Cutoff Rate and Signal Design For The Quasi-Static Rayleigh-Fading Space-Time Channel

2400
IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 47, NO. 6, SEPTEMBER 2001
Cutoff Rate and Signal Design for the Quasi-Static

Rayleigh-Fading SpaceTime Channel
Alfred O. Hero, III, Fellow, IEEE, and Thomas L. Marzetta, Senior Member, IEEE
AbstractWe consider the computational cutoff rate and its

implications on signal design for the complex quasi-static Rayleigh
flat-fading spatio-temporal channel under a peak-power constraint where neither transmitter nor receiver know the channel
matrix. The cutoff rate has an integral representation which is
an increasing function of the distance between pairs of complex
signal matrices. When the analysis is restricted to finite-dimensional sets of signals, interesting characterizations of the optimal
rate-achieving signal constellation can be obtained. For an arbitrary finite dimension, the rate-optimal constellation must admit
an equalizer distribution, i.e., a positive set of signal probabilities
which equalizes the average distance between signal matrices in
of receive antennas is
the constellation. When the number
large, the distance-optimal constellation is nearly rate-optimal.
When the number of matrices in the constellation is less than the
ratio of the number of time samples to the number of transmit
antennas, the rate-optimal cutoff rate attaining constellation is a
set of equiprobable mutually orthogonal unitary matrices. When
the signal-to-noise ratio (SNR) is below a specified threshold, the
matrices in the constellation are rank one and the cutoff rate is
achieved by applying all transmit power to a single antenna and
using orthogonal signaling. Finally, we derive recursive necessary
conditions and sufficient conditions for a constellation to lie in the
feasible set.
Index TermsFlat Rayleigh fading, minimal-distance codes, optimal signal design, peak-power constraints, spacetime coding,
spatial diversity, unitary constellations.
I. INTRODUCTION
N this paper, we investigate the single-user cutoff rate for

a Rayleigh-fading spatio-temporal channel model under
a maximum peak transmitted power constraint. For a wider
perspective on different channel fading models we refer the
reader to [4]. Multichannel flat Rayleigh-fading models have
been widely used to investigate spacetime channel capacity
[30], [8], [14], [20], [31], [12], to investigate random coding
error exponents [1], [31], and to motivate and evaluate robust
Manuscript received March 30, 2000; revised December 13, 2000. The work
of A. O. Hero, III, was supported in part by the Mathematical Sciences Research
Center during his sabbatical leave there in spring 1999 and subsequently by the
Army Research Office under Grant ARO DAAH04-96-1-0337. The material
in this paper was presented in part at the IEEE Workshop on Sensors, Arrays
and Multichannel Signal Processing (SAM), Cambridge MA, March 2000, and
at the IEEE International Symposium on Information Theory (ISIT), Sorrento,
Italy, June 2000.
A. O. Hero, III, is with the Department of Electrical Engineering and Computer Science, The University of Michigan, Ann Arbor, MI 48109-2122 USA
(e-mail: hero@eecs.umich.edu).
T. L. Marzetta is with the Mathematical Sciences Research Center, Bell Laboratories, Lucent Technologies, Murray Hill, NJ 07974 USA (e-mail: tlm@research.bell-labs.com).
Communicated by M. L. Honig, Associate Editor for Communications.
Publisher Item Identifier S 0018-9448(01)06216-2.
spacetime coding schemes [15], [16], [29], [7]. Similarly to

Marzetta and Hochwald [20], our results are derived under the
quasi-static Rayleigh- fading model and under the assumption
that neither transmitter nor receiver knows the channel. The
quasi-static fading model assumes a propagation matrix of
zero-mean complex Gaussian-distributed coefficients which
remain constant over several time samples, called the coherent
fade sampling interval, but change independently over successive fade intervals. The signal set for this channel consists
of complex matrices whose rows are indexed over temporal
coordinates that span the fade interval, and whose columns are
indexed over spatial coordinates that span the locations of the
transmitter antennas.
is a lower bound on the Shannon
The channel cutoff rate
also specifies an upper bound on the
channel capacity .
error rate of an optimal decoder operating at a given symbol
rate which can be used to majorize the minimum probability of
decoding error. Cutoff rate analysis has frequently been adopted
to establish practical coding limits [32], [11] as the cutoff rate
specifies the highest information rate beyond which sequential
decoding becomes impractical [27], [34] and as it is frequently
simpler to calculate than channel capacity. Cutoff rate analysis
has also been used to evaluate relative merits between different
coding and modulation schemes [22], signal design for optical
communications [28], and establishing achievable rate regions
for multiple-access channels [24], [25].
The following are some of the principal results obtained.
1) An integral representation for the cutoff rate is obtained
(Lemma 1) which depends on a pairwise dissimilarity
measure over the set of signal matrices. This dissimilarity
measure is a decreasing function of the spatial correlation
between pairs of signal matrices.
2) For low signal-to-noise ratio (SNR), the dissimilarity
measure reduces to a distance metric equal to the trace
norm of pairwise differences between outerproducts of
the signal matrices (Lemma 7).
3) The -dimensional cutoff rate, defined as the cutoff rate
for constellations whose dimension does not exceed ,
is the appropriate limiting factor for finite-dimensional
spacetime coding/decoding and when signal sets must
be generated with finite numerical precision. A necessary
condition for such a signal constellation to attain cutoff
rate is that there exist a probability distribution on the constellation which equalizes the average exponentiated distance between signal matrices in the constellation (Proposition 3). We call this the equalization condition and it
plays a central role in this work.
00189448/01$10.00 2001 IEEE
HERO AND MARZETTA: CUTOFF RATE AND SIGNAL DESIGN FOR THE QUASI-STATIC RAYLEIGH-FADING SPACE-TIME-CHANNEL
4) The determination of the

-dimensional cutoff rate
reduces to maximization of a quadratic form over the set
of feasible constellations, defined as those constellations
which satisfy both the peak-power constraint and the
finite-dimensional equalization condition (Proposition
4). This quadratic form is similar to that arising in the
Capon/MVDR method for adaptive beamforming arrays.
If the feasible set of -dimensional constellations is
empty then the rate-optimal constellation is necessarily
of dimension less than .
5) For low symbol-rate, the rate-optimal constellation is a set
of scaled mutually orthogonal unitary matrices in
(Proposition 9). This constellation is also distance-optimal in the sense of having the largest possible minimal distance over all constellations of the same dimension (Corollary 2). When SNR is low, the rank of the
signal matrices in the constellation is one and cutoff rate is
achieved by transmitting all power from a single antenna
element. As the SNR increases, the rank of the signal
matrices increases and more and more antenna elements
are utilized. Interestingly, the number of receive antennas
plays no role whatsoever in determining how many
transmit antennas should be used.
exceeds
but is finite, the -dimensional
6) When
cutoff rate is not closed-form and must be determined
by numerical optimization of the cutoff rate objective
function over the feasible set of constellations. We derive simple necessary conditions and simple sufficient
conditions for a constellation to be in the feasible set
(Lemma 3). The necessary conditions characterize properties of the optimal cutoff rate achieving constellation
while the sufficient conditions can be used to generate
feasible signal sets. To satisfy these conditions, among
other attributes, the signal matrices should have pairwise
distances of low variability. As these conditions can be
used to check if a candidate set of constellations is in the
feasible set or not, they provide a potentially useful signal
design tool, especially when dealing with large constellations of high-dimensional signal matrices for which computing the globally rate-optimal signal set may be impractical.
of receive antennas increases the
7) When the number
distance-optimal constellation becomes nearly rate-optimal and the -dimensional cutoff rate becomes approx. Furthermore, the dimension of
imately equal to
the rate-optimal constellation among all constellations of
increases (Proposicountable dimension increases as
tion 6).
The outline of this paper is as follows. In Section II, the
Rayleigh- fading measurement model is introduced, we briefly
review the computational cutoff rate. and we characterize general structural properties of spacetime codes that achieve cutoff
rate. In Section III, an integral expression is specified for cutoff
rate involving the pairwise dissimilarity function. In Section IV,
we specialize to the finite-dimensional cutoff rate which reduces
to maximization of a quadratic form over signal dissimilarity
matrices. In Section V, we consider quadratic programming for
2401
computing the cutoff rate. In Section VI, general characterizations for the feasible set are given. In Section VII, we specialize
to the case of finite constellations with dimension smaller than
, for which the cutoff rate is closed-form.
II. PRELIMINARIES
A. Rayleigh-Fading Spatio-Temporal Channel
We use substantially the same notation as in the papers of
transmitter
Hochwald and Marzetta [20], [14]. Let there be
receiver antennas and let the
channel
antennas and
fading coefficients be constant over an interval of length
time periods, called the coherent fade sampling interval. A
matrix having complex valued
transmitted signal is a
denote the set of all possible signal
entries. Let
is defined as
matrices. The norm of a matrix
where
denotes the Hermitian transpose of and
are the singular values of . Note that if
only
of
these singular values will be nonzero.
For receiver antennas and an observation time interval of
time periods, the received signal is a sequence
of
complex-valued
matrices
which
has the representation [20]
(1)
is the th transmitted signal,
is the norwhere
the expected SNR at each receiver per
malized SNR with
is an
matrix of complex channel
transmit antenna,
is a
matrix of complex noises. The
coefficients, and
quasi-static Rayleigh flat fading model corresponds to taking
elements of the matrices
and
the
to be independent and identically distributed (i.i.d.) complex
zero-mean (circularly symmetric) Gaussian random variables
with unit variance. Therefore, the joint conditional probability
density of the observations factors into a product of marginals
where
(2)
is the
identity matrix, and
denotes
the magnitude determinant of square matrix .
denote the probability of decoding error for a block
Let
code of rate (nats) with block length . It is well known [34],
[10], [9] that for below capacity , the minimum decoding
of the best code satisfies
error probability
(3)
2402
where
is a reliability function, called the random
coding error exponent, controlling the error rate of the optimal
decoder
nats/symbol
(4)
where the inner maximization is performed over a suitably constrained set of probability distributions defined over the set
of signal matrices . Additional constraints on are determined
by factors such as total power budget or envelope constraints
which atand are left implicit in (4). The distribution
gives an optimal signal distributains the maximum in
tion for which there exists a decoder achieving minimum probability of decoding error for sufficiently large block lengths. Genis not discrete and an optimal set of signal matrices
erally,
can only be constructed by a random coding procedure. The
was studied for the spatio-temporal Rayleigh
function
quasi-static fading model by AbouFaycal and Hochwald [1]
and Telatar [31] under a mean power constraint on transmitted
signal matrices. Unfortunately, the double maximization in (4)
is generally very difficult since the inner integral is raised to a
. The cutoff rate specified
fractional exponent when
which is
below is a lower bound on the error exponent
frequently simpler to analyze.
can be lowerThe random coding error exponent
having slope
:
bounded by the tangent line
, where
is this lines intercept on both the
and the axes [34]. The point of tangency
is the critical rate and the intercept

is called the computational cutoff rate. The cutoff rate is given by
nats/symbol
(5)
in (4) with the

which is equivalent to the error exponent
replaced by the simpler evaluation
maximization over
.
at the point
While use of cutoff rate for evaluating system performance
has its limitations, in particular it is an upper bound on rate
only for sequential decoding [5], [21], it satisfies useful properties which in our opinion justify its application to spacetime
coding:
, the cutoff rate provides a lower bound on
Since
channel capacity.
when the symbol rate is close to .
which attains the

Thus, the signal distribution
maximum in the definition of

will attain optimal error
rates for near .
is the maximum
For sequential decoding strategies,
practical symbol transmission rate. Specifically, even
though Shannons coding theorems ensure that there
exists a low error decoding algorithm for all such that
, if
the probability of arbitrary long
decision delays becomes significant [27]. Even though it
has recently been demonstrated that nonsequential turbo
[3], sequential
decoders can achieve rates greater than
decoding remains of interest and cutoff rate analysis
continues to give useful insights [4], [33], [18], [32], [11].
Combining (3)(5), a useful upper bound is obtained on
the minimum probability of decoder error for transmission
rates less than
(6)
is an upper bound on the error rate
Thus,
of the optimal decoder regardless of the decoding algorithm.
is difficult or in When analysis of channel capacity
offers an alternative which may be easier to
tractable,
analyze.
B. Structure of SpaceTime Codes Attaining Cutoff Rate
Below we give a result that parallels Theorems 1 and 2 of
Marzetta and Hochwald for channel capacity [20], but covers
the case of peak-power constrained signal sets.
Proposition 1: Assume that the transmitted signal is con.
strained to satisfy the peak-power constraint
The peak-power constrained cutoff rate attained with
transmit antennas is the same as that attained with
antennas. Therefore, there is no advantage to using more than
transmit antennas. Furthermore, for
the signal matrix which achieves peak-power constrained cutoff rate can be
where
is a
unitary matrix,
expressed as
is a
matrix, and
is a diagonal
matrix.
.
Proof of Proposition 1: We first treat the case that
Let have singular value decomposition (SVD)
where is a
unitary, is an
unitary, and
is a
matrix partitioned into a diagonal
matrix
and a
matrix of zeros. The fading model
specifies the received signal as the
matrix
(1) for
where the
matrix and the
matrix
are mutually independent matrices of i.i.d. zero-mean complex Gaussian
random variables. Substituting the SVD of into the model we
obtain
2403
Note that, as
is unitary,
is a
matrix of i.i.d. zero-mean complex Gaussian random variables
, the
trans(r.v.s). Note also that, as
satisfies the same peak-power constraint as
formed signal
signal . Thus, the signal
measured at
the original
antennas
the receiver after transmission of the signal on
is
is statistically equivalent to received when the signal
transmitted on only antennas.
, the SVD of is identical to
When
above except that now
, where
is an
diagonal matrix. Therefore,
The inner integral has the explicit form
where
where in the first line, the normalization constant of the

multivariate complex Gaussian p.d.f. has been identified
) and in the second
(
line, the determinant property
has been used.
Readers familiar with [20, Theorems 1 and 2] might suspect that characterization of the statistical distribution of the
optimal cutoff achieving signal matrix can be obtained. Indeed, paralleling the arguments of [20], it can be shown that, as
is invariant to unitary premultiplication of
and as the maximization in the definition of
is over a concave function of , the peak-power constrained cutoff rate is
where is
attained by random matrices of the form
isotropically distributed matrix,
is a
a
diagonal matrix, and and are statistically
random
independent.
(9)
The form of the cutoff rate given in (7) suggests defining

the pairwise dissimilarity measure for any two signal matrices
(10)
Thus, the cutoff rate can be equivalently expressed as
III.
REPRESENTATION FOR QUASI-STATIC RAYLEIGH FLAT

FADING CHANNEL
nats/symbol
Here we give integral representations for the cutoff rate.

Lemma 1: The cutoff rate for the spatio-temporal fading
model (1) is given by
nats/symbol
(7)
where the maximization is performed over the set of distributions of suitably constrained signal matrices.
Proof of Lemma 1: For notational convenience define the
matrix
. Then the conditional probability density function (p.d.f.) (2) takes the form
The integral in (5) is the triple integral
As
is nonnegative definite
and, therefore,
. Thus, by Fubini we can interchange
order of integration in the triple integral to obtain
(11)
As the integral (9) is upper-bounded by one (Cauchy

is nonnegative. By Lemma 7
Schwarz inequality),
(given in the Appendix)
so that in the asymptotic regime of low SNR ( small),

is to a good approximation proportional to the
of the matrix outerproduct
squared norm
difference. We define the minimal distance (dissimilarity) of a
as
set of signal matrices
A set of signal matrices which is close to optimal, in terms of

nearly attaining the cutoff rate, might be expected to have large
. Indeed, Proposition 9 implies that the rate-opvalue of
for
timal peak constrained signal constellation maximizes
low symbol-rates.
In the following sections, we specialize to the case of discrete signal constellations for which equalizer distributions are
always optimal.
IV. FINITE ( )-DIMENSIONAL CUTOFF RATE
(8)
For practical coding schemes, it is of interest to restrict attention to finite sets of signal matrices. Let be a prespecified
finite positive integer. The cutoff rate (11) restricted to discrete
2404
distributions concentrated on at most signal matrices will be

called the -dimensional cutoff rate and takes the form
On the other hand, when a peak-power constraint is imposed,

the optimization (12) is performed over the simpler product set
(12)
where
(14)
is the quadratic form

where
(13)
(15)
is a suitably constrained set of signal matrices
and where
and signal probabilities
(see the following
subsection for examples).
the signal is never transmitted which motiWhen
vates a natural definition of a signal constellation.
in
Definition 1: A set of matrices
constellation if all assigned signal probabilities
.
positive,
and
(16)
In particular, under the constraint (14), the optimization (12)
decomposes into two nested minimizations
is a signal
are strictly
As the optimization in (12) is performed over a restricted set

is a lower bound on
. Hence, in particular,
can be used in (6) to specify an upper bound on probability of
decoding error or a lower bound on capacity. If the optimizing
distribution in (11) over actually concentrates on a constellation consisting of a discrete finite number of signal mais a discrete probability
trices then the bound is tight, i.e.,
for some finite . This hypothesis may
and
, the optimal
not be unreasonable, e.g., when
was shown to be discrete by Abou-Faycal and Hochwald [1] for
codes that achieve the random coding error exponent. However,
and , this
for cutoff rate and for more general values of
property remains to be verified. On the other hand, when the
signal matrices are generated under finite precision arithmetic
is always tight for an appropriately chosen
the bound
, e.g.,
for
bit register resolution on each
complex entry of the signal matrix.
(17)
,
where, for fixed
is the optimal probability assignment to the signal

(18)
B. Signal Dissimilarity Matrix

(12) is obtained by putting
A more compact form for
the quadratic form into vector form with a dissimilarity matrix
. For signal matrices
, and for
is the
matrix
A. Peak Power Versus Average Power Constraints

When an average transmitted power constraint is imposed,
such as adopted in [20], the optimization (12) must be performed
over the restricted set
..
.
Note that
where
is the average transmitter power budget. Thus, the
average power constraint introduces additional dependency between signal matrices and signal probabilities which complicates the optimization problem.
..
..
..
..
..
.
(19)
is symmetric with nonnegative entries.
is nonnegative definite. If the outerproduct

Lemma 2:
are distinct then: i)
is positive definite;
matrices
,
.
and ii)
Proof of Lemma 2: From the defining relations (10) and
representation (9) in the proof of Lemma 1 we have
For assertion i) consider for any
2405
V. GENERAL PROPERTIES OF SOLUTIONS TO

In this and the following sections we establish properties of
solutions to (12) under the peak-power constraint (14).
First, we give a relation between the average distance of the
rate-optimal constellation and the largest possible minimum distance of any constellation of identical dimension. Define this
latter distance
Thus
A constellation whose minimum distance

said to be distance-optimal.
is equal to zero iff
Proposition 2: Let
-dimensional cutoff rate
where
almost everywhere ( ). In view of (2)
where
is a multivariate circular complex Gaussian denand
sity with zero mean and covariance matrix
For a reference signal

, an easy calculation establishes
the KullbackLeibler divergence [6] of
that for any
from
is
where
are the eigenvalues of the positive definite matrix
attains
is
be a constellation that attains the

. Then
is the mean value
As
combination of Proposition 2 and a sphere-packing bound on

derived in [13, Proposition 2] gives the following bound
on the maximum distance of the constellation attaining cutoff
rate:
Proof of Proposition 2: Let

of the quadratic form
imum value
(18). First we show that
achieve the minin
(20)
As
that
, with equality iff

, we see
, as
is symmetric. Thus,
iff
and, therefore, for distinct
the density functions
are almost ev[17]. Gathering
erywhere linearly independent functions of
are distinct then
together the above results: if
iff
(a.e.)
are identically zero. Hence,
is positive defimplies
inite.
As for assertion ii) of the Lemma we have by Cauchy
Schwarz
with equality iff

(a.e. ). As shown above,
. Hence,
this equality condition can only occur if
under the hypotheses of the lemma.
Assume the contrary. Then
On the other hand,
2406
which is a contradiction. Proposition 2 follows after application

of Jensens inequality to the left-hand side of (20).
is equivalent to equalizing over

tance from to all other codewords
the average dis,
A. Optimality of Equalizer Distributions

to the minimization (18)
The solution
can be found by convex quadratic optimization subject to linear
equality and inequality constraints. Define the Lagrangian
(21)
where
.
We have thus shown that rate-optimal finite-dimensional constellations, defined in Definition 1, must have equalizer distributions. These results are summarized in the following.
,
is a -elewhere
is an undetermined multiplier
ment vector of ones, and
.
that must be chosen to enforce the equality constraint
is convex cup, the KuhnTucker conditions [26] assert
As
that the minimum exists and must satisfy
achieves
Proposition 3: A constellation of dimension
-dimensional peak constrained cutoff rate (12) only
the
over signal matrices
if the optimal distribution
in the constellation is an equalizer distribution of the form
.
An equivalent condition to (24) is that there exists a vector
, not identically zero, lying in the positive
which satisfies
orthant
if
if
or, equivalently,
(28)
if
if
(22)
(23)
is the th column of
. By reordering the entries
where
and the rows and columns of
we can assume, without
of
s satisfying (22) are the first
loss of generality, that the
columns of
with the remainder satisfying (23),
.
Making this assumption, (22) and (23) imply that
and
(24)
are distinct (see
Furthermore, we can assume that
is positive definite.
Lemma 9) and, therefore, by Lemma 2,
Therefore, from (22)(24), the strictly positive component of
is given by
the minimizer
. Define the feasibility set

or, equivalently,
of -dimensional constellations
(29)
is attained by a constellation of diBy Proposition 3,
is nonempty. Substitution of the form
mension only if
specified in Proposition 3
of the optimal probability vector
into (13) we obtain the following alternative characterization of
.
Proposition 4: Let be a positive integer. The peak-power
constrained -dimensional cutoff rate is
(30)
(25)
where is determined via the constraint
or
,
(26)
Note that
since
is positive definite and
since
and
as the elements of
are
. Thus, in view
satisfy the equalof (24), the strictly positive elements of
ization condition
(27)
are the signal matrices in

with strictly
where
. For low SNR it is
positive assigned probabilities
easily shown using Lemma 7 that the equalization condition (27)
in Proposition
Observe that by taking the limit
4, we obtain the cutoff rate of constellations of countable, but
possibly infinite, dimension. The objective function
maximized in (30) is similar to the criterion used in Capons
method, also known as minimum variance distortionless
response (MVDR), for adapting the weights of a beamforming
array of antenna elements and for high-resolution spectral
estimation [19].
and the rate-achieving signal set can
The cutoff rate
be iteratively computed using (30) in the following steps: 1) sein
, 2) use
lect a candidate set of signal matrices
, speciquadratic programming to find the zero entries of
fied by condition (23), e.g., using slack variable or active set
methods [23]; 3) reorder the zero and nonzero entries of
and permute the rows and columns of
as described above;
for ; 5) compute inner product
and
4) solve
take its natural logarithm; 6) perturb the candidate set of signal
matrices and repeat 1)5). This iterative procedure is repeated
for each new set of candidate signal matrices until
attains a maximum. The procedure becomes computationally

heavy when becomes large. Proposition 4 suggests an alternative optimization method which bypasses steps 1) and 2) by
to lie
preselecting the candidate set of signal matrices
.
in
which
It will thus be of interest to establish conditions on
guarantee existence of a positive vector satisfying (28) or,
lies in the
equivalently, which guarantee that the vector
positive orthant. The first result in this direction is that this alof receiver anways holds for a sufficiently large number
tennas.
2407
The left-hand inequality of (33) follows from the sequence of

inequalities
B. A Large Number of Receiver Antennas

The equalization condition (27) is easily manipulated to yield
,
. This suggests that, for
, if is sufficiently large the
any constellation having
,
will satisfy
equiprobable distribution
this condition. We give stronger results below.
The inequality on the second line follows from Cauchy

Schwarz
be a set of peak constrained

Proposition 5: Let
. Define
signal matrices with distinct outerproducts
the finite positive integer
. Then, for
where
timal distribution defined in (18) is
(31)
and the inequality on the third line follows from nonnegativity

. The right-hand inequality of (33) follows from
of
the op-
(34)
(32)
where
is no
The significance of the above proposition is that if
, any peak constrained set of
greater than
is in the feasible set
and thus
signal matrices
. Furthermore,
has an equalizer distribution
,
converges to the equiprobable distribution
as
. Using [13, Proposition 2], the following bound on
for the distance-optimal constellation can be derived:
Now (28) can be expressed as the perturbed linear system by

expressing
As
is positive definite
is invertible. Furthermore,
are strictly less than one and
from (33), the eigenvalues of
by elementary matrix manipulations
The th element of the vector
which by (34) is positive for

small).
Proof of Proposition 5: First observe that under the
since, by Lemma 2,
hypothesis of the proposition
,
. Define the symmetric matrix
.
of are
The elements
is
. Also,
and
Therefore,
are bounded
. When
the eigenvalues
has positive entries. This implies that for

,
and the th element of
is upper-bounded by
and, consequently, as it is nonnegative, of order
by (34). Hence,
of
(33)
From the proof of Proposition 5 we can identify a weaker condition on the number of receiver antennas necessary to ensure
2408
be an equalizer distribution: be sufficiently large to make

a diagonally dominant matrix, i.e., to make
For large , a significantly stronger version of Proposition

and
5 can be established. For arbitrary signal matrices
define the quantity
probabilities
as
since
and
. Therefore,
we have
and the two lower bounds in (35) are established. Furthermore,

let
for arbitrary
With this definition and (12) we observe that the cutoff rate is
where, as above,
expressed as
is the rate-optimal set of signals and probabilities.
Proposition 6: Fix
and assume
Then, as
. Then
(35)
is the distance-optimal constellation defined earwhere

lier in Section V
is its minimal distance, and

corresponds to the
increases to inequiprobable distribution. Furthermore, as
finity the size of any rate-optimal countable constellation and
the cutoff rate both increase without bound.
The above Proposition implies that for a sufficiently large
of receive antennas the -dimensional cutoff rate
number
which is attained by an
takes the form
equiprobable -dimensional distance-optimal constellation.
Proof of Proposition 6: The upper bound
holds independently of and is an immediate consequence of
the inequality
which, when substituted into the discrete form of (5), yields
both and
in as
are monotone increasing

.
VI. CHARACTERIZATION OF FEASIBLE SET

CONSTELLATIONS
OF
Here we derive sufficient conditions and necessary conditions

to lie in the positive orthant. Several of these confor
ditions will be defined recursively in . So as to not confuse
the reader, and to improve clarity of the equations, we will use
lower case throughout this section to distinguish it from the
fixed index .
matrix consisting of columns
For a square
the positive polyhedral cone generated by is [2], [26]
This cone is convex and each of its faces are contained in

.
one of the hyperplanes
are the positive rays
,
The extremals of
.
contains the vector
then, as
It is evident that if
is positive definite,
must have positive elements,
is the optimal equalwhich would guarantee that
izer distribution.
A. Conditions for Feasibility
where the last line follows from the elementary inequality

, for any set of probabilities
, with
.
equality when
in (35) first observe
As for the two lower bounds on
as defined in Proposition 6
that, with and
Here we give two conditions, one sufficient and one neces. The sufficient condition reduces
sary, to ensure
to specifying the largest inscribed right circular cone which fits
while the necessary condition is equivalent
inside
to specifying the smallest right circular cone which contains
. This sufficient condition is slightly weaker than the
sufficient condition presented in Lemma 4 of the next section
and obtained by recursive inscription of a polyhedral cone. On
the other hand, as contrasted with the condition of Lemma 4, the
sufficient condition in this section is not easy to express recursively in .
will denote the th column of
As in (22) and (23),
For
, define
the
by deleting its th column. Define
2409
matrix obtained
(36)
matrix which orthogonally projects vectors

the idempotent
onto the column span of
.
in
Lemma 3: Let be any vector in
to be contained in
dition for
. A sufficient conis
(37)
A necessary condition for
to be contained in
is
(38)
Both of the conditions specified in Lemma 3 require that the

. As a particularly
test vector be close to the ray
simple application of the Lemma, consider setting
; the arithmetic mean of the columns of . In this case,
are orthogonal to
,
as all but the th column of
the sufficient condition becomes
This condition requires that the distance between

and
be
times the prediction error squared of the most
less than
using the remaining columns of
predictable column of
as predictor vectors.
and let
. As
Proof of Lemma 3: Fix
there exists a sphere centered at that fits inside
(see Fig. 1). Such a sphere can have radius no larger
.
than the distance between and the closest face of
are contained in the column span
As the faces of
,
and
is convex, the radius
of
of the largest sphere is given by the distance between and its
orthogonal projection onto this span
Fig. 1. Two-dimensional (2-D) slice of three-dimensional (3-D) cone

cone(E ) (k = 3) with inscribed and circumscribed 3-D spheres used to
establish sufficient and necessary conditions, respectively, for establishing that
1 2 cone(E ). Interior of triangle is slice of cone(E ) in a plane orthogonal
to the vector cu. Vertices of triangle are located at points fc " (i)g
along
extremal rays of cone(E ). The inner dotted circle is the corresponding slice
of the largest possible sphere centered at cu that can be inscribed in cone(E ).
The outer dotted circle is the slice of the smallest possible sphere centered at
cu through which all rays in cone(E ) must pass.
the rays
of
,
, which are the extremals
. Therefore, the radius of this smallest sphere is
Finally, if
this sphere must intersect the ray

in which case
B. A Recursive Construction
Here the objective will be to specify conditions for which
lies in
when it is known that
lies in
. To proceed, we will need a recursive update for
in terms of
. For a set of signal matrices
let the
matrix
be
partitioned as
which is the right-hand side of (37). If the positive ray

passes through this sphere then, by convexity of
must lie in
. This occurs iff the distance from
this ray is less than , i.e.,
,
to
(39)
where
is the vector of pairwise dissimilarity functions

(40)
Since
is arbitrary, we obtain the inequality (37).
there exists a
As for the necessary condition, for any
through which all positive rays in
sphere centered at
must pass (see Fig. 1). The smallest such sphere has
radius equal to the maximum distance between and the faces of
. Points at maximum distance must occur along one of
We also recall the partitioned matrix inverse identity for the case
is positive definite
that
(41)
2410
where
partitioned form
. Let
where
have
is a scalar.
C. Feasibility via Polyhedral Inscribed Cones

Here we specify a simpler positive cone which is inscribed
inside
and is generated by the positive definite
matrix
Proposition 7: For a given set of signal matrices

with distinct outerproducts
(45)
(42)
. We recall the ShermanMorrisseyWoodbury

where
identity for the inverse
where
(43)
(46)
is positive
Proof of Proposition 7: By Lemma 2,
definite. Applying the partitioned matrix inverse identity (41)
to the right-hand side of
is centered along the positive ray

. The set
is equal to the positive
for
while it approaches the aforementioned
orthant
is
positive ray as approaches . Furthermore,
decreasing in in the sense that
for
.
Define the maximum, minimum, mean, and root-mean square
(rms) values of a real vector
Using the definitions of
and
The set
yields (42).
Corollary 1: For a given set of signal matrices

with
assume that the -element
distinct outerproducts
lies in
. Let
be a signal such that
vector
Then the
-element vector
is in
iff a)
and b)
is in
. Furthermore,
.
condition b) implies that
Proof of Corollary 1: It only need be shown that condition
. If this condition holds
b) of the corollary implies that
as
has nonnegative elements.
then
Therefore,
(44)
Lemma 4: Let
tinct outerproducts
sume that
be a signal such that
Then the
if
be a set of signal matrices with dissuch that

. Asfor
. Let
-element vector
lies in
a)
and
(47)
b)
where
and
are defined in (40) and (43).
An alternative form for b) in Lemma 4 is
By definition (43)
(48)
where
Adding this latter equation to the inequality (44) gives
Corollary 1 motivates an iterative procedure, specified

in Hero and Marzetta [13], for generating a monotonically
improving sequence of constellations lying in the feasibility set
defined in Proposition 4. At iteration , assume that a
-dimensional peak constrained constellation
has
attaining the
been constructed whose optimal distribution
.
minimum in (18) is an equalizer distribution:
, with
Consider adding a candidate signal
distinct from
, having associated dissimilarity
given by (40). The two conditions a) and b) of the
vector
to ensure that the optimal
corollary may be used to select
for the updated constellation
distribution
is also an equalizer distribution.
is monotonic increasing over

, under condition a)
and, as by Lemma 2
Interpretation of the sufficient conditions in Lemma 4 is similar to the interpretation of Lemma 3. Condition a) is equivalent
to
(49)
lies close to the column span of
then
If
is small and condition a) restricts the inner product of
the previous unnormalized probability vector
and
to
2411
a narrow range near . Thus, we can expect that constraint a)

will become active only for densely packed signal constellations
). When
is close to one, which occurs for the
(large
(low SNR), condicase of small values of
tion b) places restrictions on the elements of feasible vectors
to ensure low variation about the mean value
. This can
can be selected to minimize the variation of
be ensured if
.
its pairwise dissimilarities
Proof of Lemma 4: Condition a) of Lemma 4 obviously
implies condition a) of Corollary 1. Under the assumption
we will show conditions that a) and
b) of Lemma 4 jointly imply condition b) of Corollary 1. First,
is in
iff
observe that
Second, since
show that
contains
it will suffice to
Fig. 2. Interior of outer triangle is slice of cone(E ) for k = 3 in a plane

orthogonal to the vector c1 . Vertices of outer triangle are located at points
c " (i)
along the extremal rays of cone(E ). The interior of the smaller
equilateral triangle in the center is the slice of cone(F ( )) within the same
plane. The equilateral triangle has center point c1 . The dotted circle is the slice
of a sphere with radius d equal to the distance between c1 and the vertices of
the smaller triangle. b is the distance of c1 to the closest face of cone(E ).
cone(F ( ))
cone(E ) as long as the smaller triangle is inscribed in the
larger triangle; i.e., d
b.
or, equivalently,
is in
, under the conditions of Lemma 4. Applying the ShermanMorrisonWoodbury identity (46) to
where
, and
(50)
(52)
is the orthogonal projector defined in (36).
. Let be the distance
Proof of Lemma 5: Fix a value
and the closest face of
(see
between the vector
Fig. 2). As each face is contained in the column span of
and as
is convex and contains
, is specified by
the projection theorem
(53)
and
The minimum element of the vector (50) is
is an idempotent matrix
Since
As the elements of
which is positive under conditions a) and b) of the lemma. Thus,
is in the positive orthant which establishes condition b) of Corollary 1.
Since
is positive definite and
and
are convex,
does not lie on the boundary of
and
such that
there always exists a value
for all
. As the set
is monotone decreasing in , the largest possible inscribed cone is obtained by using the minimum possible value of . This gives the
least restrictive sufficient condition b) in (47). The form of this
in the Lemma below.
minimum is specified for
with
Lemma 5: For a given set of signal matrices
assume that the -element
distinct outerproducts
lies in
. For
, the minimum value
vector
ensuring
is the value
of
given by
(51)
which occurs for

and
Next let
are nonnegative the projection error norm

is maximum when
is largest,
. In this case
be the maximum distance from

to
(see Fig. 2). As the extremals of
are symmetric about the ray
, each extremal has
. Consider the
a point at identical maximum distance from
where
is the first column of
.
extremal
We thus have
(54)
as long as
.
Evidently,
be the minimum corresponding value of
for
Let
. Equating (53) and (54) yields the following
which
quadratic equation for :
2412
where
. For
, the
quadratic equation has two nonnegative solutions only one of
. For
,
which is contained in
and there is one solution in
. In both cases this solution is
given by (51).
VII. LOW-RATE OPTIMALITY OF UNITARY ORTHOGONAL
CONSTELLATIONS
to be considered is
When the number of signal matrices
sufficiently small, significant simplification of the optimization
(17) is possible. In particular, one obtains optimality of a set of
scaled mutually orthogonal unitary signal matrices and a simple
.
form for
The first result specifies the solution to optimization of the
defined in (10).
dissimilarity measure
define the integer
For given , , and
(55)
is the rank of
We will see below that under some conditions
the signal matrices in the optimal -dimensional constellation.
. Then
Proposition 8: Let
Fig. 3. Top panel shows

given by (55) as a function of the SNR parameter
T M . Bottom panel is a blowup of the first panel over a reduced range of
SNR. The straight line is a least squares linear fit to the upper panel. The linear
approximation has slope 0:32 and zero intercept 0:08. Average residual error
between linear fit and exact M versus SNR step function is less than 0:09 and
maximum error is less than 0:52. By Corollary 2, for T; K; M; M satisfying
T KM and M
M , the curve gives the number of antennas utilized by
the optimal constellation for various SNRs.
signal matrices apply all available transmit power to a single

antenna element over the coherent fade interval .
(56)
Furthermore, the optimal signal matrices which attain
can be taken as scaled rank
mutually orthogonal unitary
matrices of the form
where, for
and
The assumption
is critical and ensures that the singular vectors of and can be chosen as mutually orthogonal
for any set of singular values.
of the optimal matrices
and
increases
The rank
increases from to
from to as the SNR parameter
(see Fig. 3). Numerical evaluation has shown that the functional
and SNR is well approximated by the
relationship between
relation
Proof of Proposition 8: Let

value decompositions
and
have the singular
(57)
unitaries, i.e.,
,
are
unitaries, and
are
real
and
,
diagonal matrices of singular values
respectively.
of
and
only depends
The function
and
through
. Thus, unitaries
and
on
can be arbitrarily chosen without affecting
and
. We denote this functional dehence we can choose them as
. Therependence by writing
fore,
(58)
where
are
Using the form for
given in Lemma 6 (see the Appendix)
where , are the slope and intercept of the least squares linear
fit to the function
(59)
The approximation is a lower bound and underestimates the
, given by (55), by at most over less than
exact value of
. If
0.5% of the SNR range shown in Fig. 3
for
the SNR is sufficiently large, e.g., (from Fig. 3)
and
,
and the optimal signal matrices
transmit antennas. On the other hand, for small
utilize all
,
and the optimal
SNR, i.e., (from Fig. 3)
first consider the inner maximizaFor fixed arbitrary

of the right-hand
tion in (58), i.e., maximization over
side of (59). Observe that the first term in (59) depends only on
and not on
. Observe that as
the singular values
, for any
the product
can be
and
to be mutually
forced to zero by selecting matrices
. Recall
orthogonalpossible under the assumption
that
where
and , defined in Lemma 6, have
and . Therefore, when
and
the same column spaces as
are mutually orthogonal
and the second term in (59)
. This establishes that the opis equal to its maximum value
and
have orthogonal singular vectimum signal matrices
and
and that
tors
Due to the assumed peak-power constraint, each of the above

maximizations on the right-hand side are performed over the
inequality constraint sets
2413
denote the number of nonzero-valued

for one of these candidate maximizers. Then, invoking the constraint
and noting that the ordering of the indexes of
is irrelevant, any candidate maximizer of can be put in the form
(62)
is the only remaining free parameter. Substituting
and now
into (60) we obtain the expression
this expression for
equal to the optimal value of
given by (55).
(56) with
it is
Using the SVD representation (57) with
easily seen that the optimal signal matrices, corresponding to
, have the form
(62) with
where the
s are rank
mutually orthogonal rectangular
unitaries as defined in the statement of the Proposition.
Proposition 8 implies that for low symbol rate, the distanceoptimal signal constellations are constellations of scaled mutu.
ally orthogonal unitary matrices of rank
As these maximizations are of identical form
(60)
where the maximization on the right-hand side is subject to
. Observe that as each summand is monotone
increasing in the inequality constraint in the maximization is
always met with equality. Consider the Lagrangian
where
is an undetermined multiplier. For a suitable value
is identical
of , the maximum of over unconstrained
to the maximum of the right-hand side of (60) over power con. The derivative of with respect to is
strained
where
. Therefore, for each
stationary points of given by either
is one of the roots
, there are five possible

or by
where
(61)
With respect to (61) there are three cases which must be conboth roots are real negative and for
sidered. For
both roots are complex. For
both roots are
real positive. Thus, only the latter case is relevant. It can be veris equal to one of these two positive roots,
ified that, when
the second derivative of is
which is negative only for the root

.
Therefore, we can restrict attention to a candidate maximizer
of for which each can take on either of two
. Let
values: zero or
and let
be as defined in (55).
Corollary 2: Let
and define the set of signal
Assume
by
where
are a
matrices
mutually orthogonal rectangular unitary matrices
set of
(
and
,
). This set of
of rank
signal matrices are equidistant in the sense
and they attain the maximum possible value of
over all signal sets

of dimension .
Proof of Corollary 2: First observe that if
then such an orthogonal unitary set
exists, e.g., the set of
mutually orthogonal permutation
whose columns are formed from
matrices of dimension
disjoint
-dimensional subsets of the columns of . The
satisfies the peak-power constraint
,
set
, with equality and, by Proposition 8, it has the
property that
Let
be any set of signal matrices in
the same power constraint. Then
and thus
maximizes
satisfying
It can be easily verified that the mutually orthogonal unitary

constellation of Corollary 2 satisfies the sufficient conditions a)
is not empty. For this constellaand b) of Lemma 4 when
,
, and
tion
where
. Therefore,
is defined below (48) with

. Furthermore, as
2414
, condition b) reads:
, which is satisfied regardless of the value of .
The final result of this section is an expression for the cutoff
rate.
Now, using the identity (46),

tion in the positive orthant
has a unique solu-
and let
be as defined in (55).
Proposition 9: Let
. Then the peak constrained
Suppose that
dimensional cutoff rate (12) is
Thus, the minimizer
tribution
(63) establishes that
of
is the uniform dis. Substitution of this solution back into
and
is given by (56). Furthermore, an optimal constellais the set of rank
mutually orthogtion attaining
onal matrices specified in Corollary 2 and the optimal proba,
.
bility assignment is uniform:
Any unitary transformation on the columns (spatial coordinates) of a set of signal matrices produces a set of signal ma. In particular, any set of
mututrices with identical
permutation matrices, specified in the
ally orthogonal
course of proof of Corollary 2, has optimal distance properties.
This simple set of signal matrices corresponds to transmitting
energy on a single antenna element at a time, among a total of
elements, in each of the available time slots. Since
is increasing in , the maximum cutoff rate achievable
using these mutually orthogonal unitary matrices is obtained by
.
using the maximum possible number of them:
Observe that the resulting optimal constellation may correspond
to a code of quite low symbol rate, e.g., for
the symbol rate is only one bit per symbol.
It is noteworthy that the optimal peak constrained signal constellation specified by Proposition 9 does not include the zero. Including zero in the signal convalued signal matrix
stellation would allow signaling using ONOFF keying. ONOFF
keying is often proposed for average power constrained signaling over low-SNR channels since it permits energy discrimination at the receiver. As contrasted with ON-OFF keying all signals in the optimal peak constrained signal set have equal power.
We conjecture that the zero signal would result from replacing
the peak-power constraint with an average power constraint in
Proposition 9.
which is
Proof of Proposition 9: Define
strictly less than one under the hypotheses of the proposition.
satisfying the assumpFor any set of signal matrices
tions of Proposition 9, Proposition 8 asserts that
,
, with equality when
consists of the specified mutually orthogonal matrices. Using this inequality and the
,
fact
which, when substituted into (12), yields the expression given

specified in Proposition 9.
for
VIII. CONCLUSION
We have derived representations for the single-user computational cutoff rate for spacetime coding under the Rayleigh
quasi-static fading channel model under a peak transmitted
power constraint. For finite-dimensional constellations the
cutoff rate and the optimal signal distribution were specified
as a solution to a quadratic optimization problem and it was
shown that optimal constellations have codeword distributions
which satisfy an equalization condition. This characterization
of optimality motivated us to study properties of the set of
feasible constellations which satisfy the equalization property.
In particular, we showed that distance-optimal constellations
are close to rate-optimal for large number of receive antennas.
Easily verifiable necessary and sufficient conditions were given
for validating that a given signal constellation lies in the feasible
set. A recursive form was given for one of these conditions
which may be useful for design of feasible constellations.
APPENDIX
PROPERTIES OF
Here we collect various properties of the signal dissimilarity
measure (10).
A. Alternative Form for
can be equivalently expressed in terms of a signal
multiple correlation matrix .
Lemma 6: The signal dissimilarity measure (10) for the
spatio-temporal fading model (1) has the equivalent form
where
(63)
is the
multiple correlation matrix
and
2415
real and positive. Thus, the dissimilarity measure

given in (10), can be written as
are the prewhitened signal matrices
Using the relation

and
denotes the positive-definite square-root factor of positive-definite square matrix .
Proof of Lemma 6: Consider the quantity in the numerator
of expression (10)
where, in the second line, we have used the property

.
For and Hermitian positive-definite matrices we recall
the following determinant identity for block partitioned matrices:
and the fact that
Since
Lemma 7 is established.
Lemma 8: Let
the trace norm of
and
be
satisfies
complex matrices. Then
(64)
Furthermore, denoting by
of , where
where
the invertible square-root factor
. Identifying
, and
,
finishes the proof of
Lemma 6.
and
are the diagonal elements of and
where
, respectively. Equality occurs in (64) iff the trace norm of the
is equal to zero.
antidiagonal matrix
have diagonal eleProof of Lemma 8: Let
and consider the decomposition
ments
into diagonal matrix
and antidiagonal matrix .
Then
As
B. Bounds on
Lemma 7: The dissimilarity measure
where
is a nonnegative function of
satisfies
such that
Proof of Lemma 7: Let

and
denote the
and
, respectively, and
eigenvalues of
denote the eigenvalues of
. Since these are
let
Hermitian nonnegative-definite matrices, these eigenvalues are
with equality iff
The following establishes that it is safe to assume that the

signal matrices in the optimal constellation have distinct outer
products. While this result can also be obtained from the statistical invariance of the Rayleigh- fading model, the lemma below
is proven using a more elementary nonstatistical argument.
have disLemma 9: Let the set of signal matrices
and assume that for some
:
similarity matrix
. Then there exists a
-dimensional subset of
having dissimilarity matrix
such that
2416
Proof of Lemma 9: Without loss of generality, we can as. As

is a function of
sume that
and
only through
and
,
takes the form
(65)
where
is a vector with entries
Let
and
Then using (65)
Since
the Lemma follows.

ACKNOWLEDGMENT
The authors wish to thank the reviewers and the associate
editor for their thoughtful comments on this work.
REFERENCES
[1] I. Abou-Faycal and B. M. Hochwald, Coding requirements for multiple-antenna channels with unknown Rayleigh fading, Bell Labs., Lucent Technol., Murray Hill, NJ, Tech. Rep., 1999.
[2] A. Berman and R. J. Plemmons, Nonnegative Matrices in the Mathematical Sciences. Philadelphia, PA: SIAM, 1994.
[3] C. Berrou, A. Glavieux, and P. Thitimajshima, Near Shannon limit
error-correcting coding and decoding: Turbo codes, in Proc. Int. Conf.
Communications, 1993, pp. 10641070.
[4] E. Biglieri, J. Proakis, and S. Shamai (Shitz), Fading channels: Information-theoretic and communications aspects, IEEE Trans. Inform.
Theory, vol. 44, pp. 26192692, Oct. 1998.
[5] R. Blahut, Principles and Practice of Information Theory. Reading,
MA: Addison-Wesley, 1987.
[6] T. Cover and J. Thomas, Elements of Information Theory. New York:
Wiley, 1987.
[7] G. J. Foschini, Layered space-time architectures for wireless communication in a fading environment when using multi-element antennas,
Bell Labs Tech. J., vol. 1, no. 6, pp. 4149, 1996.
[8] G. J. Foschini and M. J. Gans, On limits of wireless communication in
a fading environment when using multiple antennas, Wireless Personal
Commun., vol. 6, no. 3, pp. 311335, 1998.
[9] R. G. Gallager, A simple derivation of the coding theorem and some applications, IEEE Trans. Inform. Theory, vol. IT-11, pp. 38, Jan. 1965.
, Information Theory and Reliable Communication. New York:
[10]
Wiley, 1968.
[11] J. Hagenauer, Iterative decoding of binary block and convolutional
codes, IEEE Trans. Inform. Theory, vol. 42, pp. 429445, Mar. 1996.
[12] B. Hassibi and B. M. Hochwald, How much training is needed in multiple-antenna wireless links, Bell Labs., Lucent Technol., Murray Hill,
NJ, Tech. Memo, 2000.
[13] A. O. Hero and T. L. Marzetta, On computational cutoff rate for space
time coding, Bell Labs., Lucent Technol., Murray Hill, NJ, Tech.
Memo, 2000.
[14] B. M. Hochwald and T. L. Marzetta, Unitary space-time modulation for
multiple-antenna communications in Rayleigh flat fading, Bell Labs.,
Lucent Technol., Murray Hill, NJ, Tech. Memo, 1999.
, Unitary space-time modulation for multiple-antenna communi[15]
cation, IEEE Trans. Inform. Theory, vol. 46, pp. 543564, Mar. 2000.
[16] B. M. Hochwald, T. L. Marzetta, T. J. Richardson, W. Sweldens, and
R. Urbanke, Systematic design of unitary space-time constellations,
IEEE Trans. Inform. Theory, vol. 46, pp. 19621973, Sept. 2000.
[17] I. A. Ibragimov and R. Z. Hasminskii, Statistical Estimation: Asymptotic Theory. New York: Springer-Verlag, 1981.
[18] S. Kallel and K. Li, Biderectional sequential decoding, IEEE Trans.
Inform. Theory, vol. 43, pp. 13191326, July 1997.
[19] S. M. Kay, Modern Spectral Estimation: Theory and Application. Englewood-Cliffs, NJ: Prentice-Hall, 1988.
[20] T. L. Marzetta and B. M. Hochwald, Capacity of a mobile multipleantenna communication link in Rayleigh fading, IEEE Trans. Inform.
Theory, vol. 45, pp. 139158, Jan. 1999.
[21] J. L. Massey, Variable length codes and the Fano metric, IEEE Trans.
Inform. Theory, vol. IT-19, pp. 196198, Jan. 1972.
[22]
, Coding and modulation in digital communications, in Proc.
1974 Int. Zurich Sem., Digital Communication, Mar. 1974, pp.
E2.1E2.4.
[23] J. J. Mor and S. J. Wright, Optimization Software Guide, ser. Frontiers
in Applied Mathematics. Philadelphia, PA: SIAM, 1993.
[24] P. Narayan and D. L. Snyder, The two-user cutoff rate region for an
asynchronous and a synchronous multiple access channel are the same,
IEEE Trans. Inform. Theory, vol. IT-27, pp. 414419, July 1981.
[25]
, Signal set design for band-limited memoryless multiple access
channels with soft decision demodulation, IEEE Trans. Inform. Theory,
vol. IT-33, pp. 539556, July 1987.
[26] T. Rockafellar, Convex Analysis. Princeton, NJ: Princeton Univ. Press,
1970.
[27] J. E. Savage, Sequential decodingThe computation problem, Bell
Syst. Tech. J., vol. 45, pp. 149175, Jan. 1966.
[28] D. L. Snyder and I. B. Rhodes, Some implications of the cutoff-rate criterion for coded direct detection optical communications, IEEE Trans.
Inform. Theory, vol. IT-26, pp. 327338, May 1980.
[29] V. Tarokh, N. Seshadri, and A. R. Calderbank, Space-time codes for
high data rate wireless communication: Performance criteria and code
construction, IEEE Trans. Inform. Theory, vol. 44, pp. 744765, Mar.
1998.
[30] I. E. Telatar, Capacity of multi-antenna Gaussian channels, AT&T Bell
Labs., Tech. Memo, 1995.
[31]
, Capacity of multi-antenna Gaussian channels, Europ. Trans.
Telecommun., vol. 10, no. 6, pp. 585595, Nov.Dec. 1999.
[32] F.-Q. Wang and D. J. Costello, Probabilistic construction of large
constraint length trellis codes for sequential decoding, IEEE Trans.
Commun., vol. 43, pp. 24392448, Sept. 1995.
[33] L. Wei and H. Qi, Near-optimal limited-search detection on ISI/CDMA
channels and decoding of long convolutional codes, IEEE Trans. Inform. Theory, vol. 46, pp. 14591482, July 2000.
[34] J. M. Wozencraft and R. S. Kennedy, Modulation and demodulation
for probabilistic coding, IEEE Trans. Inform. Theory, vol. IT-12, pp.
291297, July 1966.

Cutoff Rate and Signal Design For The Quasi-Static Rayleigh-Fading Space-Time Channel

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Originaltitel

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Cutoff Rate and Signal Design For The Quasi-Static Rayleigh-Fading Space-Time Channel

Hochgeladen von

Copyright:

Verfügbare Formate

2400

IEEE TRANSACTIONS ON INFORMATION THEORY, VOL. 47, NO. 6, SEPTEMBER 2001