Bhuvaneswara -International Journal of Computer Science information and Engg.

, Technologies ISSN 2277-4408 || 01062013-006

A New Approach for Finite-Length Framework of Multi-Input Multi-Output

MMSE
1

A. Bhuvaneswara Rao, 2 Sreekanth Kavuri, 3 Tirupathi Rao Gullipalli

1,2
IT Mentor,CSE Faculty, RGUKT Nuzvid
3
Asst.Prof, JBIET,Rangareddy,Hyderabad
1
bhuvan.allu@gmail.com, 2skavuri47@gmail.com ,3gullipallitirupathi@gmail.com

Abstract
In this paper this framework is introduced for
analyzing the performance of a finite-length
minimum-mean-square error decision feedback
equalizer (MMSE-DFE) in a multi-input multi-output
(MIMO) environment. The framework includes
transmit and receive diversity systems as special
cases and quantifies the diversity performance
improvement as a function of the number of
transmit/receive antennas and equalizer taps. Fast and
parallelizable algorithms for computing the finitelength MIMO MMSE-DFE are presented for three
common multi-user detection scenarios.
Keywords Equalizers, FIR digital filters, mean
square error methods, MIMO systems, Explicit rate,
flow control.

I. Introduction
MULTI-USER
communication
over
linear,
dispersive, and noisy channels, the received signal is
composed of the sum of several transmitted signals
corrupted by inter-symbol interference (ISI), interuser interference (IUI), and noise. Examples include
TDMA digital cellular systems with multiple
transmit/receive antennas [4], [25], wideband
asynchronous CDMA systems [3], where IUI is also
known as multiple access interference (MAI),
wideband transmission over digital subscriber lines
(DSL) [1], where IUI takes the form of near-end and
far-end crosstalk between adjacent twisted pairs, and
high-density digital magnetic recording, where IUI is
due to interference from adjacent tracks [15]. This
work is motivated by the desire to increase the
capacity of digital TDMA wireless networks by
allowing multiple transmissions sharing the same
time slot and frequency band and separating them
spatio-temporally at the receiver. On the uplink, this
corresponds to, for example, multiple synchronous
co-channel and co-cell users, where each is equipped
with single (or multiple) antennas transmitting in the
same time slot and received using an antenna array at
the base station. On the downlink, the base station
antenna array provides transmitter spatial diversity
IJCSIET-ISSUE3-VOLUME2-SERIES2

(and possibly additional coding gain using space-time

coding technology [25]), and spatio-temporal
processing is performed at the mobiles to detect the
multiple-input signals Multiuser detection techniques
for MIMO systems have been shown to offer
significant performance advantages over single-user
detection techniques that treat IUI as additive colored
noise and lump its effects with background noise.
Recently, it has been shown that the presence of ISI
in these MIMO systems could enhance overall
system capacity significantly, provided that effective
multiuser detection techniques are employed [12],
[13] The optimum maximum likelihood sequence
estimation (MLSE) receiver for MIMO channels was
developed in [26]; however, its exponential
complexity increase with the number of users and
channel memory makes its implementation costly for
multiuser detection on severe-ISI channels, especially
as the input signal constellation size increases to
improve spectral efficiency.2 Two alternative lower
complexity transceiver structures, which are widely
used in practice for single-input single-output (SISO)
dispersive channels, namely, discrete multi-tone
(DMT) and minimum-mean-square-error decision
feedback equalizer (MMSE-DFE), have been
recently proposed for MIMO dispersive channels as
well [2], [3]. In this paper, we present a new
analytical framework for analyzing the MIMO
MMSE-DFE that extends the work in [6] to the
MIMO case in a manner that is distinct from the
work in [2], [3], [16], and [17] in three key aspects.
First, the MIMO MMSE-DFE feed forward and
feedback matrix filters are restricted to be finite
impulse response (FIR) for practical implementation,
3 and the decision delay is optimized, thus
establishing finite-length analogs of the results in [2],
[3], [16], and [17]. Second, the assumption of an
equal number of channel inputs and outputs made in
[2] and [3] is relaxed.4 Third, the special structure of
the problem is exploited to derive fast and
parallelizable MIMO MMSE-DFE computation
algorithms suitable for real-time implementation. As
shown in [3], computing the MIMO MMSE-DFE for
the infinite-length case requires computationally
intense spectral factorizations of matrix rational
spectra. The organization of this paper is as follows.

Page 1

Bhuvaneswara -International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062013-006

We start in Section II by developing the input-output

model for the MIMO MMSE-DFE and deriving
closed-form expressions for its optimal filter settings.
Several special cases of the MIMO MMSE-DFE
structure are discussed. The notation adopted in this
paper conforms to the following convention
Scalars are denoted in lower case. Unless otherwise
stated, vectors are column vectors and are denoted in
lower case bold.
denotes the unit vector (it has a one in the th
position and zeros everywhere else).
In situations where the components of the vectors
are to be emphasized, the first and last components,
separated
by a colon, are given as a subscript to the vector:
Matrices are upper case bold: .
denotes the identity matrix of size .
denotes an all-zeros matrix with rows and columns.
denotes the determinant of matrix .
trace denotes the trace of matrix .
denotes the expected value operator.
A diagonal matrix with elements on the main
diagonal will be denoted.
The symbol will be used to denote the complexconjugate transpose of a matrix or a vector and the
complex
conjugate of a scalar.
The symbol will be used to denote the transpose of
a matrix or a vector.
When a sub-matrix of a given matrix needs to be
specified, indices of the first and last rows (and
columns) of
the sub-matrix, separated by colons, are used as the
first (and second) component of an argument to the
matrix
For convenience, we summarize in Table I the key
matrices used in this paper and their sizes.

II. MIMO MMSE-DFE

We start in this section by describing the input
output model assumed throughout the paper. Then,
we derive closed-form expressions for the optimum
filter settings of the finite-length MIMO MMSE-DFE
under three multiuser detection scenarios. In the first
scenario, only previous decisions of other users are
available at the input of the feedback filter for any
user. In the second scenario, previous decisions as
well as current decisions of lower indexed users are
available. This scenario assumes that decisions are
made sequentially, starting with lower indexed users.
In the third scenario, previous and current decisions
of all others users are assumed available (e.g., from a
previous detection stage).
InputOutput Model
IJCSIET-ISSUE3-VOLUME2-SERIES2

We consider the general case of a linear, dispersive,

and noisy digital communication system with inputs
and outputs.

Fig1. Block Diagram of the multi-input multi-output

channel
We use the standard complex-valued baseband
equivalent signal model. Assuming an oversampling
factor of , the samples at the channel output ( ) have
the standard form (see Figs. 1 )

All of these three quantities are column vectors

corresponding to the time samples per symbol in the
assumed temporally oversampled channel model.
Furthermore, the overall channel impulse response
between the input and the output is represented by
the vector . By grouping the received samples from
all channel outputs at symbol time into an column
vector , we can relate to the corresponding column
vector of input.

III. Related Work

The studied multicast congestion control scheme, we
focus on the following two simulation models, and
are mostly interested in analyzing the transient
behaviors of the network. In the performance
analysis, the duration of response time, receiving rate
of receivers, and steady state of buffer occupancy are
the main concerns. From the view of control theory, a
control scheme with short response time has the

Page 2

Bhuvaneswara -International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062013-006

following advantages: when the buffer of receiver

nodes is close to the threshold, one may tell the
sending node to reduce the sending rate and prevent
the loss of packets as soon as possible; while when
the available bandwidth increases, the sending node
increases the sending rate as soon as possible and
enhances the utilization rate of the bandwidth. In
simulations, we process the nodes that have small
differences of time delay and sending rate together.
Then we unify the time delay and sending rate. Since
the situation of every node in each group (about 20
receivers) is similar, we only choose one node from
each group as a representative. We assume that the
link delay is dominant compared to the other delays,
such as processing delays and queuing delay. The
FIR MIMO MMSE-DFE consists of a feed forward
filter matrix

with matrix taps , each of size , and a feedback filter

matrix equal to

where we introduced the matrices . When dealing

with
multidimensional error random processes,
either the trace or the determinant of can be used as a
mean square error measure. The trace measure is
equal to the arithmetic average of the given values of
, whereas the determinant measure is related to their
geometric average. For the MIMO MMSE-DFE, it
turns out that the same feed forward and feedback
coefficients minimize both measures; hence, either
one can be adopted. Corresponding to these two
measures, we define the following two decision-point
SNR performance measures for the MIMOMMSEDFE. A second approach for computing the optimum
FIR MIMO MMSE-DFE filter settings under
IJCSIET-ISSUE3-VOLUME2-SERIES2

Scenario 2 is by performing a standard (i.e., not

block) Cholesky factorization of the matrix in the
form . Then, the feedback filter matrix is given by the
adjacent columns of that correspond to a diagonal
matrix with the smallest trace. Therefore, (20) and
(23) are used to compute MIMO MMSE-DFE filter
settings with the understanding that is now a lower
triangular matrix and not a block lower triangular
matrix. The equivalence of the two approaches can be
easily shown using the nesting property of Cholesky
factorization. This property states that if , then the
lower triangular and diagonal Cholesky factors of the
sub-matrix of formed from its first rows and columns
are equal to the first rows and columns of and ,
respectively. Conversely, if , then the uppertriangular and diagonal Cholesky factors of the submatrix of formed from its last rows and columns are
equal to the last rows and columns of and ,
respectively. When multistage detectors [27] are
employed, current decisions from all other users,
which are obtained from a previous detection stage,8
are available for the detection of the user of interest.
Therefore, suppressing their interfering effects would
improve the performance of the MIMO MMSE-DFE.
This detection scenario has the same mathematical
formulation presented for completeness of the
analysis; however, our main focus in this paper will
be on multistage detection will not be discussed any
further. Remark: In concluding this section, we
would like to emphasize the importance of
optimizing the decision delay parameter in all three
scenarios, especially for short feed forward filters.
Using a suboptimum can result in significant
performance degradation, as it will be demonstrated
in the simulation results.. This is also the case for the
SISO MMSE-DFE, as shown in [7]. For the MIMO
case, allowing the different users to have different
decision delays could result in improved decisionpoint SNR for the MIMO MMSE-DFE. However, it
might violate the assumption of available previous
decisions from all other users in Scenario 1 and the
additional assumption of availability of current
decisions from lower indexed users made in Scenario
2. Other drawbacks of allowing variable user
decision delays include the increased computational
complexity in optimizing these variable delays and
computing the optimum MIMO MMSE-DFE settings
and the fact that the resulting feedback filters will be
of different lengths in general. For all of the above
reasons, we do not explore this variable-delay
detection strategy in this paper.

Cooperative MIMO scheme

For data transmission from source node S to
destination node D over distance d, instead of SISO
direct transmission which is not practical for long

Page 3

Bhuvaneswara -International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062013-006

range, we can create a C-MIMO transmission, to

reduce the transmit energy. In the transmission side,
node S can cooperate with its neighbors and
exchange its data (the distance between cooperating
nodes dc << d). MIMO techniques (STBC, SpaceTime Trellis Code (STTC), Spatial Multiplexing,
etc.) are then employed to transmit their data
simultaneously to the destination node (or multidestination cooperative nodes) like a multi-antenna
diversity system (each cooperative node plays role of
one antenna of MIMO system). In the reception side
the cooperative neighbors of destination node D
receive the MIMO modulated symbols and,
respectively, retransmit them to the destination node
D for joint MIMO signals combination. However, if
the C-MIMO scheme can exploit the energy-efficient
transmission of MIMO technique, the local data
transmission at TX and RX sides of C-MIMO
scheme costs an extra transmission energy due to the
extra circuit consumption of the cooperative nodes
and the more complex MIMO digital signal
processing. For short range transmission, this extra
energy consumption can be greater than the
transmission energy saved by using C-MIMO (or
MISO) instead of SISO technique. Another C-MIMO
trade-off is the delay of the cooperative local data
transmission. Nevertheless, transmission delay is a
less important design criterion than energy
consumption and in comparison to multi-hop
technique.
Application of STBC
Among MIMO diversity coding techniques (Space
Time Block or Trellis Codes, Spatial Multiplexing),
STBC is the most practical for WSN [9]. The
simplicity of ST coding and combination is very
interesting due to the calculation limitation of the
sensor node (decoding algorithm of STBC is only
based on linear processing). The diversity Alamouti
code [7] is used, for systems with 2 transmission
antennas, whereas the orthogonal STBC or complex
symbol signals developed by Tarokh [8] is used for
systems with 38 transmission antennas. Due to the
diversity of transmission and reception, BER
performance of MIMO STBC can easily outperform
SISO system under the same Signal-to-Noise Ratio
Eb=N0. In other words, with the same BER
requirement, MIMO system requires less energy for
the transmission than SISO system. We assume that
we have perfect synchronization, perfect channel
estimation and Maximum Likelihood detection in the
receiver.
A note on time synchronization
In principle, the RTS/CTS transmission between
source and destination exists in many medium access
IJCSIET-ISSUE3-VOLUME2-SERIES2

control (MAC) protocols. The reception of the CTS

packet triggers at each relay the initiation of the
timing process, within an uncertainty interval that
depends on different propagation times[10]. Since
source CH and receiver CH have direct
communication link, therefore, an explicit time
synchronization protocol among the cooperative
nodes is not required [10]

MMSE Algorithm

In this section, we give a rough order-of-magnitude

estimate of the computational complexity involved
in computing the optimum filter coefficients of the
MIMO MMSE-DFE using (20) and (22) Several
assumptions are made. First, we use the number of
instructions per coefficient update as a complexity
measure. Second, we only count complex multiplies
and assume that each complex multiply is equivalent
to six instructions (four real multiplies and two real
adds).10 Third, we assume white input and noise
sequences. Finally, the complexity incurred in
estimating the MIMO channel matrix is not included.
We evaluate the computational complexity as a
function of the number of transmit and receive
antennas, number of equalizer taps, and the MIMO
channel memory. The results are shown in Figs. 4
and 5. The instruction counts given in these figures

Page 4

Bhuvaneswara -International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062013-006

can be readily translated into MIPS estimates through

multiplication by the MIMO MMSE-DFE coefficient
update rate, which is a design parameter set
according to the dynamics of channel time variations.
As an example, assuming a 2-ms update rate,11 then
computing the coefficients of a -spaced MIMO
MMSE-DFE with (i.e., a total of 16 feed forward
taps), four feedback taps for a channel memory of 4
and two antennas at each of the transmit and receive
ends requires around 100 MIPS. This is well within
the
processing
power
of
state-of-the-art
programmable DSP chips like the TMS320C6X
family of processors from Texas Instruments.

IV. CONCLUSION
In this paper, we propose a solution for non-uniform
energy consumption in the cluster based multi-hop
WSN through adaptive selection of transmission rate
and MIMO schemes. We have shown that the multihop MIMO in transit clusters results in reduced
energy consumption as compared to the noncooperative case, and MIMO transmission provides
uniform energy consumption within a tolerance range
e in all the clusters. The proposed MIMO
communication architecture can also offer substantial
energy saving in Wireless Sensor Networks provided
that the system is designed judiciously. These include
careful consideration of transmission distance
requirements and joint rate and number of
cooperative nodes optimization. New closed-form
expressions for the optimum finite-length MIMO
MMSE-DFE filters and decision-point SNR were
derived under the three most common multiuser
detection scenarios. Fast and parallelizable
algorithms for computing these filters were derived
by exploiting the block-Toeplitz structure of the
MIMO channel matrix. The presented analytical
framework allows for quick MIMO MMSE-DFE
performance evaluation as a function of the number
of transmit antennas, receive antennas, and filter taps.

Future Work
The Future Enhancement of proposed solution for
non-uniform energy consumption in the cluster based
multi-hop WSN through adaptive selection of
transmission rate and C-MIMO schemes. We have
shown that the multi-hop C-MIMO in transit clusters
results in reduced energy consumption as compared
to the non-cooperative case, and ARC-MIMO
transmission provides uniform energy consumption
within a tolerance range e in all the clusters. The
IJCSIET-ISSUE3-VOLUME2-SERIES2

proposed ARC-MIMO communication architecture

can also offer substantial energy saving in Wireless
Sensor Networks provided that the system is
designed judiciously. These include careful
consideration of transmission distance requirements
and joint rate and number of cooperative nodes
optimization. The conclusions drawn in this paper
only apply to the dense WSN where appropriate
number of close cooperative nodes are available.

References
[1] Special Issue on High-Speed Digital Subscriber
Lines, IEEE J. Select. Areas Commun., vol. 9, Aug.
1991.
[2] A. Duel-Hallen, Equalizers for multiple input
multiple output channels and PAMsystems with
cyclostationary input sequences, IEEE J. Select.
Areas Commun., vol. 10, pp. 630639, Apr. 1992.
[3] , A family of multiuser decision-feedback
detectors for asynchronous code-division multipleaccess channels, IEEE Trans.
Commun., vol. 43, pp. 421434, Feb./Mar./Apr.
1995.
[4] A. Naguib, V. Tarokh, N. Seshadri, and A. R.
Calderbank, A space-time coding modem for highdata-rate wireless communications, IEEE J. Select.
Areas Commun., vol. 16, pp. 14591477, Oct. 1998.
[5] N. Al-Dhahir, A computationally-efficient FIR
MMSE-DFE for CCIimpaired dispersive channels,
IEEE Trans. Signal Processing, vol. 45,
pp. 3241, Jan. 1997.
[6] N. Al-Dhahir and J. M. Cioffi, MMSE decision
feedback equalizers: Finite-length results, IEEE
Trans. Inform. Theory, vol. 41, pp. 961976, July
1995.
[7] , Efficient computation of the delay-optimized
finite-length MMSE-DFE, IEEE Trans. Signal
Processing, vol. 44, pp. 12881292,
May 1996.
[8] P. Balaban and J. Salz, Optimum diversity
combining and equalization in digital data
transmission with applications to cellular mobile
radioPart I: Theoretical considerations, IEEE
Trans. Commun., vol. 40, pp. 885894, May 1992.
[9] A. W. Bojanczyk, R. P. Brent, F. R. de Hoog, and
D. R. Sweet, On the stability of the babeiss and
related toeplitz factorization algorithms,
SIAM J. Matrix Anal. Applicat., vol. 16, pp. 4057,
Jan. 1995.
[10] S. Chandrasekaran and A. H. Sayed, Stabilizing
the generalized Schur algorithm, SIAM J. Matrix
Anal. Applicat., vol. 17, no. 4, pp. 950983, Oct.
1996.

Page 5

Bhuvaneswara -International Journal of Computer Science information and Engg., Technologies ISSN 2277-4408 || 01062013-006

[11] A. Furuskar, S. Mazur, F. Muller, and H.

Olofsson, EDGE: Enhanced data rates for GSM and
TDMA/136 evolution, IEEE Personal
Commun. Mag., June 1999.
[12] G. Raleigh and J. Cioffi, Spatio-temporal
coding for wireless communication, IEEE Trans.
Commun., vol. 46, pp. 357366, Mar. 1998.
[13] G. J. Foschini and M. J. Gans, On limits of
wireless communication in a fading environment
when using multiple antennas, Wireless Personal
Commun., pp. 311335, Mar. 1998.
[14] G. Golub and C. Van Loan, Matrix
Computations, 2nd ed. Baltimore, MD: Johns
Hopkins University Press, 1989.
[15] J. Cioffi, W. Abbott, H. Thapar, C. Melas, and
K. Fisher, Adaptive equalization for magnetic disk
storage channels, IEEE Commun. Mag.,
pp. 1429, Feb. 1990.
[16] T.L. Marzetta, B.M. Hochwald, Capacity of a
mobile multiple-antenna communication link in
Rayleigh flat fading, IEEE Transactions on
Information
Theory 45 (1999) 139157.
[17] S. Cui, A.J. Goldsmith, A. Bahai, Energyconstrainted
modulation
optimization,
IEEE
Transactions on Wireless Communications 4 (5)
(2005) 23492360.
[18] M. Gastpar, B. Rimoldi, M. Vetterli, To code or
not to code: Lossy sourcechannel communication
revisited, IEEE Transactions on Information Theory
49
(2003) 11471158.
[19] J.-H. Chang, L. Tassiulas, Energy conserving
routing in wireless ad-hoc networks, in: Proc. IEEE
INFOCOM, 2000, pp. 2231.
[20] S.K. Jayaweera, Energy analysis of mimo
techniques in wireless sensor networks, in: 38th
Annual Conference on Information Sciences and
Systems,
Princeton University, USA, March 2004.
[21] . Cui, A.J. Goldsmith, A. Bahai, Energyefficiency of mimo and cooperative mimo techniques
in sensor networks, IEEE Journal on Selected Areas
in
Communications 22 (6) (2004) 10891098.
[22] S.M. Alamouti, A simple diversity technique for
wireless communications, IEEE Journal on Selected
Areas in Communications 16 (8) (1998) 14511458.
[23] V. Tarokh, H. Jafarkhani, A.R. Calderbank,
Space-time block codes fromorthogonal designs,
IEEE Transactions on Information Theory 45 (5)
(1999) 14561467.

IJCSIET-ISSUE3-VOLUME2-SERIES2

[24] L.X.H. Dai, Q. Zhou, Energy efficiency of

mimotransmission strategies in wireless sensor
networks, in: International Conference on
Computing,
Communications
and
Control
Technologies (CCCT), Austin, TX, USA, August
2004.
[25] H. Lev-Ari and T. Kailath, Triangular
factorization of structured hermitian matrices, in
Oper. Theory: Adv. Appl.,, I. Gohberg, Ed., 1986,
vol. 18, pp. 301324.
[26] M. Stewart and P. Van Dooren, Stability issues
in the factorization of structured matrices, SIAM J.
Matrix Anal. Appl., pp. 104118, 1997.
[27] J. Proakis, Digital Communications, 3rd ed. New
York: McGraw-Hill, 1995.
[28] T. Kailath and A. H. Sayed, Eds., Fast Reliable
Algorithms for Matrices with Structure. Philadelphia,
PA: SIAM, 1999.
[29] V. Tarokh, N. Seshadri, and A. R. Calderbank,
Space-time codes for high data rate wireless
communications: Performance criterion and code
construction, IEEE Trans. Inform. Theory, vol. 44,
pp. 744765, Mar. 1998.
[30] S. Verdu, Minimum probability of error for
asynchronous Gaussian multiple-access channels,
IEEE Trans. Inform. Theory, vol. 42, pp.
8596, Jan. 1996.
[31] , Multiuser Detection. Cambridge, U.K.:
Cambridge University Press, 1998.
Bhuveneswara Rao working as CSE
faculty in RGUKT College Nuzvid.

Sreekanth Kavuri working as CSE faculty

in RGUKT College Nuzvid.

Tirupathi Rao Working as Assistant

Professor in JBIET College Rangareddy,
Hyderabad.

Page 6