Lecture 13 Wireless Communication

Monsoon semester 2023-24

Topic: Antenna Diversity

Vinay Joseph

NIT Calicut

October 11, 2023

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Plan for studying fading channel and recap
Key problem: how to send and receive bits using a wireless channel?

Step 1: Additive White Gaussian Noise channel without fading

y [m] = x[m] + w [m]
▶ pe ∝ e −SNR possible (approximately).
Step 2: Flat-fading channel (i.e., single tap)
y [m] = h[m]x[m] + w [m]
1
▶ pe ∝ SNR without diversity =⇒ high SNR needed for low pe
1
▶ With time diversity using L diversity branches, pe ∝ SNR L possible
⋆ So, low pe can be attained with much lower SNRs
▶ This lecture focuses on another way to achieve diversity called antenna
diversity.

Step 3: More general frequency-selective channel (i.e., multi-tap)

X
y [m] = hl [m]x[m − l] + w [m]
l

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Antenna diversity: Motivation, Introduction
Time diversity limitation: delays in channels with long coherence time
▶ For a channel with long coherence time, interleaving is needed.
▶ Interleaving lead to delays, since transmission of any codeword needs to
span symbols from multiple coherence time durations
▶ Other diversity forms useful when delay needs to be small
Antenna diversity or spatial diversity: using multiple transmitter
and/or receiver antennas
Types of antenna diversity (figure from [1])

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Antenna separation needed for antenna diversity

Is it enough to just have multiple antennas? No.

Need sufficient antenna separation to ensure different antenna pairs

fade independently
▶ Similar to ”different coherence periods” requirement with time diversity

Required antenna separation depends on

▶ local scattering conditions
▶ carrier frequency
With more scatterers, channel decorrelates over shorter spatial
distances (recall reflecting wall example in Module 1). Examples:
▶ Mobile near ground: One carrier wavelength enough (many scatterers
around)
▶ Base-stations on high towers: Tens of wavelengths may be required

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Receive diversity: Setting

Consider flat-fading channel with one transmit

antenna and L receive antenna:

yl [m] = hl [m]x[m] + wl [m], l = 1, ..., L

where w1 [m],..., wL [m] are independent CN (0, N0 )

random variables for any m
Assume that hl [m],..., hL [m] are independent
CN (0, 1) random variables for any m
▶ Assuming L receive antennas are sufficiently apart

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Receive diversity: Detection and probability of error

Receive diversity setting is equivalent to time diversity with repetition

coding.
▶ Receive diversity model yl [m] = hl [m]x[m] + wl [m], l = 1, ..., L. That
is, y [m ] = h [m ]x[m] + w [m ]
▶ Model for time diversity with repetition: y = h x1 + w

We can thus reuse coherent MRC-based detection. Thus,

▶ Probability of error with BPSK for a given h is given by

a2
q 
P {error|h } = Q 2∥h ∥ SNR , where SNR =
2
N0
▶ Probability of error taking expectation over h ,
h  i 1 1
pe ≤ E exp −∥h ∥ SNR =
2
∝ for large SNR
(1 + SNR)L SNR L

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Transmit Diversity: Introduction

Consider L transmit antennas and one receive

antenna (MISO channel)
▶ Common in older cellular downlinks (basestation
antennas usually cheaper than antennas on each
mobile)
Simple approach: transmit using one antenna at a
time and we get time diversity (but at low rate)
▶ E.g.: Same symbol repeated across antennas is
equivalent to repetition coding
Can we transmit information at higher rate (i.e.,
get coding gain) using all antennas all the time
and still get diversity? Yes. Two options:
▶ Almouti scheme, and more generally using
space-time codes.
▶ Precoding of transmitted information

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Alamouti Scheme

Alamouti Scheme is a simple space-time code for two transmit

antennas and one receive antenna
Setting: Received signal depends on two transmitted symbols
y [m] = h1 [m]x1 [m] + h2 [m]x2 [m] + w [m]
▶ w [m]: CN (0, N0 ) for any m
▶ hl [m] and h2 [m] are independent CN (0, 1) random variables for any m.
⋆ Independence relies on sufficient antenna separation
▶ Assume channel does not change much over two consecutive symbols,
i.e., hi [m] ≈ hi [m + 1]
Alamouti Scheme: symbols u1 and u2 are transmitted in two slots
▶ In time slot m = 1, transmit x1 [1] = u1 , x2 [1] = u2
▶ In time slot m = 2, transmit x1 [2] = −u2∗ , x2 [2] = u1∗

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Alamouti Scheme: Simplified setting

Transmission and reception can be done per pair of slots, and hence
setting can be simplified as follows:
y [1] = h1 u1 + h2 u2 + w [1]
y [2] = −h1 u2∗ + h2 u1∗ + w [2]
▶ w [1] and w [2] are independent CN (0, N0 )
▶ hl and h2 are independent CN (0, 1) random variables

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Alamouti Scheme: Analysis (separating u1 and u2 )
Rewriting the equations (to isolate u1 and u2 )

y [1] = h1 u1 + h2 u2 + w [1] (1)

∗
y [2] = −h1∗ u2 + h2∗ u1 ∗
+ w [2] (2)

Let h = [h1 h2 ]t . So, ∥h ∥2 = |h1 |2 + |h2 |2 .

h1∗ Eqn(1) + h2 Eqn(2) cancels u2 term, and we have
 
h1∗ y [1] + h2 y [2]∗ = |h1 |2 + |h2 |2 u1 + (h1∗ w [1] + h2 w [2]∗ )
=⇒ r1 = ∥h ∥u1 + w1 , normalizing by 1/∥h ∥

h2∗ Eqn(1) - h1 Eqn(2) cancels u1 term, and we have

 
h2∗ y [1] − h1 y [2]∗ = |h1 |2 + |h2 |2 u2 + (h2∗ w [1] − h1 w [2]∗ )
=⇒ r2 = ∥h ∥u2 + w2 , normalizing by (1/∥h ∥)

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Alamouti Scheme: Decoupling u1 and u2 detection
Using the property (related to mean and variance) of linear
combinations of Gaussian random variables, we can show
1
w1 = (h∗ w [1] + h2 w [2]∗ ) ∼ CN (0, N0 )
∥h ∥ 1
1
w2 = (h∗ w [1] − h1 w [2]∗ ) ∼ CN (0, N0 )
∥h ∥ 2
Further, w1 and w2 are also independent
▶ We skip the proof. Can be proved using (i) independence of w [1] and
w [2]; (ii) w1 and w2 obtained using orthogonal transformation of w [1]
and w [2]; (iii) Independent if joint PDF can be expressed as product.
Hence, detection of u1 and u2 can be viewed as two separate
independent detection problems:

r1 = ∥h ∥u1 + w1
r2 = ∥h ∥u2 + w2

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Alamouti Scheme: Probability of error
Using BPSK (±a) for u1 and u2 , we can reuse analysis steps from
coherent detection for detection using r1 and r2 separately, and have:
▶ Probability of error for a given h

a2
q 
P {error|h } = Q 2∥h ∥ SNR , where SNR =
2
N0
▶ Probability of error taking expectation over h ,
h  i 1 1
pe ≤ E exp −∥h ∥ SNR =
2
2
∝ for large SNR
(1 + SNR) SNR 2

Probability of error is same as that for Repetition Coding with MRC.

So, what did we gain from Alamouti Scheme?
▶ Alamouti Scheme (with BPSK) allows us to send one bit per slot
whereas Repetition Coding only sends one bit per two slots.
Thus, Alamouti Scheme provides Coding Gain over Repetition Coding
while providing same diversity benefits (for L=2).

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Precoding for transmit diversity: Model
Model with L transmit antennas: symbol received in slot m
L
y [m] = h [m ] x [m ] + w [m] =
T
X
hl [m]xl [m] + w [m], l = 1, ..., L
l=1

▶ xl [m]: symbol transmitted by l’th antenna in slot m

▶ hl [m]: channel gain associated with l’th transmit antenna and receiver
in slot m, and are independent CN (0, 1) random variables for any m
▶ w [m] ∼ CN (0, N0 )
Precoding using precoding vector p [m ] ∈ CL , information symbol
x̃[m] ∈ C:
x [m] = p [m ]x̃[m],
Maximum Ratio Transmission (MRT) Precoding:

h [m ]∗
p [m ] =
∥h [m ]∥

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 MRT Precoding: Performance

With MRT precoding, resulting model is

y [m] = ∥h [m ]∥x̃[m] + w [m]

▶ The transmissions of multiple transmit antennas are thus combined via

beamforming
▶ We effectively have a model similar to receive diveristy with MRC
combining.
Diversity gain: Using arguments like those for MRC, can show that
MRT precoding provides diversity gain of L

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 MIMO: Outline
MIMO channel: multiple transmit and multiple receive antennas
▶ Assume transmit antennas sufficiently apart, receive antennas too are

Recall: 1 x nR SIMO channel provides receive-diversity gain up to nR

Recall: nT x 1 MISO channel provides transmit-diversity gain up to
nT
nT x nR MIMO channel
▶ can provide up to diversity gain of nR nT
▶ also allows spatial multiplexing of up to min(nR , nT ) streams
⋆ useful in increasing throughput
▶ there is tradeoff between diversity gain and number of spatially
multiplexed streams

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 Takeaways

Antenna diversity utilizes multiple transmitter and/or receiver

antennas
Space-time codes (e.g., Alamouti Scheme) achieve transmit diversity
and coding gain
Comparison of various diversity methods
Bits/
Diversity pe Scheme details
symbol
1
No 1 ∝ SNR Coherent detection
1 1
Time L ∝ SNR L
Repetition Coding and MRC
1 Coherent detection with L an-
Receive 1 ∝ SNR L tennas
1
Transmit 1 ∝ SNR L
Using MRT Precoding

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication

 References

[1] Tse, D., Viswanath, P. (2005). Fundamentals of Wireless

Communication. Cambridge: Cambridge University Press

Vinay Joseph (NIT Calicut) Lecture 13 Wireless Communication