Sunteți pe pagina 1din 7

962

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 53, NO. 6, JUNE 2005

Switching Between Diversity and Multiplexing in MIMO Systems


Robert W. Heath, Jr. and Arogyaswami J. Paulraj

AbstractMultiple-input multiple-output (MIMO) wireless


communication systems can offer high data rates through spatial
multiplexing or substantial diversity using transmit diversity. In
this letter, switching between spatial multiplexing and transmit
diversity is proposed as a simple way to improve the diversity
performance of spatial multiplexing. In the proposed approach,
for a fixed rate, either multiplexing or diversity is chosen based
on the instantaneous channel state and the decision is conveyed
to the transmitter via a low-rate feedback channel. The minimum Euclidean distance at the receiver is computed for spatial
multiplexing and transmit diversity and is used to derive the
selection criterion. Additionally, the Demmel condition number
of the matrix channel is shown to provide a sufficient condition
for multiplexing to outperform diversity. Monte Carlo simulations
demonstrate improvement over either multiplexing or diversity
individually in terms of bit error rate.
Index TermsMultiple-input multiple-output (MIMO) systems,
space-time coding, transmit diversity.

I. INTRODUCTION
UTURE wireless systems will employ multiple antennas
at both transmitter and receiver to improve quality, capacity, and reliability. Achieving these goals, however, requires
careful design of the modulation and coding scheme to fully
exploit all the degrees of freedom offered by the multiple-input
multiple-output (MIMO) channel. Two popular approaches
for communicating in the MIMO channel are diversity and
multiplexing. MIMO diversity,1 or transmit diversity (e.g.,
[1][5]), is an approach whereby information is spread across
multiple transmit antennas to maximize the diversity advantage
in fading channels. Spatial multiplexing, on the other hand, is
an approach where the incoming data is divided into multiple
substreams and each substream is transmitted on a different
transmit antenna [6], [7]. In independent identically distributed
(i.i.d.) Rayleigh matrix channels, spatial multiplexing can
achieve the full ergodic capacity (with a suitable outer code)
unlike diversity transmission [8], [9] but does not offer the same
diversity gain as with MIMO diversity [10]. Since practical

Paper approved by H. El-Gamal, the Editor for Space-Time Coding and


Spread Spectrum of the IEEE Communications Society. Manuscript received
April 9, 2001; revised January 29, 2003, and September 1, 2004. A portion of
these results appeared in the Proceedings of the 2000 Allerton Conference on
Communication, Control, and Computers, Urbana, IL, and the Proceedings of
the 2001 International Conference on Communications, Helsinki, Finland. This
work was completed while R. W. Heath, Jr. was a Ph.D. student at Stanford
University.
R. W. Heath, Jr. is with the Department of Electrical and Computer Engineering, Wireless Networking and Communications Group, The University of
Texas at Austin, Austin, TX 78712-1084 USA (e-mail: rheath@ece.utexas.
edu).
A. J. Paulraj is with the Information Systems Laboratory, Stanford University,
Stanford, CA 94305-9510 USA (e-mail: apaulraj@stanford.edu).
Digital Object Identifier 10.1109/TCOMM.2005.849774
1We use the term MIMO diversity, instead of transmit diversity, to emphasize
that we are considering transmit diversity with multiple receive antennas.

systems require some trade-off between rate and reliability, it is


unclear how to choose between spatial multiplexing or MIMO
diversity.
In this letter we consider the problem of switching between
multiplexing and diversity based on the instantaneous channel
state. This requires a low-rate (one binary decision per update)
feedback channel from the receiver to the transmitter, which
is assumed to be zero-delay and error-free. We formulate the
switching criterion from the point-of-view of the minimum
Euclidean distance of the codebook (or constellation) at the
receiver (i.e., the codebook constructed when the channel
operates on each codeword). This is motivated by the fact that,
assuming maximum-likelihood detection, the conditional error
probability given a channel realization is determined by the
distance properties of the codebook at the receiver. We focus
on the minimum Euclidean distance since, using the union
bound, it can be used to bound the codeword error probability
[11][13] (see also examples in ([14], p. 31) for discussion of
the tightness of this bound). We consider only communication
in frequency-flat channels where the channel is known perfectly
to the receiver but not to the transmitter. Error correction coding
is not considered. To make the comparison fair, the overall data
rate is fixed (note that this means that the size of the constellations for diversity and multiplexing will be different). Since
an analysis that is valid for all types of diversity techniques is
difficult, we consider only a restricted class of space-time block
codes that includes the Alamouti scheme [3] and orthogonal
designs as special cases. These codes have a Euclidean distance
that is proportional to the Frobenius norm of the matrix channel
and provide a relevant upper bound on the Euclidean distance
for other practical space-time codes such as delay-diversity [1],
[2]. We consider a standard spatial multiplexing system [15],
or BLAST system [6], assuming the use of the maximum-likelihood receiver. In related work we have considered extensions
to other spatial multiplexing receivers as well other diversity
techniques such as selection diversity [16].
To motivate selection, we compute the Euclidean distance
at the receiver for each of spatial multiplexing and space-time
block coding. The outcomes of this computation are two-fold.
First, we find a useful measure for characterizing matrix channels based on the Demmel condition number of the channel [17].
Other authors have used the regular condition number (or its reciprocal) to evaluate the quality of matrix channels since it provides some intuition on channel quality (see, e.g., [18], [19]).
Our results, however, show that the Demmel condition number
may be more useful than the regular condition number since we
can relate the Demmel condition number to a sufficient condition for multiplexing to be better than diversity. Second, we
show that the minimum Euclidean distance is a relevant measure
for selecting between multiplexing and diversity even when the
overall data rate is fixed. This is because the transformation by

0090-6778/$20.00 2005 IEEE

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 53, NO. 6, JUNE 2005

the channel generally changes the Euclidean distance properties


of the codebook at the receiver. Since multiplexing and diversity
codebooks generally have very different distributions of codewords, one codebook or the other is typically more affected
by the channel. Simulations illustrate improvements over static
spatial multiplexing and MIMO diversity.
Recently there has been interest in diversity and multiplexing
from the point-of-view of making a diversity-multiplexing
tradeoff [20][23]. Compared with this work, we take a pragmatic view based on instantaneous channel state information.
The work in [20] studies the multiplexing and diversity tradeoff
in terms of the average probability of error (specifically for
polarized channels) which requires making certain assumptions
about the distribution of the channel. This comparison is useful
when either diversity or multiplexing is chosen for a succession
of channel realizations but may vary as a function of the second
order statistics of the channel. This approach allows one to
formulate selection based on average probability of error. The
work in [21][23], studies multiplexing and diversity from an
information theoretic point-of-view based on achievable rates.
In particular, the work in [21] uses an approach based on an
outage probability formulation that quantifies the amount of
diversity gain and multiplexing gain achievable for any code.
Compared to this work, our approach is more pragmatic since
we focus on the choice of either multiplexing or diversity for
a given channel as opposed to the more fundamental characterization in [21]. An advantage of our formulation is that we
obtain a constructive means for switching between diversity
and multiplexing that is useful in actual wireless systems (see,
e.g., [24], [25] for details).
Recently, code designs have appeared that provide both the
benefits of diversity and multiplexing in terms of error probability and capacity [8], [10], [26]. At first it seems that this makes
the diversity versus multiplexing tradeoff uninteresting. These
results, however, are based on ergodic capacity or average error
probability and do not adapt to the instantaneous channel realization. In fact, we can easily extend the methodology proposed
herein to compare the minimum distance of these space-time
codes, however, we omit this due to lack of space.
This letter is organized as follows. In Section II, we review
space-time signaling and motivate the minimum Euclidean distance performance metric. Next, we derive the minimum Euclidean distance for spatial multiplexing in Section III, while in
Section IV we present the minimum Euclidean distance for a
class of space-time block codes. In Section V we show how the
minimum distance provides a relevant measure for space-time
link adaptation and leads to a notion of channel characterization. In Section VI we present some simulation results while in
Section VII, we present our conclusions.
II. SYSTEM OVERVIEW
A MIMO communication link consisting of
transmit
antennas and
receive antennas is illustrated in Fig. 1. The
system contains a transmitter with a space-time modulator
that maps bits to space-time codewords, a matrix propagation
channel that is a function of the wireless environment, and a
space-time receiver that uses an estimate of the propagation
channel to decide on the transmitted bit stream.

963

Fig. 1. Generic MIMO system employing a space-time modulator and spacetime demodulator.

A space-time modulator maps a sequence of bits


to a sequence of
vector symbols
. The coded
is pulse-shaped, up-consymbol stream on the th antenna
verted, and launched into the RF propagation environment.2 The
bits/s/Hz
average rate, or spectral efficiency, of the code is
since bits are transmitted in symbol periods.
Assume that the transmission bandwidth is much less than the
coherence bandwidth of the channel and that the block length
(symbol period times ) is much less than the coherence time.
In this case we can describe the propagation channel from the
matrix whose th
transmitter to the receiver as an
column describes the complex scalar coefficients from the th
transmit antenna to each receive antenna. Neglecting symbol
received signal
timing errors and frequency offsets, the
vector after matched filtering and sampling is written
(1)
where
is an
vector realization of i.i.d. complex
circularly symmetric additive white Gaussian noise (AWGN)
per dimension,
process with distribution
is the average energy per symbol, and is the block length. For
convenience the transmitted signal vectors are assumed to be
.
normalized to have unit energy,3 i.e.,
The channel is assumed to be perfectly known at the receiver
(via training symbols) but unknown to the transmitter.
In this letter we consider maximum-likelihood (ML) codeword detection thus the metric of interest is the probability
of codeword error. Since the exact probability of codeword
error does not have a closed form solution we instead use the
Union upper bound on the probability of codeword error [12].
index the th codeword in the codebook and
Let
define the squared minimum Euclidean distance of the received
codebook as

We use the term received codebook since this is the codebook


constructed when the channel operates on each codeword. Given
, it is possible to use the union bound to upperbound
2In this letter we use boldface lowercase letters do denote vectors, boldface
uppercase to denote matrices. is the k -th element of the vector while [ ]
is the element in the k -th row and l-th column of the matrix . When
is a
matrix, k k is the matrix norm on induced by the usual Euclidean norm on
a vector and k k is the Frobenius norm.
3We use for conjugate,
for transpose, for conjugate transpose, tr() for
the trace operator that gives the sum of the diagonal elements of a matrix, and
E to denote expectation with respect to random variable s.

S
A

964

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 53, NO. 6, JUNE 2005

the conditional error probability, conditioned on a given channel


realization, as
(2)
Fig. 2.

This bound gets looser as increases. We compare exact error


rates with the union bound for MIMO diversity and spatial multiplexing in [14] and find the bound in (2) is a reasonable indicator of the relative performance. Therefore we choose the minimum Euclidean distance of the received codebook as the metric
for comparing the performance of different space-time codes as
a function of the channel.
III. SPATIAL MULTIPLEXING
In this section we derive the minimum Euclidean distance for
spatial multiplexing transmission with a maximum-likelihood
receiver. First we describe the spatial multiplexing system. Then
we compute the minimum Euclidean distance of the received
codebook as well as upper and lower bounds.
A. Spatial Multiplexing System Description
Spatial multiplexing is a space-time modulation technique in
which the core idea is to send independent streams of data from
each transmit antenna. This is motivated by the spatially white
property of the distribution which achieves capacity in MIMO
i.i.d. Rayleigh matrix channels [27] and [28]. Although spatial
multiplexing was first proposed by Paulraj and Kailath in 1994
[7], major interest is a result of the work in [6], [27], [29] by
Foschini and others at Bell Labs that shows it to be of both theoretical and practical significance. Initial spatial multiplexing
systems were narrow-band with small delay spread, however,
spatial multiplexing is being considered for wideband channels
in conjunction with OFDM modulation [20], [30].
The spatial multiplexing system we consider for the duration
of this work is illustrated in Fig. 2. During every discrete-time
complex symbols
symbol period, the encoder multiplexes
, scales by
and stacks to form a complex
vector where we have simplified the notation since
for
spatial multiplexing. To ensure that a rate of bits per codeword
is maintained, the scalar symbols are derived from a constellabits per symbol.
tion with
Let
be the set of
possible transmitted constellaand let
be
tions points with minimum distance
the set of all possible spatial multiplexing codewords. Note that
as expected since we have fixed the
bits per symbol ratio to . The receiver solves the ML detection problem

to determine the transmitted symbols. A codeword error occurs


. A simple implementation
when is decoded as some
possible transmitted codewould require searching over all
words though more efficient near-optimal methods are available
[31].

Spatial multiplexing MIMO wireless system.

B. Minimum Distance Computation


To compute the squared minimum Euclidean distance at the
receiver, consider two transmitted codewords
such that
. The squared Euclidean distance between
two possible transmitted codewords at the receiver is
. The minimum squared Euclidean distance at the receiver
is found by minimizing this difference over all possible codewords and is written
(3)
is not readily a function of
, the
Observe that
minimum Euclidean distance of the transmit constellation. This
is because the codewords undergo a linear transformation by the
channel that does not preserve their distance properties (unless
is unitary).
The computation in (3) requires at most a search over
vectors which is prohibitive for large
requires a search, to develop
constellations. Since
intuition on properties of the minimum Euclidean distance it
is desirable to find upper and lower bounds on the minimum
Euclidean distance.
First observe that

Now applying the RayleighRitz theorem [32] to (3) it follows


that

(4)
where we have assumed that
otherwise the lefthand side will be zero. Equality on the left occurs if there exthat is a scalar multiple of the right
ists an error vector
singular vector corresponding to the minimum singular value
. Equality on the right occurs when there exists an
that is a scalar multiple of the right singular
error vector
vector corresponding to the maximum singular value
.
Equality occurs in both inequalities simultaneously if and only
, i.e., if the channel is unitary.
if
The bound of interest in (4) is the lower bound. This is
, and thus the error vectors, will
because the vectors in
be distributed symmetrically in the space of all vectors since
the entries of are all derived from identical constellations.
As the rate increases, the density of the error vectors increases
and it becomes more likely that there will be an error vector
close to the minimum right singular vector of the channel.

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 53, NO. 6, JUNE 2005

965

Since we are interested in the minimum Euclidean distance, it


is the lower bound that conveys the relevant intuition for spatial
multiplexing. An illustration of the exact versus lower bound
on the minimum distance is available in ([14], p. 34).
IV. MIMO DIVERSITY

Fig. 3.

In this section we derive the minimum Euclidean distance for


MIMO diversity. First we provide a brief review of MIMO diversity modulation. Then we present an expression for the minimum Euclidean distance that is valid for a number of popular
space-time coding techniques.
A. Diversity System Description
Like spatial multiplexing, diversity modulation is another
special case of the general space-time modulator/demodulator
pair in Fig. 1. In contrast to spatial multiplexing, however,
the core idea is to send dependent streams of data from each
transmit antenna. This is motivated by the desire to reduce
the probability that a fade on any one of the transmit-receive
antenna links will cause a codeword error.
A model for the MIMO diversity system considered herein is
illustrated in Fig. 3. There are two important differences with
the system in Fig. 2. First, the symbols are mapped to a conbits per symbol instead of
bits
stellation of size of
per symbol. Second, the stream of symbols is then passed to
a space-time mapper that spreads the input symbols onto multiple transmit antennas. Fig. 3 captures (in spirit) the operation
of a number of diversity schemes including linear filtering [1],
delay diversity [1], [2], and space-time block codes [3], [5], [33].
We use the term MIMO diversity to distinguish between the
class of codes considered herein and more general linear dispersion codes (e.g., [8], [10], [26]) that combine diversity and
multiplexing.

MIMO diversity system.

the derivation of
, the minimum Euclidean distance
for our candidate set of space-time codes, follow the derivation of the maximum signal-to-noise (SNR) criterion for code
design (see (14)(16) in [33]). From the presentation in [33],
it can be verified that (5) holds with equality for the Alamouti
scheme [3]. Additionally, based on the work in [34], it is possible to show that for delay diversity with maximum-likelihood
sequence estimation, (5) is an upper bound that becomes tight
as the block length becomes large. For more than two transmit
antennas, however, the bound is not achievable [1] but the difference becomes negligible for large numbers of antennas [34].
When full-rate space-time block codes based on orthogonal de, (5) provides an uppersigns [5] do not exist for the choice of
bound on the minimum Euclidean distance (cf. maximum SNR
criterion in [33]).
To obtain some insight into the benefits of diversity, in the
deterministic sense, we expand the squared Frobenius norm

(6)

or alternatively use the property that the squared Frobenius norm


of a matrix equals the sum of its squared singular values [35]

(7)

B. Computation of the Minimum Euclidean Distance


Computation of the minimum Euclidean distance for an arbitrary space-time code is difficult since it inherently depends
on the specification of the code. There are, however, a class of
codes that have a simple expression for the minimum Euclidean
distance. Conveniently, these are also among the most practical
codes for transmit diversity.
In this letter, we consider the class of linear full-rate spacetime block codes [3], [5], [33]. These codes are full-rate in the
sense that they transmit an average of one symbol per symbol
period. Using the representation described in [33], it is possible
to express the codewords as linear combinations of certain basis
matrices, with the weights determined by the transmit symbols.
The basis matrices are designed to obtain maximum diversity
advantage and are independent of a given channel realization.
For the class of full-rate linear space-time block codes, it is possible to show that
(5)
is defined as the minimum Euclidean diswhere
tance for our candidate set of space-time codes. The details for

where
and is the th singular value of .
Observations: From (5) the channel influences the minimum Euclidean distance through the Frobenius norm. Therefore good channels are those that convey a large total power
to the receiver [cf. (6)]. The diversity advantage is illustrated
through the summation of the powers of each path. This sum
implies that a large minimum distance is possible even if there
is only a single nonnegligible path between one of the
transmit antennas and
receive antennas. In spite of the
fact that the channel is a matrix channel, its rank has no immediate effect on the performance except through the sum of
singular values in (7). Spatial multiplexing, on the other hand,
depends significantly on the rank of the channel since performance is strongly determined by the minimum singular value
of the channel in (4). If the rate is fixed, however, both schemes
will use constellations with different minimum distances. Thus
while spatial multiplexing is more sensitive to the rank of the
channel it also uses a constellation with a larger minimum distance. The implication of this is that the choice of multiplexing
or diversity is dependent on the matrix channel, the constellation, and the desired rate.

966

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 53, NO. 6, JUNE 2005

V. EUCLIDEAN DISTANCE-BASED
CHANNEL CHARACTERIZATION
The observation that multiplexing and diversity perform
differently for different channels motivates characterizing
the quality of a MIMO channel based on whether it is more
capable of supporting multiplexing or MIMO diversity. In this
section we illustrate one such characterization motivated by
our Euclidean distance calculations in Sections III and IV.
Channel characterization is useful for qualitative analysis of
measurement data, e.g., [18], [19], [36]. Other examples of
channel characterizations include the Frobenius norm of the
channel and the capacity of the channel at a given SNR.
Our proposed characterization is based on whether a given
channel is more suitable for multiplexing or MIMO diversity.
We could simply compute (3) and (5); however, the dependence
on the error vectors in (3) is somewhat inconvenient. Therefore
let us instead determine when the lower bound on the minimum
distance for spatial multiplexing in (4) exceeds the upper bound
on the minimum distance for MIMO diversity transmission in
(5). In this case, the worst minimum distance of spatial multiplexing is better than the best minimum distance for diversity
transmission. This comparison will allow us to identify channels
for which spatial multiplexing will always out-perform diversity
transmission in terms of the union bound (2) on the codeword
error probability. Essentially we are finding a sufficient condition that multiplexing will be better than diversity for a given
matrix channel. Of course, since bounds on the minimum distance are being compared, this condition is not necessary.
Comparing (5) and the lower bound in (4), we find that multiplexing is preferred when
(8)
Define
. Rearranging (8) and taking the
square root of both sides we find that multiplexing is preferred
if
(9)
is known as the Demmel condition number [17] and for
is defined as
. Like the regular
a matrix
, the Demmel condition
condition number
number measures how ill-posed, that is, how invertible, is a
given matrix. Both the Demmel condition number and the
relative condition number are infinite when is singular. The
utility of (9) is that it is possible to characterize the suitability
for multiplexing based on only
of a given matrix channel
. As the
a single numberthe Demmel condition number
threshold on the right-hand side of (9) is increased, more
poorly conditioned channels become suitable for multiplexing
transmission.
An advantage of the Demmel condition number is that it
is scale-invariant; thus it is possible to decouple the gain
or loss of the channel with the span of its row and column
spaces. In particular, we can without loss of generality nor(as is common in many measurement
malize
papers, e.g., [36]) and then multiplexing is preferred when
. From this point of view,

Fig. 4. MIMO system where the transmitter switches between spatial


multiplexing and MIMO diversity based on feedback from the receiver.

(which is less than one) reduces the effective minimum distance of the spatial multiplexing constellation. Multiplexing is
preferred as long its scaled minimum distance is still greater
than the minimum distance of the diversity constellation.
As a tool for analyzing channel measurements, the Demmel
condition number fits in nicely with other forms of channel characterization. For example, histograms of the Frobenius norm are
often computed to study the empirical distribution of effective
SNR (gain or loss) in the channel. Histograms of the Demmel
condition number would complement these studies by giving an
empirical estimate of the probability that multiplexing is preferred for a given channel distribution as well as a quantitative
measure of channel invertibility. Both would be useful for understanding the severity of real-world propagation effects such
as correlation or line-of-sight terms. Due to space constraints we
defer further discussion and applications to real channel data to
future work.
VI. ADAPTIVE SPACE-TIME MODULATION
The characterization in Section V confirms our intuition that
channels may be suitable for either multiplexing or diversity depending on their salient characteristics. In this section, we exploit this fact by proposing a system that adaptively switches between multiplexing and diversity. We consider fixed rate adaptation where the total rate for each approach is fixed thus only
the spatial modulation scheme and not the rate is varied.
The proposed system, illustrated in Fig. 4, consists of a
transmitter with a switch between a diversity and a multiplexing modulator, a receiver unit with the corresponding pair
of receivers, and a low-bandwidth feedback path. When the
channel is slowly varying, the receiver computes the optimum
signaling technique and sends this information (a single bit)
to the transmitter. The transmitter then employs the corresponding spatial signaling method for the next channel use.
Low-bandwidth feedback channels, for example, are present
in most cellular systems for the purpose of power control. At
present, perfect channel knowledge is assumed at the receiver
and zero feedback delay. Delay in the feedback path or channel
estimation errors will create decision errors and thus will
degrade switching performance. Study of these degradations is
left to future work.
Optimum performance for a fixed rate system is obtained by
choosing the signaling scheme with the lowest bit error rate.
In Rayleigh fading channels, optimum switching guarantees
full diversity advantage, measured in terms of the ratio of the
asymptotic slope of the bit error rate curves with and without

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 53, NO. 6, JUNE 2005

diversity. The reason is the following. In Rayleigh fading channels, transmit diversity schemes by construction guarantee full
diversity order. Optimum switching always picks the scheme
with the lowest bit error rate. Consequently it can only do better
than transmit diversity. Thus even though the diversity advan, with optimum switching we
tage of spatial multiplexing is
by including the diare guaranteed to get a diversity of
versity mode of operation. More importantly, this holds even
for spatial multiplexing with suboptimum receivers, however,
we defer this to other work [16].
Since optimizing the bit error probability is difficult due to
the bit-to-symbol mapping, we resort to formulating a selection
probability based on the minimum Euclidean distance since it
plays a major role in the codeword error probability for a given
space-time code.
Selection Criterion: Minimum Euclidean Distance:
from (5) and
from (3).
Compute
Choose the scheme with the largest minimum Euclidean
distance.
Based on previous discussions about using the minimum
Euclidean distance to predict error rate, we still expect to
observe full diversity advantage at high SNRs and we observe
this in the simulation example. Other suboptimal selection
criteria are possible with reduced computational requirements,
for example based on the Demmel condition number, however
these approaches may incur additional diversity loss. Study of
other selection criteria is a topic for future research.
While we explicitly deal with fixed-rate adaptation, it is clear
that the proposed approach can be combined with adaptive modulation (see, e.g., discussion in [37]). A rather transparent way
of doing this would be to compute the largest rate that can be
supported with diversity or multiplexing, given a target packet
error rate, and then to send back the desired space-time modulation scheme and the requested rate. Thus the selection of
a space-time modulation scheme would form an additional bit
of feedback in the adaptive modulation. The diversity gain improvement would be transformed into higher average throughputs for each user.
We have focused on switching from the point-of-view of improving the quality of the communication link. The proposed
switching architecture especially with adaptive modulation extensions, however, can also provide benefits that are realized
at the system level in cellular systems. To see this, realize that
transmit diversity primarily improves resistance to fading and
thus is of the most interest to users at the edge of the cell who
are most susceptible to fades. Allowing a diversity mode allows
the range of the cell to be improved. On the other hand, users
that are closer to the base station have larger SNRs that can
be used to achieve higher rates. Combined with spatial multiplexing, these users can realize even higher throughputs. Consequently, switching between diversity and multiplexing may
provide a mechanism for making range and rate tradeoffs in cellular systems though switching criteria that also depend on the
interference level need to be developed.
VII. SIMULATION EXAMPLE OF SWITCHING
To illustrate potential performance improvements of the
system in Fig. 4, we provide an example of fixed-rate spa-

967

Bit error rate versus SNR for a M = M = 2 MIMO system with


R = 4 bits/symbol and optimal selection of spatial multiplexing or MIMO
Fig. 5.

diversity based on minimum Euclidean distance.

tial-adaptation that employs minimum Euclidean distance


based switching. Let
, and
bits per
symbol. The encoder switches between the Alamouti MIMO
diversity scheme [3] with 16-QAM modulation and spatial
multiplexing with 4-QAM transmit constellations. We assume
that the channel is i.i.d. complex Gaussian and is constant over
a block of 10 symbols and is drawn independently in each
block. Perfect channel knowledge/prediction and zero-delay
feedback are assumed.
We estimate the probability of bit error, averaged over all
channel realizations, using Monte Carlo simulations by evaluating the bit error performance for
channel realizations.
The results are illustrated in Fig. 5. The first curve is that
obtained through the use of spatial multiplexing. It performs
slightly better at low SNR but slightly worst than MIMO diversity at higher SNR. This is due to the fact that spatial
multiplexing has a diversity advantage on the order of the
number of receive antennas [38], in this case second order,
while MIMO diversity has a diversity advantage on the order
of the product of the number of transmit and receive antennas,
in this case fourth order [3]. Diversity advantage changes the
slope of the average probability for Rayleigh fading channels
at high SNR.
The MIMO diversity and spatial multiplexing curves cross at
approximately
dB. If one were to choose one modulation scheme for all channel realizations then spatial multiplexing is preferred for
dB while MIMO diversity
is preferred for
dB. On the other hand, if it is possible to adapt based on the channel state then the best of spatial
multiplexing or MIMO diversity is chosen for each channel realization. The results in a gain of about 1 dB at higher SNRs and
a marginal gain at lower SNRs. The reason for the lack of gain
at lower SNRs is that the union bound is a good approximation
of the codeword error probability only at higher SNRs. It does
not accurately predict performance at lower SNRs thus another
switching criterion should be derived that is specific to low SNR
operation.

968

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 53, NO. 6, JUNE 2005

VIII. CONCLUSION
In this letter we proposed and analyzed a MIMO communication system that switches between spatial multiplexing and
transmit diversity based on instantaneous channel state information. We based the selection between spatial mapping schemes
on the minimum Euclidean distance of the received codebook
because this measure reveals dependencies on the channel realization and provides an approximate measure of error rate performance. As a byproduct of our switching criterion, we found
that the Demmel condition number of the matrix channel can be
used to characterize the ability of a channel to support spatial
multiplexing using a constellation dependent threshold. Specifically, the Demmel condition number provides us with a sufficient condition to test if multiplexing will outperform diversity
for a given choice of constellation rate.
The switching architecture presented in this paper was described specifically for narrow-band MIMO channels. It is clear,
however, that this approach can be applied to broad-band channels using MIMO-OFDM. In such a realization, a choice of multiplexing or diversity would be made for clusters of subcarriers
corresponding to the coherence bandwidth of the channel, and
the selection results would be sent back for each cluster. While
we did not consider such generalizations it is also clear that the
proposed methodology can be applied to systems with more sophisticated space-time coding schemes that provide more granularity between spatial multiplexing and transmit diversity. One
example is to employ antenna subset selection with different
sized subsets; preliminary results are described elsewhere [16].
REFERENCES
[1] A. Wittneben, Base station modulation diversity for digital simulcast,
in Proc. IEEE Vehicular Technology Conf., May 1991, pp. 848853.
[2] N. Seshadri and J. H. Winters, Two signaling schemes for improving
the error performance of frequency-division-duplex (FDD) transmission
systems using transmitter antenna diversity, Int. J. Wireless Inform.
Netw., vol. 1, no. 1, pp. 4960, Jan. 1994.
[3] S. M. Alamouti, A simple transmit diversity technique for wireless
communications, IEEE J. Sel. Areas Commun., vol. 16, no. 8, pp.
14511458, Oct. 1998.
[4] V. Tarokh, N. Seshadri, and A. R. Calderbank, Space-time codes for
high data rate wireless communication: Performance criterion and code
construction, IEEE Trans. Inf. Theory, vol. 44, no. 2, pp. 744765, Mar.
1998.
[5] V. Tarokh, H. Jafarkhani, and A. R. Calderbank, Space-time block
codes from orthogonal designs, IEEE Trans. Inf. Theory, vol. 45, no.
5, pp. 14561467, Jul. 1999.
[6] G. J. Foschini, Layered space-time architecture for wireless communication in a fading environment when using multiple antennas, Bell
Labs. Tech. J., vol. 1, no. 2, pp. 4159, 1996.
[7] A. Paulraj and T. Kailath, Increasing capacity in wireless broadcast systems using distributed transmission/directional reception (DTDR), U.
S. Patent 5,345,599, Sep. 6, 1994.
[8] B. Hassibi and B. M. Hochwald, High-rate codes that are linear in space
and time, IEEE Trans. Inf. Theory, vol. 48, no. 7, pp. 18041824, Jul.
2002.
[9] S. Sandhu and A. Paulraj, Space-time block coding: A capacity perspective, IEEE Commun. Lett., vol. 4, no. 12, pp. 384386, Dec. 2000.
[10] R. W. Heath Jr. and A. Paulraj, Linear dispersion codes for MIMO systems based on frame theory, IEEE Trans. Signal Process., vol. 50, no.
10, pp. 24292441, Oct. 2002.
[11] J. G. Proakis, Digital Communications. New York: McGraw Hill,
1995.
[12] J. Cioffi. Digital transmission: Volume i. Stanford University, EE 379A.
[Online]. Available: http://www.stanford.edu/class/ee379a/
[13] S. Verdu, Multiuser Detection. Cambridge, U.K.: Cambridge Univ.
Press, 1998.

[14] R. W. Heath Jr., Space-time signaling in multiantenna systems, Ph.D.


thesis, Stanford Univ., Stanford, CA, Nov. 2001.
[15] R. W. Heath Jr. and A. Paulraj, Switching between spatial multiplexing
and transmit diversity based on constellation distance, in Proc. Allerton
Conf. Communication, Control, and Computers, Oct. 2000.
[16] R. W. Heath Jr. and D. J. Love, Multi-mode antenna selection for
spatial multiplexing systems with linear receivers, in Proc. Allerton
Conf. Communication, Control, and Computers, Monticello, IL, Oct.
2003.
[17] J. W. Demmel, The probability that a numerical analysis problem is
difficult, Math. Comput., pp. 449480, 1988.
[18] V. Erceg, P. Soma, D. Baum, and A. J. Paulraj, Capacity obtained from
multiple-input multiple-output channel measurements in fixed wireless
environments at 2.5 GHz, in Proc. Int. Conf. Communications, vol. 1,
May 2002, pp. 396400.
[19] M. D. Batariere et al., Wideband MIMO mobile impulse response measurements at 3.7 GHz, in Proc. IEEE Vehicular Technology Conf., vol.
1, 2002, pp. 2630.
[20] R. U. Nabar, H. Bolcskei, V. Erceg, D. Gesbert, and A. J. Paulraj, Performance of multiantenna signaling techniques in the presence of polarization diversity, IEEE Trans. Signal Process., vol. 50, no. 10, pp.
25532562, Oct. 2002.
[21] L. Zheng and D. Tse, Diversity and multiplexing: A fundamental
tradeoff in multiple antenna channels, IEEE Trans. Inf. Theory, vol.
49, no. 5, pp. 10731096, May 2003.
[22] B. Varadarajan and J. R. Barry, The rate-diversity trade-off for linear
space-time codes, in Proc. IEEE Vehicular Technology Conf., vol. 1,
2002, pp. 6771.
[23] M. Godavarti and A. O. Hero, Diversity and degrees of freedom in
wireless communications, in Proc. ICASSP, vol. 3, May 2002, pp.
28612854.
[24] H. Sampath, S. Talwar, J. Tellado, V. Erceg, and A. Paulraj, A fourthgeneration MIMO-OFDM broad-band wireless system: Design, performance, and field trial results, IEEE Commun. Mag., vol. 40, no. 9, pp.
143149, Sep. 2002.
[25] R. W. Heath Jr., S. K. Peroor, and A. J. Paulraj, Methods of controlling
communication parameters of wireless systems, U.S. 6,298,092, Oct.
2, 2001.
[26] H. El Gamal and M. O. Damen, An algebraic number theoretic framework for space-time coding, in Proc. Int. Symp. Information Theory,
2002, p. 132.
[27] G. J. Foschini and M. J. Gans, On limits of wireless communications
in a fading environment when using multiple antennas, Wireless Pers.
Commun., vol. 6, no. 3, pp. 311335, Mar. 1998.
[28] I. E. Telatar, Capacity of multiantenna Gaussian channels, AT&T Bell
Labs., Internal Technical Memorandum 1995, published in the Eur.
Trans. Telecommun., Nov./Dec. 1999. .
[29] G. D. Golden, G. J. Foschini, R. A. Valenzuela, and P. W. Wolniansky,
Detection algorithm and initial laboratory results using the V-BLAST
space-time communication architecture, Electron. Lett., vol. 35, no. 1,
pp. 1415, 1999.
[30] G. G. Raleigh and J. M. Cioffi, Spatio-temporal coding for wireless
communication, IEEE Trans. Commun., vol. 46, no. 3, pp. 357366,
Mar. 1998.
[31] O. Damen, A. Chkeif, and J.-C. Belfiore, Lattice code decoder for
space-time codes, IEEE Commun. Lett., vol. 4, no. 5, pp. 161163,
May 2000.
[32] R. A. Horn and C. R. Johnson, Matrix Analysis. Cambridge, U.K.:
Cambridge Univ. Press, 1985.
[33] G. Ganesan and P. Stoica, Space-time block codes: A maximum SNR
approach, IEEE Trans. Inf. Theory, vol. 47, no. 4, pp. 16501656, May
2001.
[34] J. H. Winters, The diversity gain of transmit diversity in wireless systems with Rayleigh fading, IEEE Trans. Veh. Technol., vol. 47, no. 1,
pp. 119123, Feb. 1998.
[35] H. Ltkepohl, Handbook of Matrices. New York: Wiley, 1996.
[36] M. J. Gans, N. Amitay, Y. S. Yeh, H. Xu, T. C. Damen, R. A. Valenzuela, T. Sizer, R. Storz, D. Taylor, W. M. MacDonald, C. Tran, and A.
Adamiecki, Outdoor BLAST measurement system at 2.44 GHz: Calibration and initial results, IEEE J. Sel. Areas Commun., vol. 20, no. 3,
pp. 570583, Apr. 2002.
[37] S. Catreux, V. Erceg, D. Gesbert, and R. W. Heath Jr., Adaptive modulation and MIMO coding for broad-band wireless data networks, IEEE
Commun. Mag., vol. 40, no. 6, pp. 108115, Jun. 2002.
[38] B. A. Bjerke and J. G. Proakis, Multiple-antenna diversity techniques
for transmission over fading channels, in Proc. Wireless Communications and Networking Conf., vol. 3, Sep. 1999, pp. 10381042.

S-ar putea să vă placă și