EEE 437 Wireless Communication

Space-Time Communications

Dr. Md. Forkan Uddin

Professor
Department of EEE, BUET
 Space-Time (ST) Configurations
The use of multiple antennas at the receiver and/or transmitter in a
wireless system, popularly known as space-time (ST) wireless or multi-
antenna communications or smart antennas.

2
 ST Communications Benefits

 Higher data rates

 Improved spectral
efficiency
 Better user coverage
 Fewer dropped calls/
improved link reliability

Data rate (at 95% reliability) vs SNR for different antenna

configurations. Channel bandwidth is 200 KHz. 3
 ST Communication System

MT transmit MR receive
antennas antennas

 The input data bits enter a ST coding block that adds parity bits for protection against noise and
also captures diversity from the space and possibly frequency or time dimensions in a fading
environment. After coding, the bits (or words) are interleaved across space, time and frequency
and mapped to data symbols (such as QAM) to generate MT outputs.
 MT symbol streams may then be ST pre-filtered before being modulated with a pulse shaping
function, translated to the passband via parallel RF chains and then radiated from MT antennas.
 These signals pass through the radio channel where they are attenuated and undergo fading in
multiple dimensions before they arrive at the MR receive antennas. Additive thermal noise in the
MR parallel RF chains at the receiver corrupts the received signal.
 The mixture of signal plus noise is matched-filtered and sampled to produce MR output streams.
Some form of additional ST post-filtering may also be applied. These streams are then ST de-
interleaved and ST decoded to produce the output data bits. 4
 Signal Models: SISO Channel

If a signal s(t) is
transmitted, the
received signal:

5
 Signal Models: SIMO Channel

Signal received at the ith receive antenna

Signals received at the MR receive antennas by the MR × 1 vector

6
Signal Models: MISO Channel

7
Signal Models: MIMO Channel

Tx antenna

Rx antenna

8
 Sampled Signal Model: SISO channel
If a sequence of data symbols s[l] (l = 0, 1, 2, . . . ) is to be transmitted, then the received signal

where Ts is the duration of a single symbol (1/Ts ≈ B, the bandwidth of transmission)

and n(t) is additive noise.
If this signal is sampled at instants t = kTs +Δ (k = 0, 1, 2, . . . ), where Δ is the sampling delay,
then the sampled signal response is

and may be rewritten as

h[l] (l = 0, 1, 2, . . ., L − 1) is the Ts spaced sampled channel.

9
L is the channel length measured in sampling periods.
 Sampled Signal Model: SISO channel

Frequency flat channel

Frequency selective channel

T successive received signal samples are therefore

10
 Sampled Signal Model: SISO channel
Frequency selective channel (…contd)

The input–output relation for frequency selective fading can alternatively be expressed as

11
 Sampled Signal Model: MIMO channel
Frequency flat channel

Zero-Mean Circulant Symmetric Complex

Gaussian
Independent, zero mean and equal variance.

where y[k] is the received signal vector with dimension MR × 1, s[k] is the transmit
signal vector with dimension MT × 1 and n[k] is the MR × 1 spatio-temporally white
ZMCSCG noise vector with variance No in each dimension.

Since the output at any instant of time is independent of inputs at previous times, we
can drop the time index k for clarity and express the input–output relation simply as

MR × 1

MR × 1 M R × MT MT × 1 12
 Sampled Signal Model: MIMO channel

Frequency selective channel

Both of length L

13
Sampled Signal Model: MIMO channel

14
Channel Capacity: Flat fading MIMO Channel
The maximum error-free data rate that a channel can support is
called the channel capacity.
Input–output relation for the MIMO channel
Zero-Mean Circulant Symmetric
Complex Gaussian
A complex Gaussian RV, Z = X + jY
is ZMCSCG if X and Y are
independent real Gaussian RV with
where zero mean and equal variance.

 y is the MR × 1 received signal vector

 s is the MT × 1 transmitted signal vector
 n is the MR × 1 ZMCSCG noise with covariance matrix E{nnH} = NoIMR
 Es is the total average energy available at the transmitter over a symbol period.
 The covariance matrix of s, Rss = E{ssH}, (s is assumed to have zero mean) must
satisfy Tr(Rss) = MT in order to constrain the total average energy transmitted over a
symbol period [Tr( ) is the Trace of a matrix].
15
 (Deterministic Channel
& Perfectly Known at the
Receiver) 16
Channel Capacity: Flat fading MIMO Channel

H is Deterministic and Known at the Receiver:

f(s) is the probability distribution of the vector s
I(s; y) is the mutual information between vectors s and y
H(y) is the differential entropy of the vector y
H(y|s) is the conditional differential entropy of the vector
y, given knowledge of the vector s.
H(Hs|s) =0 since H is deterministic, entronpy is zero
Since the vectors s and n are independent, H(n|s) = H(n)
Thus maximizing I(s; y) is equivalent to the maximizing H(y)
Covariance matrix of y:

17
Channel Capacity: Flat fading MIMO Channel

18
Channel Capacity: Flat fading MIMO Channel

det(AB) = det(A) det(B)

19
Channel Capacity: Flat fading MIMO Channel
Case I: H is Unknown to the Transmitter (slide 1):
Vector s may be chosen to be statistically non-preferential:
This implies: signals are independent and equi-powered (= Es/MT) at the transmit antennas.

(Eigen decomposition, Q: Unitary matrix, Λ: Diagonal matrix)

A: (m × n) matrix; B: (n × m) matrix

Q and Λ:
MR × MR matrix

r is the rank of the channel; diagonal elements of Λ are λi (i = 1, 2, . . ., r ),

which are the positive eigen values of HHH.
 Thus the capacity of the MIMO channel is the sum of the capacities of r
SISO channels, each having power gain λi (i = 1, . . ., r) and transmit power
Es/MT 20
Channel Capacity: Flat fading MIMO Channel
Case I: H is Unknown to the Transmitter (slide 2)
Given a fixed total channel power transfer, Frobenius norm
, what is the nature of the channel H that maximizes capacity?

ζ=M2
Squared Frobenius norm of H:

The capacity of an orthogonal MIMO channel is therefore M times the scalar

channel capacity. 21
Channel Capacity: Flat fading MIMO Channel
Case II: H is Known to the Transmitter (slide 1)
When the channel is known at both the transmitter and receiver, the individual channel
modes may be accessed through linear processing at the transmitter and receiver.

 Consider a ZMCSCG signal vector s of dimension r × 1, where r is the rank of the

Channel H to be transmitted.
 The vector is multiplied by the matrix V prior to transmission.
 At the receiver, the received signal vector y is multiplied by the matrix UH.
Singular value decomposition

22
 Channel Capacity: Flat fading MIMO Channel
Case II: H is Known to the Transmitter (slide 2)
Previous equation shows that with channel knowledge at the transmitter, H can be explicitly
decomposed into r parallel SISO channels satisfying

Singular value decomposition

If λi ≥ λi+1

23
Channel Capacity: Flat fading MIMO Channel
Case II: H is Known to the Transmitter (slide 3)

Since the transmitter can access the spatial sub-channels, it can allocate variable energy
across the sub-channels to maximize the mutual information. The mutual information
maximization problem now becomes

(using Water-pouring/Water-filling algorithm)

24
 Water-Pouring/Filling Algorithm (1)
Setting the iteration count p to 1, we first calculate the constant μ

Using the value of μ found above, the power allocated to the ith sub-channel can be
calculated using

count p incremented by 1.

The optimal water-pouring power allocation strategy is found when power allocated
to each spatial sub-channel is non-negative.
The capacity of the MIMO channel when the channel is known to the transmitter is
necessarily greater than (or equal to) the capacity when the channel is unknown to
the transmitter.
25
Water-Pouring/Filling Algorithm (2)

26
 (Random Channel
and Perfectly Known at
the Receiver) 27
Channel Capacity: Flat fading MIMO Channel
Let consider the cases of random MIMO channels.
 For now, we assume the elements of H, hi,j (i = 1, 2, . . ., MR, j = 1, 2, . . ., MT) are
normalized so that E{|hi, j|2} = 1.
 Each realization of the fading channel has a maximum information rate associated
with it depending on whether the channel is known or unknown to the transmitter.
The signal received at the ith receive antenna, yi, is given by

 hi: Vector of dimension 1 × MT , represents the ith row of H

 ni: ith element of n.
 Since E{|hi, j|2} = 1 and Tr(Rss) = MT, it follows that E{|yi |2} = Es + No
 Average received SNR at the ith receive antenna is ρ = Es/No
 Es/(MTN0 ): Average SNR at any receive antenna contributed by a single transmit
antenna (assuming equal power allocation).
 Clearly, if MT = MR = 1, ρ is the average SNR at the receiver.
28
 Ergodic Capacity (1)
The ergodic capacity of a MIMO channel is the ensemble average of the
information rate over the distribution of the elements of the channel matrix H.
Ergodic capacity is also called the Shannon capacity of the channel.

When the channel is unknown to the transmitter, the ergodic capacity is given by

Ergodic capacity when the channel is known to the transmitter is the ensemble
average of the capacity achieved when the waterpouring optimization is performed for
each realization of H, and is given by

29
 Ergodic Capacity (2)

Channel is unknown to the transmitter

30
 Ergodic Capacity (3)
MT = MR = 4
Hw MIMO channel

Q. Explain why the gap in capacity between the channel known and unknown cases
decreases at higher SNR for Hw MIMO channels with MT = MR = M.
31
 Outage Capacity (1)
 Outage analysis quantifies the level of performance (in this case capacity) that is
guaranteed with a certain level of reliability.
 q% outage capacity Cout,q is defined as the information rate that is guaranteed for
(100 − q)% of the channel realizations, i.e., P(C ≤ Cout,q) = q%

CDF of information rate for

the Hw MIMO channel with
MT = MR = 2 and a SNR of
10 dB.

32
 Outage Capacity (2)

Channel is unknown
to the transmitter

 Outage capacity increases with SNR and is higher for larger antenna configurations.

33
 Outage Capacity (3)
MT = MR = 4
Hw MIMO channel

10% outage capacity of a MT = MR = 4 Hw MIMO channel with

and without channel knowledge at the transmitter.
34
(Perfect Channel Knowledge at
the Receiver)

35
 Diversity Gain (1)
• Diversity provides the receiver with multiple (ideally independent)
looks at the same transmitted signal
• Assume that a symbol s, drawn from a scalar constellation with unit
average energy, is to be transmitted.
• There are M identical independent Rayleigh fading links between the
transmitter and receiver.
• Received signal

 Es/M is the symbol energy available to the transmitter for each of the M diversity
branches
 yi is the received signal corresponding to the ith diversity branch
 hi is the channel transfer function corresponding to the ith diversity branch
 ni is additive ZMCSCG noise with variance No. Furthermore, consider E{ninj*} = 0,
which ensures that the additive noise is uncorrelated across the diversity branches.
36
 Diversity Gain (2)
• Given multiple faded versions of the transmitted signal s, the post-processing SNR
η at the receiver can be maximized through a technique known as maximal ratio
combining (MRC).
• Assuming perfect channel knowledge at the receiver, the M received signals are
combined according to

• SNR η is given by

where ρ = Es/No may be interpreted as the average SNR at the receive antenna in a SISO
fading link.
• Assuming ML detection at the receiver, the corresponding probability of symbol error is

37
 Diversity Gain (4)
Applying the Chernoff bound, Pe can be upper-bounded as,

x=|h|^2 with unit power

f(x)=exp(-x)

In the high SNR regime (ρ >>1),

38
 Coding Gain vs Diversity Gain
While both diversity and coding improve system performance (decrease error rate), the nature of
these gains is very different. While diversity gain manifests itself in increasing the magnitude
of the slope of the SER curve, coding gain shifts the error rate curve to the left. The SER for
a system employing both coding and diversity techniques at high SNR can be approximated by

 c is a scaling constant specific to

the modulation employed and the
nature of the channel
 γc (γc ≥ 1) denotes coding gain
 M is the diversity order of the
system

The SNR advantage due to diversity gain increases with increasing diversity order and lower
target error rate. On the other hand, the coding gain is typically constant at a high enough SNR.
39
 Spatial Diversity: Receive Antenna Diversity (1)
Consider a system with a single antenna at the transmitter and multiple antennas at
the receiver (SIMO channel).
Assuming flat fading conditions, the channel vector h for such a system is given by

Assuming that the symbol s to be transmitted is drawn from a scalar constellation with
unit average energy, the input-output relation for the channel may be expressed as
y is the MR × 1 received signal vector
n is MR × 1 ZMCSCG noise vector with E{nnH} = NoIMR.

To maximize the received SNR, the receiver performs MRC, and output z becomes

Since the noise vector n is spatially white, the SNR at the receiver η is given by

40
 Spatial Diversity: Receive Antenna Diversity (2)
If the separation between the antennas at the receiver is greater than the coherence
distance (DC) and if we assume a rich scattering environment, then h = hw.
Average probability of symbol error for such a channel is given by

In the high SNR regime,

Thus the diversity order of the system is equal to the number of antennas at the
receiver MR. Furthermore, since E{||h||F2} = MR for h = hw, the average SNR at the
receiver, E{η}, is given by

Hence, in addition to diversity gain, the average SNR at the receiver is enhanced
by a factor of MR over a standard SISO link due to array gain which is expressed as
10 log10MR in dB. 41
 Spatial Diversity: Receive Antenna Diversity (3)
 BPSK modulation
 AWGN SISO channel (no fading)

With MR = 4, system outperforms an

AWGN link for a target BER greater
than 10−5. This can be explained by
the role of array gain (which is 6 dB
for four antennas) which provides
the initial advantage. However, at
lower target BER, the penalty due to
fading overwhelms the array gain
advantage.

Receive diversity techniques are capable of extracting full diversity gain and
array gain. Performance improvement is proportional to the number of receive
antennas used. However, deploying multiple antennas at the terminal receiver is often
not feasible due to cost or space limitations. Instead, multiple antennas at the
transmitter in combination with transmit antenna diversity techniques is more popular.
42
Spatial Diversity: Transmit Antenna Diversity (1)
Exploiting spatial diversity in systems with multiple antennas at the transmitter
requires that the signal be pre-processed or pre-coded prior to transmission.
Why is the pre-processing of the transmit signal required?
Consider a symbol s that is to be transmitted over a system with two transmit antennas
and a single receive antenna. A simplistic attempt to exploit diversity would be to
transmit the signal from both transmit antennas at the same time. Assuming a flat
fading environment, where the channel signatures corresponding to the transmit
antennas are given by h1 and h2, the received signal y may be expressed as

Es is the average energy available at the transmitter over a symbol period evenly
divided among the transmit antennas and n is the ZMCSCG noise at the receiver. Now
the sum of two complex Gaussian random variables is also complex Gaussian (h1 +
h2)/sqrt(2) is ZMCSCG with unit variance. So we get
where h is ZMCSCG with E{|h|2} = 1.

Therefore, this naive technique does not provide diversity. 43

Spatial Diversity: Transmit Antenna Diversity (2)
Case I: Channel unknown to the transmitter: MISO
We assume two antennas at the transmitter and a single receive antenna.
Alamouti scheme:

1st symbol 2nd symbol

period period

We assume that the channel remains constant over the two symbol periods, and is
frequency flat. Therefore, h = [h1, h2] and the signals y1 and y2 received over the two
symbol periods are given by
n1 and n2 are ZMCSCG noise with
E{|n1|2} = E{|n2|2} = No and Es/2 is the
average transmit energy per symbol
period per antenna.

44
Spatial Diversity: Transmit Antenna Diversity (3)
Case I: Channel unknown to the transmitter: MISO (contd.)

Hence, the effective channel output for symbols si (i = 1, 2) becomes

and the received SNR, η, per symbol is given by

Assuming h = hw, for high SNR regime

45
 Spatial Diversity: Transmit Antenna Diversity (4)
Case I: Channel unknown to the transmitter: MISO (contd.)
 The Alamouti scheme therefore extracts a diversity order of 2 (full MT diversity), even in the
absence of channel knowledge at the transmitter.

 Thus, the absence of channel knowledge at the transmitter does not allow array gain.

While both schemes outperform a

SISO fading link and extract the
same diversity gain (the SER
curves have the same slope),
receive diversity (MT = 1, MR = 2)
outperforms the Alamouti scheme
(MT = 2, MR = 1) (transmit
diversity) because of array gain.

46
 Spatial Diversity: Transmit Antenna Diversity (5)
Case II: Channel known to the transmitter: MISO
The vector channel:

To exploit spatial diversity, the same signal is transmitted from each transmit
antenna after being weighted appropriately, so that the signals arrive in phase at the
receive antenna and add coherently. The signal at the receiver is given by

 y is the received signal

 w is a weight vector of dimension MT × 1
 n is ZMCSCG noise.

 w must be chosen subject to ||w||F2 = MT, to ensure that the average total power of the
transmitted signal is Es .
 Clearly, w that maximizes the received SNR is given by

 This scheme is known as transmit-maximal ratio combining (transmit-MRC).

 The SNR at the receiver η is given by

47
 Spatial Diversity: Transmit Antenna Diversity (6)
Case II: Channel known to the transmitter: MISO (contd.)
If h = hw, then the average probability of symbol error
in the high SNR regime is upper-bounded by

The average SNR at the receiver

is improved by a factor of MT over
a SISO link, and is the transmit
SISO Link array gain. Hence, if perfect
channel knowledge is available to
MISO: MT = 2 the transmitter, transmit-MRC
will deliver array gain and
diversity gain.

Transmit-MRC outperforms the

Alamouti scheme on account of
BPSK modulation array gain while providing the
same diversity gain. 48
 Spatial Diversity: Transmit Antenna Diversity
Case III: Channel unknown to the transmitter: MIMO
 Consider a MIMO system with MT = 2, MR = 2 and use Alamouti scheme.
 We assume that the channel remains constant over consecutive symbol periods.
 Assume also a frequency flat channel.

Signals received at the receive antenna array over consecutive symbol periods are y1 and y2.

n1, n2, n3 and n4 are uncorrelated ZMCSCG noise samples with E{|ni|2} = No (i = 1, 2, . . ., 4).

As in the MISO channel (with the channel unknown to transmitter), the energy available at the
transmitter is evenly divided between the transmit antennas.

The receiver now forms a signal vector y according to

49
 Spatial Diversity: Transmit Antenna Diversity
Case III: Channel unknown to the transmitter: MIMO (contd.)

Hence, the effective channel for either data symbol si (i = 1, 2) becomes

with the corresponding received SNR given by

50
 Spatial Diversity: Transmit Antenna Diversity
Case III: Channel unknown to the transmitter: MIMO (contd.)
Assuming H = Hw, it follows that the average probability of symbol error in the high
SNR regime is upper-bounded by

Therefore, the Alamouti scheme extracts order MT MR diversity (= 4 in this case),

though channel knowledge is not available to the transmitter.

Therefore, in the absence of channel knowledge at the transmitter, the Alamouti

scheme is capable of extracting only receive array gain.

Alamouti technique may be used to extract diversity in MIMO systems with two
transmit antennas and any number (MR) of receive antennas – we get 2MR order
diversity (full diversity) and an array gain of MR.
51
 Spatial Diversity: Transmit Antenna Diversity
Case IV: Channel known to the transmitter: MIMO
 We consider a system with MR X MT MIMO system.
 When the channel is known to the transmitter, spatial diversity may be extracted
through a technique known as dominant eigenmode transmission. Here, as with
transmit-MRC for MISO systems, the same signal is transmitted from all antennas in
the transmit array with weight vector w.

The received signal vector is then given by

y is the MR × 1 received signal vector
H is the MR × MT channel transfer function
w is the MT × 1 complex weight vector
n is spatially white ZMCSCG noise
We note that w must satisfy ||w||F2 = MT to maintain total average transmitted energy
Let the receiver form a weighted sum of antenna outputs according to

g is a MR × 1 vector of complex weights

52
 Spatial Diversity: Transmit Antenna Diversity
Case IV: Channel known to the transmitter: MIMO (contd.)
SNR at the receiver, η:

Maximizing the SNR at the receiver is equivalent to maximizing

With the appropriate (SNR maximizing) choice of w and g, the effective input–output
relation for the channel reduces to
n is ZMCSCG noise with variance No
σmax2 = λmax , where λmax is the maximum
eigenvalue of HHH

SNR at the receiver: Therefore, array gain = E{λmax} in

dominant eigenmode transmission
λmax may be upper- and lower-bounded according to
r is the rank of H
53
 Spatial Diversity: Transmit Antenna Diversity
Case IV: Channel known to the transmitter: MIMO (contd.)
Since H = Hw is full-rank with probability 1, r = min(MT , MR). Hence,
the average SNR at the receiver may be upper- and lower-bounded as

Thus, array gain when the channel is known to the transmitter is greater
than or equal to the array gain when the channel is unknown.

SER at high SNR:

Thus, SER must maintain a slope of magnitude MT MR, as a function of

SNR (on a log–log scale). Hence, we can conclude that dominant
eigenmode transmission extracts a full diversity order of MT MR. 54
 Spatial Diversity: Transmit Antenna Diversity
Case IV: Channel known to the transmitter: MIMO (contd.)

BPSK modulation
MT = 2, MR = 2

 As expected, both schemes

extract the same order of
diversity.
 However, dominant
eigenmode transmission
outperforms the Alamouti
scheme due to the higher
array gain.

55
 Spatial Diversity: Summary

Array gain and diversity order for different multiple antenna

configurations (Channel known to the receiver):

(CU = channel unknown to the transmitter

CK = channel known to the transmitter)
56
57
 Spatial Multiplexing (SM)
 SM (a special case of ST coding) sends MT independent
symbols per symbol period, so rs = MT
 Depending on the choice of ST coding, rs ϵ [0, MT]
 Two types - Uncoded and Coded SM
 For uncoded SM, diversity order is MR
 SM with no coding may be considered as a ST code with
spatial rate MT with MR order diversity

58
 Coded SM: Horizontal Encoding (HE)
Horizontal encoding (HE)

 Bit stream is first demultiplexed into MT separate streams.

 Each stream undergoes independent temporal coding, interleaving and symbol
mapping, and is transmitted from one antenna.
 The spatial rate: independent data streams can be transmitted simultaneously or number
of symbol transmission per frame length time
So spatial rate is clearly rs = MT =N/T
Frame length=the duration of time over which the multiple antennas are used to transmit
independent data streams, T
 The signaling rate is the number of bit transmission per frame length time =qK/T
Temporal code rate/channel coding rate, rt =K/N
59
 Horizontal Encoding (HE)

 Signalling rate= qrtMT bits/transmission

 HE scheme (like uncoded SM) can at most achieve MR order diversity,
since any given symbol is transmitted from only one transmit antenna and
received by MR receive antennas. This is a source of the sub-optimality of
this particular encoding architecture.
 But it does simplify receiver design.
 Coding gain depends on the strength of the temporal code
 Array gain= Signal strength gain=MR is achievable.

60
 Coded SM: Vertical Encoding (VE)
Vertical encoding (VE)

 Bit stream undergoes temporal coding, interleaving and symbol mapping after which
it is demultiplexed into MT streams that are transmitted over the antennas.
 This form of encoding can reach optimality since potentially each information bit can
be spread across all antennas.
 However, VE requires joint decoding of the substreams at the receiver and can be
very complex.
 The spatial rate is rs = MT and the signaling rate is qrtMT bits/transmission.
 Since the information symbols are spread over more than one antenna, VE can
achieve a diversity order greater than MR.
 Coding gain will depend on the temporal code design and array gain of MR is
achievable.
61
 Coded SM: Diagonal Encoding (DE)
Diagonal encoding (DE)

 Incoming data stream first undergoes HE encoding, after which each

codeword is split into frames/slots.
 These frames pass through a stream rotator that rotates the frames in a
round robin fashion so that the bit stream–antenna association is
periodically cycled.
 Making the codeword large enough ensures that the codeword from
any one demultiplexed stream is transmitted over all MT antennas.

62
 Coded SM: D-BLAST Encoding
D-BLAST Numerals represent layers
encoding belonging to the same codeword

 The D-BLAST transmission technique follows DE type encoding strategy

(includes an initial wasted triangular block where no transmission takes place).
 This initial wastage is required to enable optimal decoding.
The encoder uses a space time arrangement that corresponds to a diagonal
layering. The information bit stream coming from the source is demultiplexed into
several substreams (serial to parallel), and each substream is coded separately and
mapped to complex symbols. Then the symbols of each substream are dispersed
“diagonally” across antennas and time.
63
 The spatial rate is MT and the signaling rate is qrtMT bits/transmission.
 D-BLAST like schemes can achieve full MTMR diversity if the temporal
coding with stream rotation is optimal (Gaussian code books with infinite
block size). Coding gain will depend on the temporal code design and an
array gain of MR is achievable.

64
65
 Some Identities and Properties
IID (spatially white) MR × MT MIMO channel Hw

Correlated channels imply that elements of H are correlated

R is the MT MR × MT MR covariance matrix:

R is a positive semi-definite Hermitian matrix

If R = IMT MR, then H = Hw

66
 Some Identities and Properties
Singular values of H (MR x MT matrix with rank r)
U: MR × r matrix
V: MT × r matrix
σi ≥ 0 and σi ≥ σi+1,
σi is the ith singular value of H

HHH is an MR × MR positive semi-definite Hermitian matrix.

 Eigen decomposition of HHH = QΛQH
Q is an MR × MR matrix satisfying
QHQ = QQH = IMR and Λ = diag{λ1, λ2,· · ·, λMR} with λi ≥ 0.

If λi ≥ λi+1

Since H is random, λi (eigenvalues of HHH) is also a random variable.

67
 Some Identities and Properties
The squared Frobenius norm of H:
HHH is an MR × MR
positive semi-definite
Hermitian matrix.

The statistics of determines diversity performance.

A Hermitian matrix is a complex square matrix that is equal to its own conjugate transpose.

68
 Some Identities and Properties
Orthogonal matrix:
A square matrix Q with real numbers is said to be an orthogonal matrix, if its
transpose is equal to its inverse matrix (i.e., QT = Q-1) or we can say, when the product
of a square matrix and its transpose gives an identity matrix (=> QTQ = QQT = I), then
the square matrix is known as an orthogonal matrix. An orthogonal matrix Q is
necessarily invertible and unitary.

Unitary matrix:
In linear algebra, a complex square matrix U is unitary if its conjugate transpose UH is
also its inverse, that is, UUH= UHU = I.

Trace of a square matrix:

In linear algebra, the trace of a square matrix A, denoted tr(A), is defined to be the sum
of elements on the main diagonal (from the upper left to the lower right) of A.
Rank of matrix:
1. If A is a m-by-n matrix, Rank(A) ≤ min(m, n)
2. Rank(AB) ≤ min(rank(A), rank(B))
3. Rank(AAH) = Rank (AHA) = Rank (A) = Rank (AH)

69
 References
Text Book:
Introduction to Space-Time Wireless Communications – A.
Paulraj, R. Nabar and D. Gore
(Cambridge University Press)

Chapter: 3, 4, 5, 6

70
 Additional Information (David Tse)
We make the standard assumption that w(t) is zero-mean additive white Gaussian noise (AWGN) with
power spectral density N0/2, i.e., E[w(0)w(t)] = N0/2 δ(t).

Hence the discrete-time noise process w(m) is white, i.e., independent over time; moreover, the real and
imaginary components are i.i.d. Gaussians with variances N0/2.

Complex noise, w ∼ CN(0, N0)

The noise energy per complex symbol time is N0

The assumption of AWGN essentially means that we are assuming that the primary source of the noise is at
the receiver or is radiation impinging on the receiver that is independent of the paths over which the signal is
being received. This is normally a very good assumption for most communication situations.

The orthogonal modulation scheme considered here uses only real symbols and hence transmits only on the
I channel. Hence it may seem more natural to define the SNR in terms of noise energy per real symbol, i.e.,
N0/2. However, later we will consider modulation schemes that use complex symbols and hence transmit on
both the I and Q channels. In order to be consistent throughout, we choose to define SNR this way.

Channel tap gains: For Rayleigh fading, all the taps are modeled as: h ∼ CN(0, σ2). Then the magnitude |h|
of each tap is Rayleigh distributed and |h|2 is exponentially distributed. For flat fading, only one tap in the
channel model.
Usually, for Rayleigh fading, h ∼ CN(0, 1), where we normalize the variance to be 1.
Recall that we normalized the channel gain such that E[h2] = 1. 71