NETW701

Wireless Communications
Dr. Wassim Alexan
Winter 2018

Lecture 8
 NETW705

Mobile Communication
Networks
Dr. Wassim Alexan
Winter 2018

Lecture 8

Wassim Alexan 2
 What is Diversity?
◆ Diversity techniques offer two or more inputs at the Rx such that the fading
phenomena among these inputs are uncorrelated

◆ If one radio path undergoes a deep fade at a particular point in time, another
independent (or at least highly uncorrelated) path may have a strong signal at
that input

◆ By having more than one path to select from, both the instantaneous and
average SNR at the Rx could be improved

◆ Diversity techniques include



◆ Space diversity

◆ Frequency diversity

◆ Time diversity

Wassim Alexan 3
 Diversity Techniques: Space Diversity

Fig. 1. Space diversity at the Rx side.

(Figure is copied from Prof. Tallal’s lecture slides, with his permission)

◆ Rx space diversity: M different antennas appropriately separated and deployed

at the Rx to combine uncorrelated fading signals

Wassim Alexan 4
 Diversity Techniques: Space Diversity

Fig. 2. Space diversity at the Tx side.

(Figure is copied from Prof. Tallal’s lecture slides, with his permission)

◆ Tx space diversity: M different antennas appropriately separated and deployed

at the Tx to obtain uncorrelated fading signals at the Rx

◆ Tx space diversity: The total transmitted power is split among the antennas

Wassim Alexan 5
 Diversity Techniques: Frequency Diversity

Fig. 3. Frequency diversity.

(Figure is copied from Prof. Tallal’s lecture slides, with his permission)

◆ Modulate the signal through M different carriers

◆ The separation between the carriers should be at least the coherence

bandwidth BC
◆ Different copies of the signal undergo independent fading

◆ Only a single antenna is needed

◆ The total transmitted power is split among the carriers

Wassim Alexan 6
 Diversity Techniques: Time Diversity

Fig. 4. Time diversity.

(Figure is copied from Prof. Tallal’s lecture slides, with his permission)

◆ Transmit the desired signal in M different time periods (i.e. each symbol is
transmitted M times)

◆ The interval between the transmission of the same symbol should be at least
the coherence time TC

◆ This allows for different copies of the signal to undergo independent fading

Wassim Alexan 7
 Diversity Combining Techniques
◆ Selection combining, SC

◆ Maximal ratio combining, MRC

◆ Equal gain combining, EGC

Wassim Alexan 8
 Selection Combining

Fig. 5. Selection combining.

(Figure is copied from Prof. Tallal’s lecture slides, with his permission)

◆ Select the strongest signal

Wassim Alexan 9
 Selection Combining
◆ Consider M independent Rayleigh fading channels available at the receiver

◆ The average SNR at all diversity branches is SNR = Γ

◆ The instantaneous SNR at diversity branch i is SNR = γi

◆ Then, the Rayleigh fading voltage would be

1 γi
f (γi) = exp-  (1)
Γ Γ

◆ The outage probability of a single branch is then given by

1 γ
i γ γ
Pr[γi < γ] = exp -  ⅆ γ = 1 - exp-  ( 2)
Γ 0 Γ Γ

Wassim Alexan 10
 Selection Combining
◆ The outage probability of all M independent branches less than γ would be
γ M
Pr[γ1, γ2, ..., γM < γ] = 1 - exp-  (3)
Γ

Wassim Alexan 11
 Exercise 1
Assume four–branch diversity is used, where each branch receives an indepen-
dent Rayleigh fading signal. If the average SNR is 20 dB, determine the probabil-
ity that the SNR will drop below 10 dB.

Compare this with the case of a single receiver without diversity

Wassim Alexan 12
 Exercise 1 Solution
◆ For this exercise, the specified threshold is γ = 10 dB, Γ = 20 dB, and there are
four branches. Using equation (3)

γi M
Pr[γ1, γ2, γ3, γ4 < 10] = 1 - exp- 
Γ
4
1010/10
= 1 - exp- 20/10 
10
= 0.000082
= 8.2 × 10-5

◆ For this exercise, the specified threshold is γ = 10 dB, Γ = 20 dB, and there are
four branches. Using equation (2)
γi
Pr[γ < 10] = 1 - exp- 
Γ
1010/10
= 1 - exp- 20/10 
10

Wassim Alexan 13
 Exercise 1 Solution
= 0.095
= 9.5 × 10-2

◆ Notice that without diversity, the SNR drops below the specified threshold with
a probability that is three orders of magnitude greater than if four branch
diversity is used!

Wassim Alexan 14
 Maximal Ratio Combining

Fig. 6. Maximal ratio combining (MRC).

(Figure is copied from Prof. Tallal’s lecture slides, with his permission)

◆ Selection combining does not benefit from the power received across all
diversity branches

◆ MRC conducts a weighted sum across all branches with the objective of
maximizing the SNR

Wassim Alexan 15
 Maximal Ratio Combining
◆ Consider M independent Rayleigh fading channels available at the receiver

◆ Envelope applied to the receiver detector

M
rMRC =  r i Gi (4)
i=1

◆ Total noise power applied to the receiver detector

M
NMRC = N  G2i ( 5)
i=1

◆ SNR at the receiver detector

(rMRC)2 M 2 M
γMRC ∝ =  ri Gi  N  G2i (6)
NMRC i=1 i=1

Wassim Alexan 16
 Maximal Ratio Combining
◆ Cauchy’s inequality (∑ ai bi)2 ≤ ∑ a2i ∑ b2i

2 2
∑M
i=1
ri
N
N Gi  ∑M ri
i=1  N 2 ∑M
i=1  N Gi
γMRC = 2
≤ 2
N ∑M
i=1 Gi N ∑M
i=1 Gi ( 7)

1 M 2 M
= r
i=1 i i=1 γi
=
N

◆ γMRC is maximized when Gi are linearly proportional to the channel coefficients

hi

◆ Thus, to optimize the performance of MRC, channel measurements must be

carried out before its application

Wassim Alexan 17
 Maximal Ratio Combining
◆ The Rayleigh fading voltage is
1 γi
f (γi) = exp-  ( 8)
Γ Γ

◆ SNR γMRC is Gamma distributed (sum of M exponential RVs)

(γMRC)M-1 exp[-γMRC / Γ]
f (γMRC) = (9)
ΓM (M - 1)!
◆ The outage probability of MRC is then

1(γMRC)M-1 exp[-γMRC / Γ]
γ
Pr[γMRC < γ] =  M
ⅆ γMRC
Γ 0 Γ (M - 1)!
(10)
k-1
γ M (γ / Γ)
= 1 - exp-  
Γ k=1 (k - 1)!

Wassim Alexan 18
 Maximal Ratio Combining

Fig. 7. BER versus average output SNR for MRC.

◆ Fig. 7 illustrates the idea that increasing the number of diversity branches
(indicated here by N), leads to better BER performance of MRC

◆ The case of no fading indicates a LOS, whereby in MRC as N → ∞, a BER

performance similar to that of a LOS is attained

Wassim Alexan 19
 Equal Gain Combining

Fig. 8. Equal gain combining.

(Figure is copied from Prof. Tallal’s lecture slides, with his permission)

◆ MRC requires estimation of the channel coefficients across all the diversity
branches

◆ EGC conducts a sum across all branches (i.e. Gi = 1 for all i)

Wassim Alexan 20
 Equal Gain Combining
◆ Consider M independent Rayleigh fading channels available at the receiver

◆ Envelope applied to the receiver detector

M
rEGC =  ri (11)
i=1

◆ Total noise power applied to the receiver detector

NEGC = M N (12)

◆ SNR at the receiver detector

(rEGC)2 M 2
γEGC ∝ =   ri   M N (13)
NEGC i=1

◆ EGC is a special case of MRC with Gi = 1

◆ SNR and outage probability performance of EGC is inferior to that of MRC

Wassim Alexan 21
 Exercise 2
Consider a four–branch diversity system where each branch received an indepen-
dent Rayleigh fading signal.

Given that the instantaneous received power over the four branches respectively
are - 90 dBm, - 92 dBm, - 95 dBm and - 98 dBm.

Also assume that the interference is negligible and background noise is mea-
sured at - 98 dBm.

◆ (a) What is the SNR at the receiver detector if SC is used?

◆ (b) What is the SNR at the receiver detector if MRC is used?

◆ (c) What is the SNR at the receiver detector if EGC is used?

◆ (d) How many branches are needed to provide an outage probability of 4%

using selection combining given that the average received power on all
branches is - 92 dBm?

Wassim Alexan 22
 Exercise 2 Solution
◆ (a) What is the SNR at the receiver detector if selection combining is used?

γSC = γmax = - 90 - (- 98) = 6.31 = 8 dB

◆ (b) What is the SNR at the receiver detector if MRC is used?

γ1 = - 90 - (- 98) = 8 dB

γ2 = - 92 - (- 98) = 6 dB

γ3 = - 95 - (- 98) = 3 dB

γ4 = - 98 - (- 98) = 0 dB
M
γMRC =  γi = 100.8 + 100.6 + 100.3 + 1 = 13.286 = 11.23 dB
i=1

Wassim Alexan 23
 Exercise 2 Solution
◆ (c) What is the SNR at the receiver detector if EGC is used?

r1 = 10(-90-30)/10 = 1 × 10-6, r2 = 10(-92-30)/10 = 7.9433 × 10-7

r3 = 10(-95-30)/10 = 5.623 × 10-7, r4 = 10(-98-30)/10 = 3.98 × 10-7

M = 4 and N = 10(-98-30)/10 = 1.585 × 10-13

γEGC =
2
∑M
i=1 ri
= 1 × 10-6 + (7.9433 + 5.623 + 3.98) × 10-72  (4 × 1.585 × 10-13)
MN
= 11.97
= 10.781 dB

Wassim Alexan 24
 Exercise 2 Solution
◆ (d) How many branches are needed to provide an outage probability of 4%
using selection combining given that the average received power on all
branches is - 92 dBm?

Γ = - 92 - (- 98) = 6 dB

γth = γmin = 0 dB = 1 and PO = 0.04

γth M
Pr[γ1, γ2, ..., γM < γth] = 1 - exp-  = 0.04
Γ
log[0.04]
M= = 2.13947
log[0.222124]

◆ We round up to further reduce the probability of outage, thus, M = 3 branches

Wassim Alexan 25