and withouttaking mutual coupling effects into account for optimized andnonoptimized cases.
carried out for the case when the transmitting and receivingsides are equipped with four-element uniform half-wavedipole arrays surrounded by circles of uniformly distributedscattering objects. The obtained simulation results haveshown that at the antenna (dipole) spacing between 0.1λ to0.5λ, mutual coupling decreases the spatial correlation and
International Journal of Antennas and Propagation 9
adversely affects the channel estimation offered by MMSEand SLS methods. When the spacing is larger than 0.5λthe MSE is slightly different when the mutual couplingis taken into account or neglected. Next considerationshave focused on optimization of training signals for MIMOchannel estimation. When the optimization of the trainingsequence is applied, an improvement in MSE has beendemonstrated irrespectively whether the mutual couplingeffects are neglected or taken into account. This improve-ment is more significant at a lower transmit SPNR. Thedifference in MSE for with and without taking into accountmutual coupling effects is unchanged when the antennaelement spacing is within 0.1λ to 0.5λ. This means thatfor a compact multiple antenna mobile communicationdevice the improvement of MSE requires decoupling of theantenna elements to achieve a better estimation of MIMOchannel.
References
[1] E. Telatar, “Capacity of multi-antenna Gaussian channels,”European Transactions on Telecommunications, vol. 10, no. 6,pp. 585–595, 1999.
[2] T. L. Marzetta and B. M. Hochwald, “Capacity of a mobilemultiple-antenna communication link in Rayleigh flat fading,”IEEE Transactions on Information Theory, vol. 45, no. 1, pp.139–157, 1999.
[3] A. Narula, M. J. Lopez, M. D. Trott, and G. W. Wornell,“Efficient use of side information in multiple-antenna datatransmission over fading channels,” IEEE Journal on SelectedAreas in Communications, vol. 16, no. 8, pp. 1423–1436, 1998.
[4] C. Budianu and L. Tong, “Channel estimation for space-time orthogonal block codes,” IEEE Transactions on SignalProcessing, vol. 50, no. 10, pp. 2515–2528, 2002.
[5] A. Grant, “Joint decoding and channel estimation for linearMIMO channels,” in Proceedings of the IEEE Wireless Commu-nications and Networking Conference, pp. 1009–1012, Chicago,Ill, USA, September 2000.
[6] A. S. Kyung, R. W. Heath, and B. K. Heung, “Shannoncapacity and symbol error rate of space-time block codesin MIMO Rayleigh channels with channel estimation error,”IEEE Transactions on Wireless Communications, vol. 7, no. 1,pp. 324–333, 2008.
[7] X. Zhang and B. Ottersten, “Performance analysis of V-BLASTstructure with channel estimation errors,” in Proceedings of the4th IEEE Workshop on Signal Processing Advances in WirelessCommunications (SPAWC ’03), pp. 487–491, Rome, Italy, June2003.
[8] P. Layec, P. Piantanida, R. Visoz, and A. O. Berthet, “Capacitybounds for MIMO multiple access channel with imperfectchannel state information,” in Proceedings of the IEEE Informa-tion Theory Workshop (ITW ’08), pp. 21–25, Porto, Portugal,May 2008.
[9] M. Biguesh and A. B. Gershman, “MIMO channel estimation:optimal training and tradeoffs between estimation tech-niques,” in Proceedings of the IEEE International Conferenceon Communications (ICC ’04), vol. 5, pp. 2658–2662, Paris,France, June 2004.
[10] M. Biguesh and A. B. Gershman, “Training-based MIMOchannel estimation: a study of estimator tradeoffs and optimaltraining signals,” IEEE Transactions on Signal Processing, vol.54, no. 3, pp. 884–893, 2006.
[11] T. Svantesson and A. Ranheim, “Mutual coupling effects onthe capacity of multi-element antenna systems,” in Proceedingsof the IEEE International Conference on Acoustics, Speech, andSignal Processing (ICASSP ’01), pp. 2485–2488, Las Vegas, Nev,USA, April 2001.
[12] J. W. Wallace and M. A. Jensen, “The capacity of MIMOwireless systems with mutual coupling,” in Proceedings of theIEEE Vehicular Technology Conference (VTC ’02), pp. 696–700,Vancouver, Canada, 2002.
[13] J. W. Wallace and M. A. Jensen, “Mutual coupling in MIMOwireless systems: a rigorous network theory analysis,” IEEETransactions on Wireless Communications, vol. 3, no. 4, pp.1317–1325, 2004.
[14] P. N. Fletcher, M. Dean, and A. R. Nix, “Mutual couplingin multi-element array antennas and its influence on MIMOchannel capacity,” Electronics Letters, vol. 39, no. 4, pp. 342–344, 2003.
[15] S. Lu, H. T. Hui, M. Bialkowski, and X. Liu, “Optimizingtransmission strategy in single-user MIMO systems under theinfluence of antenna mutual coupling,” ECTI Transactions onElectrical Engineering Electronics, and Communications, vol. 6,no. 2, pp. 73–78, 2008.
[16] S. Lu, X. Liu, H. T. Hui, M. Bialkowski, N. Seman, andH. Y. Zhang, “Power allocation strategy for compact MIMOtransmitters with Uniform Circular Arrays under the influenceof antenna mutual coupling,” in Proceedings of the IEEEAsia Pacific Microwave Conference (APMC ’07), pp. 11–14,Bangkok, Thailand, December 2007.
[17] S. Lu, H. T. Hui, M. Bialkowski, and X. Liu, “Optimizingtransmission strategy in single-user MIMO systems underthe influence of antenna mutual coupling,” in Proceedings ofthe IEEE International Symposium on Communications andInformation Technologies (ISCIT ’07), pp. 370–374, Sydney,Australia, October 2007.
[18] M. E. Bialkowski, P. Uthansakul, K. Bialkowski, and S.Durrani, “Investigating the performance of MIMO systemsfrom an electromagnetic perspective,” Microwave and OpticalTechnology Letters, vol. 48, no. 7, pp. 1233–1238, 2006.
[19] X. Liu, S. Lu, M. E. Bialkowski, and H. T. Hui, “MMSEchannel estimation for MIMO system with receiver equippedwith a circular array antenna,” in Proceedings of the Asia-Pacific Microwave Conference (APMC ’07), Bangkok, Thailand,December 2007.
[20] S. Lu, H. T. Hui, M. E. Bialkowski, X. Liu, H. S. Lui, andN. V. Shuley, “The effect of antenna mutual coupling onchannel estimation of MIMO-OFDM systems,” in Proceedingsof the IEEE Antennas and Propagation Society InternationalSymposium (APS ’07), pp. 2945–2948, Honolulu, Hawaii,USA, June 2007.
[21] S. M. Kay, Fundamentals of Statistical Signal Processing:Estimation Theory, Prentice-Hall, Upper Saddle River, NJ,USA, 1993.
[22] E. G. Larsson and P. Stoica, Space-Time Block Coding for Wire-less Communication, Cambridge University Press, Cambridge,UK, 2003.
[23] C.-N. Chuah, D. N. C. Tse, J. M. Kahn, and R. A. Valenzuela,“Capacity scaling in MIMO wireless systems under correlatedfading,” IEEE Transactions on Information Theory, vol. 48, no.3, pp. 637–650, 2002.
[24] S. Da-Shan, G. J. Foschini, M. J. Gans, et al., “Fading corre-lation and its effect on the capacity of multi-element antennasystems,” in Proceedings of the IEEE International Conferenceon Universal Personal Communications (ICUPC ’98), Florence,Italy, 1998.
10 International Journal of Antennas and Propagation
[25] X. Liu, M. E. Bialkowski, and F. Wang, “Investigations into theeffect of spatial correlation on channel estimation and capacityof multiple input multiple output system,” InternationalJournal of Communications, Network and System Sciences, vol.2, no. 3, 2009.
[26] S. Durrani and M. E. Bialkowski, “Effect of mutual couplingon the interference rejection capabilities of linear and circulararrays in CDMA systems,” IEEE Transactions on Antennas andPropagation, vol. 52, no. 4, pp. 1130–1134, 2004.
[27] H. T. Hui, “Improved compensation for the mutual couplingeffect in a dipole array for direction finding,” IEEE Transactionson Antennas and Propagation, vol. 51, no. 9, pp. 2498–2503,2003.
International Journal of
AerospaceEngineeringHindawi Publishing Corporationhttp://www.hindawi.com Volume 2010
RoboticsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Active and Passive Electronic Components
Control Scienceand Engineering
Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
International Journal of
RotatingMachinery
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporation http://www.hindawi.com
Journal ofEngineeringVolume 2014
Submit your manuscripts athttp://www.hindawi.com
VLSI Design
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Shock and Vibration
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Civil EngineeringAdvances in
Acoustics and VibrationAdvances in
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Electrical and Computer Engineering
Journal of
Advances inOptoElectronics
Hindawi Publishing Corporation http://www.hindawi.com
Volume 2014
The Scientific World JournalHindawi Publishing Corporation http://www.hindawi.com Volume 2014
SensorsJournal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Modelling & Simulation in EngineeringHindawi Publishing Corporation http://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
Navigation and Observation
International Journal of
Hindawi Publishing Corporationhttp://www.hindawi.com Volume 2014
DistributedSensor Networks
International Journal of
Top Related