Professional Documents
Culture Documents
Submitted by Supervisor
Kalpant Pathak Dr. Pravas Ranjan Sahu
Roll. No. 11410240
This is to certify that the work contained in this thesis entitled “Analysis of Switched
Diversity Receivers over κ-µ and η-µ Fading Channels” by Kalpant Pathak,
Roll no: 11410240, has been carried out at Department of Electronics & Electrical
Engineering, Indian Institute of Technology Guwahati under my supervision and that it
has not been submitted elsewhere for a degree.
i
Acknowledgement
ii
Abstract
iii
Contents
1 Introduction 1
1.1 Wireless Communication System . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Fading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 Statistical Models for Fading Channels . . . . . . . . . . . . . . . 4
1.2 Effect of Fading on System Performance . . . . . . . . . . . . . . . . . . 5
1.3 Diversity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.1 Diversity Combining Techniques . . . . . . . . . . . . . . . . . . . 8
3 BER Analysis 18
3.1 The κ-µ Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.2 The η-µ Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2.1 The η-µ Distribution: Format 1 . . . . . . . . . . . . . . . . . . . 19
3.2.2 The η-µ Distribution: Format 2 . . . . . . . . . . . . . . . . . . . 19
3.3 Probability of Error Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3.1 Moment Generating Function of Output SNR . . . . . . . . . . . 20
3.3.2 Probability of Bit Error Analysis . . . . . . . . . . . . . . . . . . 24
iv
5 Conclusion and Future Work 38
Appendices 39
v
List of Figures
4.1 BER performance of SSC receiver over κ-µ fading channel for BPSK and
BFSK modulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 BER performance of SSC receiver over η-µ fading channel for BPSK and
BFSK modulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
4.3 BER performance of SEC receiver over κ-µ fading channel for L=3 for
BPSK and BFSK modulations . . . . . . . . . . . . . . . . . . . . . . . . 29
4.4 BER performance of SEC receiver over η-µ fading channel for L=3 for
BPSK and BFSK modulations . . . . . . . . . . . . . . . . . . . . . . . . 30
4.5 BER performance of SEC receiver over κ-µ fading channel for L=5 for
BPSK and BFSK modulations . . . . . . . . . . . . . . . . . . . . . . . . 31
4.6 BER performance of SEC receiver over η-µ fading channel for L=5 for
BPSK and BFSK modulations . . . . . . . . . . . . . . . . . . . . . . . . 32
4.7 Comparison of BER performance over Rayleigh fading by varying γT . . 33
4.8 BER performance of SECps receiver over κ-µ fading channel for L=2 for
BPSK and BFSK modulations . . . . . . . . . . . . . . . . . . . . . . . . 35
4.9 BER performance of SECps receiver over η-µ fading channel for L=2 for
BPSK and BFSK modulations . . . . . . . . . . . . . . . . . . . . . . . . 36
vi
4.10 Comparison of BER performance of SC, SEC|L=2 /SSC and SECps com-
bining schemes over Rayleigh fading . . . . . . . . . . . . . . . . . . . . . 37
vii
Chapter 1
Introduction
In past few years, wireless communication has played an important role in information
technology as information can be transmitted without the need of dedicated link between
transmitter and receiver unlike wired communication, where a dedicated link/channel
exist between transmitter and receiver. Compared to wired communication systems,
wireless systems introduce a very interesting feature ‘mobility’.
In any kind of communication, wired or wireless, there are some parameters like
bandwidth, transmitted power, data rate etc. which decide the reliability of a system.
The one which optimizes all of them is said to be a perfect system. In recent years, lots
of research has been done on both kinds of communication so that a reliable system can
be designed with high bandwidth, low transmitted power, high data rates and low bit or
symbol error probability.
1
NLOS
LOS
NLOS
MS
BS
the direct path between the transmitter and the receiver is called line-of-sight (LOS) path,
whereas, the path corresponding to reflected signal is called non line-of-sight (NLOS)
path.
These multipaths have different phases corresponding to different path-delays, so that
they interfere at the receiver either constructively or destructively resulting in variation in
signal-to-noise ratio. In addition, mobility introduces time variation in channel response,
i.e. if a very short pulse is transmitted, the received signal appears as a train of pulses
due to presence of multipaths. Secondly, as a result of time varying response, if same
procedure is followed multiple times, a change is observed in the received pulse train
over time, which will include changes in the sizes of individual pulses, changes in relative
delays among the pulses and often, changes in the number of pulses observed.
Hence, the equivalent low-pass time varying impulse response of the channel can be
modeled as [1]:
∑
c(τ ; t) = αi (t)e−j2πfc τi (t) δ[t − τi (t)] (1.1)
i
where, αi (t) and τi (t) are time varying attenuation factor and path delay for ith path
respectively.
For a transmitted signal sl (t) = 1 the received signal for the case of discrete multipath
is given by [1]:
∑
rl (t) = αi (t)e−jθi (t)
i
2
random manner. So, received signal rl (t) can be modeled as a random process. For
large number of paths, central limit theorem can be applied and rl (t) can be modeled
as complex-valued Gaussian random process i.e. c(τ ; t) is also a complex-valued random
process in t variable [1].
1.1.1 Fading
Due to time varying nature of phases and amplitudes of the received signals, the
multipaths add constructively or destructively at the receiver resulting in fluctuation in
the received signal. This amplitude variations in the received signal is termed as fading.
There are some parameters which define the fading characteristics of the channel.
These parameters are:
Coherence Time
Coherence time, Tc of the channel measures the period of time over which two samples
of channel response taken at same frequency but at different time instants are correlated.
The coherence time is also related to Doppler spread, fd by [1]:
1
Tc ≃
fd
Coherence Bandwidth
Coherence bandwidth, fc of the channel measures the frequency range over which two
samples of channel response taken at same time instants but at different frequencies are
correlated. The coherence bandwidth is related to maximum delay spread, τmax by [1]:
1
fc ≃
τmax
Depending upon the parameters described above, fading can be classified into four
groups:
In slow fading, the symbol time duration Ts is smaller than the channel’s coherence
time, so that multiple symbols undergo same fading. Whereas, in fast fading, the symbol
time duration Ts is larger than the channel’s coherence time, hence, fading decorrelates
from symbol to symbol [2].
3
1.1.1.2 Frequency-Flat and Frequency-Selective Fading
Rayleigh model is used when there is no LOS component present in the received
signal. The probability density function (PDF) of the envelope α ≥ 0 of channel impulse
response is given as [2]:
2α − α2
pα (α) = e Ω, α≥0
Ω
where, Ω = E[α2 ] is the mean-square value of α.
Nakagami-n distribution is also called Rice distribution. This is suitable for the
propagation paths having one strong LOS path and many weaker NLOS paths. The pdf
is given as [2]:
( √ )
2(1 + n2 )e−n α − (1+n2 )α2
2
1 + n2
pα (α) = e Ω I0 2nα , α≥0
Ω Ω
4
where, 0 ≤ n ≤ ∞ is the Nakagami-n fading parameter and is related to Ricean factor
‘K’ by K=n2 . I0 (·) is bessel’s function of first kind and order zero [3].
Nakagami-n distribution includes the Rayleigh distribution as a special case (n=0)
and spans from Rayleigh fading (n=0) to no-fading (n=∞).
where, Γ(·) is gamma function [3] and 1/2 ≤ m ≤ ∞ is Nakagami-m fading parameter.
This fading model includes one-sided Gaussian (m=1/2), Rayleigh (m=1) fading and
no-fading (m=∞) as special cases. For m < 1, it closely approximates Hoyt distribution,
and this mapping is given by [2]:
(1 + q 2 )2
m= , m≤1
2(1 + 2q 4 )
Similarly, for m > 1, it closely approximates the Nakagami-n (Rice) distribution and this
mapping is one-to-one and given by [2]:
(1 + n2 )2
m= , n≥0
1 + 2n2
5
Performance of BPSK over fading and non−fading environments
Rayleigh Fading
AWGN Channel
−1
Probability of error, Pe 10
−2
10
−3
10
−4
10
0 5 10 15 20 25
Input average SNR (γ̄) in dB
and non-fading AWGN channel over BPSK modulation scheme. From the figure it can
be observed that, to have an BER of 10−3 , SNR needed for non-fading channel is ≃ 7dB,
where as for Rayleigh faded channel, it is ≃ 23dB. In such case, the transmitted power
for Rayleigh faded channel will be ≈ 40 times that of AWGN channel.
To mitigate the effects of fading on the system and to improve performance of the
system, Diversity techniques are used at the receiver end.
1.3 Diversity
In a typical mobile radio environment, it is observed that if two or more radio channels
are sufficiently separated in frequency, time or space, then the fading in various channels
is more or less independent. Diversity, is the key technology that uses this independence
of channels to mitigate the effect of fading [4]. The main object of diversity techniques is
to use several replicas of transmitted signal to improve system performance. There are
various ways to obtain independently faded channels:
6
Receiving Antenna
1
2
Combiner
(MRC/EGC Output
/SC/SSC)
• Time Diversity: Same message is transmitted over different time slots such that
the time separation between adjacent transmissions should be greater than channel’s
coherence time.
The multiple replicas of the transmitted signal obtained using any of the method
mentioned above, is then combined efficiently to improve the SNR at the detector input.
Figure 1.3 shows a space diversity technique with a combiner, which combines the multiple
signals received on different antennas. There are different types of diversity combining
techniques, which are discussed in next section:
7
G1
1
g1
G2
2
g2 Detector
Sum
N
gN
GN
Adaptive Control
There are different types of diversity combining techniques used in practice [2], which
are as follows:
In maximal ratio combining technique, the received multiple faded copies of the trans-
mitted signal are co-phased. The co-phased signal copies are weighted individually in
proportion to their strength to maximize SNR at the output of the combiner. Assuming
the received signal SNR at the input of the combiner is γi , i = 1, 2, ..N, the output SNR
can be shown to be [5]:
∑
N
γM RC = γi
i=1
The MRC operation requires estimation of phase and amplitude of each received input
branch signal. Hence, the complexity of implementation is high.
Different weights for each branch may not be convenient as it may increase the com-
plexity of the receiver as in the case of MRC. So it is convenient to set all the gains to
unity, while cophasing all signals before combining [2]. This technique of combining is
called Equal Gain Combining.
8
1
g1
2
Co-Phased Detector
g2 And
Sum
N
gN
Adaptive Control
where, αi is the fading amplitude for ith copy of the transmitted signal.
In selection combining (SC), the system chooses the received signal having maximum
SNR out of all copies of signals received. In this scheme the output SNR can be given as
[2]:
γSC = max{γ1 , γ2 , ..., γN }
The switch and stay combining (SSC) technique discussed here is presented in [2]
and also shown in Figure 1.7. In this system, there are only two copies of fading signals
are used. The combiner has only two antennas to receive fading signals. The received
signal is fed as shown in Figure 1.7. In this scheme the received SNR γ1 at antenna L1
is compared with a predefined threshold γT . Switching occurs to the input branch L2
if γ1 < γT . And it again switches to first branch if γ1 > γT . It may happen that after
switching the input SNR γ2 at L2 is less than γT or even less than γ1 , in such case the
switch will still be connected to L2 until the SNR of first branch becomes greater than
γT . Switching from branch L2 to branch L1 is done in similar manner.
9
1
g1
2 Select
g 2 Maximum Detector
of N
N branches
gN
L1 g 1 L2 g 2
Control Switch
Logic
Comparator
Present
Receiver Estimator Threshold
Data
Figure 1.7: Block diagram of Dual branch Switch and Stay Combiner
Unlike SSC combining scheme, switch and examine combining (SEC) adds the ben-
efit of having multiple branches at the receiver, especially when they are independent
and identically distributed (i.i.d.) or equicorrelated and identically distributed. In SSC
scheme, receiver switches between the best two paths, adding a path does not improve
the performance unless the added path is better than at least one of the best two ones.
10
In SEC combining scheme, the receiver starts examining from the first path. If first
path is acceptable, it continues to receiver from it, else, it switches and examines the
next available path. This process continues until an acceptable path is found or all paths
have been examined. In the latter case, the receiver stays on the last examined path [6]
or selects the best path for reception [7].
11
Chapter 2
A number of works have been reported on the analysis of switched diversity schemes
over different fading channel models. Selection combining, switch-and-stay combining
and multibranch switch-and-examine combining schemes have been extensively studied
over different fading channels. Also, different switching algorithms have been suggested
to improve the system performance. Different research works regarding various switched
combining schemes are listed in following section:
• D. G. Brennan in [4] has analyzed the SC, MRC and EGC combining schemes.
In this paper, some departures from ideal conditions, such as non-Rayleigh fading
and partially coherent signals are also considered. Also, merits of predetection
and postdetection combining is discussed. These schemes have been analyzed over
Rayleigh fading channel and performance comparison has been presented.
• M. A. Blanco in [8], has studied the performance of non coherent binary frequency
shift keying (NC-BFSK) with dual selection combiner over independent and iden-
tically distributed Nakagami-m fading channels.
• T. Eng et. al. in [9], has compared different diversity combining techniques over
Rayleigh fading channels. In this paper, for coherent reception, the authors have
compared performance of MRC, SC and a generalization of SC, where two/three
12
signals with largest amplitudes are coherently combined and these techniques are
denoted by second/third order SC (SC2 and SC3). Also, same techniques are inves-
tigated for non coherent reception, with EGC and non coherent SC2 and SC3. It
has been shown that SC2 and SC3 techniques have better performance than conven-
tional SC scheme and under certain conditions they may approach the performance
of EGC or MRC. Also, for non coherent reception it is shown that SC2 and SC3
perform better than EGC for higher BER values.
• In year 1997, O. C. Ugweje and V. A. Aalo in [10], have analyzed the selection com-
bining scheme over correlated Nakagami fading channels and the effect of branch
correlation and fading parameter on the performance of the system has been stud-
ied. Also, the joint distribution of the combiner output is obtained and analysis
has been done for both coherent and non coherent demodulation schemes. Differ-
ent graphs have also been shown to understand the effect of fading parameter and
branch correlation. In the paper, it has been concluded that for dual selection di-
versity system, branch correlation does not influence the BER significantly whereas,
as Nakagami fading parameter m is increased, the BER performance improves sig-
nificantly at higher SNRs.
• In [11], M. K. Simon and M. S. Alouini have analyzed dual selection combining over
correlated Rayleigh and Nakagami-m fading channels. Also, they have presented
the expressions for the outage probability and average error probability performance
of dual SC receiver over correlated Rayleigh fading in closed form, in particular for
binary differential phase shift keying (BDPSK) and the results are also extended
for Nakagami-m fading.
Switch and stay combining (SSC) has been a popular combining scheme as it is the
least complex combining scheme among all combining schemes. A number of works have
been reported on the analysis of SSC scheme. Some of them are listed below:
• In [12], A. A. Abu-Dayya and N. C. Beaulieu have studied the SSC scheme over
independent and correlated Ricean fading channels using a discrete time model.
Also, in the paper, BER for non coherent binary frequency shift keying is derived.
13
In addition, an optimum switching threshold, which gives minimum BER, has also
been obtained for independent fading case. The obtained performance is then com-
pared to conventional selection combining scheme and effect of fading severity on
BER and optimum threshold is investigated. For independent channels, it has been
shown that for given Rice parameter K, optimum threshold increases with the SNR
and for fixed SNR, optimum threshold increases with K.
• In [14], Y. C. Ko et. al. has analyzed dual SSC scheme over Rayleigh, Nakagami-m
and Nakagami-n fading channels using moment generating function (MGF) ap-
proach. In this paper, effect of unequal average SNR, branch correlation and im-
perfect channel estimation has also been studied. Also, the authors have derived
the closed form expression for MGF of SSC output SNR, γSSC considering all three
fading channels and using these MGF expressions, closed form expressions of prob-
ability of error has been obtained.
Unlike SSC, SEC adds the benefit of having multiple branches at the receiver and
hence, performs better than SSC. Some important works reported on the performance
analysis of conventional SEC, modified SEC and generalized SEC are listed below:
• In [6], H. Yang and M. S. Alouini have analyzed both SSC and SEC diversity com-
bining schemes over Rayleigh fading channels. This work presents generic formulas
for cumulative distribution function (CDF), probability density function (PDF)
and moment generating function (MGF) of combiner output. Using these formulas,
closed-form expressions for outage probability and average bit error rate have been
obtained and a comparison between the two schemes has been shown. It is observed
that, for increasing number of branches, SSC does not improve the performance as
14
long as the branches are identically distributed, whereas, SEC improves the BER
with the increase in number of branches.
• In [15], G. C. Alexandropoulos et. al. has studied the performance of SEC diver-
sity scheme over arbitrarily correlated and not necessarily identically distributed
Nakagami-m fading channels. The authors have obtained analytical expressions
for distribution of SEC output SNR for constant correlation model. For integer
and half integer values of m, under the assumption of most general case of cor-
relation, analytical expressions for the distribution of output SNR is derived for
L ≤ 3. Whereas, for L > 3 analytical approximations are presented. Also, with the
help of graphs, it is shown that with increasing branch correlation, the performance
degrades.
• In 2006, in [7], H. C. Yang and M. S. Alouini has presented a new kind of switch-
and-examine combining known as SEC with post examining selection (SECps) and
analyzed its performance over i.i.d. Rayleigh fading channels. In this technique,
when no diversity path is acceptable, the receiver selects the best diversity branch
for reception instead of choosing it randomly as in the case of conventional SEC.
Using such combining schemes, author have shown that the performance can be
improved. Authors have also shown that, using appropriate switching threshold
SECps can have same error performance as that of SC with much less path estima-
tions on average.
In past few years, some more generalized fading distributions such as κ-µ and η-µ distri-
butions have also been developed and performance of wireless communication systems is
also analyzed over these channels. In the following section, some of the research works
done on such fading channels are presented:
15
2.2 κ-µ and η-µ Fading Channels
The diversity combining schemes have been extensively studied over some general
fading distributions such as Rayleigh, Ricean, Nakagami-m, and Weibull etc. In 2007, in
[17], M. D. Yacoub has suggested κ-µ and η-µ fading channels to model the mobile radio
environment more accurately. These fading distributions are used to model nonhomoge-
neous fading environment. The research works done on κ-µ and η-µ fading channels are
listed below:
• In [17], M. D. Yacoub has presented the κ-µ and η-µ fading channels. These fading
models considers a signal composed of clusters of multipath waves. Within any one
cluster, the phases of scattered waves are random and have similar delay times with
delay-time spreads of different clusters being relatively large.
In the case of κ-µ fading model, all scattered components within each cluster have
identical powers but within each cluster, a dominant component is found. Whereas,
the η-µ fading distribution has two formats. In format-1, it is assumed that, within
each cluster the in-phase and quadrature components of scattered waves are inde-
pendent and have different powers, while in format-2, the in-phase and quadrature
components of scattered waves have identical powers but they are correlated with
each other. In the paper, it is shown that these fading models also accommodate
Hoyt, Nakagami-m, Rayleigh and Ricean distribution as their special cases.
In the paper, the author has presented the derivations of PDF and CDF for both
κ-µ and η-µ distributions. Also, author stated the usability of these fading mod-
els. The κ-µ distribution is suited for line-of-sight propagation model and includes
Rayleigh, Ricean and Nakagami-m as special cases, whereas the η-µ distribution
is best suited for non line-of-sight propagation model and includes Rayleigh, Hoyt
and Nakagami-q as special cases. Further, the author has also presented various
estimators for fading parameter estimation.
The results presented in the paper are also field verified and it is observed that in
a wide sense, Nakagami-m can be thought of as a mean distribution of κ-µ and η-µ
distributions.
• In [18], N. Y. Ermolova has derived the closed form expressions for MGF of η-
µ and κ-µ fading distributions. Also, using these expressions, BER is evaluated
16
numerically for BPSK modulation scheme.
• In [19], N. Y. Ermolova has used the derived MGF expressions in [18] to evaluate
certain integrals useful in obtaining the BER expressions for BPSK and rectangular
quadrature amplitude modulation (QAM) signalling schemes. Also, the effect of
fading parameters on the system is studied.
• In [20], authors have analyzed the L-branch SC receiver over κ-µ and η-µ fading
channels. Authors have analytically obtained the moments of SC receiver output
SNR and bit error rate for binary, coherent and non-coherent modulation schemes.
Also, the numerical results are verified by comparing them with simulation results.
• In [21], authors have studied the performance of L-branch MRC receiver over κ-µ
and η-µ fading channels. Authors have derived a highly accurate approximation
to the average symbol error rate (ASER) expression for QAM modulation scheme.
Also, the results are verified using computer simulations.
2.3 Motivation
Switched diversity combining schemes such as SSC, SEC and SECps are less complex
diversity combining schemes as they don’t require channel estimation at the receiver and
also, they reduce the switching rate required among the available diversity branches.
Though these diversity combining schemes have already been examined over different
fading channels including Rayleigh, Ricean, Nakagami [2], Weibull [16] etc., analysis over
nonhomogeneous fading distributions such as κ-µ and η-µ is not available in literature.
In this thesis work, we have analyzed SSC, SEC and SECps diversity combining
schemes over κ-µ and η-µ fading channels for BPSK and coherent BFSK modulation
schemes and studied the effect of channel parameters on bit error rate performance of the
system.
17
Chapter 3
BER Analysis
For bit error rate analysis of SSC, SEC and SECps combining schemes, the channel is
assumed to be slow and flat fading with κ-µ or η-µ distribution. For transmitted signal
s(t) with symbol energy Es , the complex low pass equivalent of the received signal at ith
(i=1,2...L) path over the symbol duration Ts second can be given as ri (t) = αi ejϕi s(t) +
ni (t), where ni (t) is the complex additive white Gaussian noise (AWGN) having two
sided power spectral density N0 /2, random variable (RV) ϕi is instantaneous phase and
RV αi is fading envelope which is either κ-µ or η-µ distributed. The PDFs of κ-µ and
η-µ distributed RVs are given in following sections.
where Ωi = E[αi2 ], E[·] is the expectation operator, κ > 0 and µ > 0 are the parameters
of the distribution and Iv (·) is the modified Bessel function of the first kind and v th
order [3]. The parameter κ is the ratio of the power due to dominant components to the
total power due to scattered components and µ is the number of multipath clusters. This
distribution includes Rice (µ=1 and κ=K), Nakagami-m (κ→0 and µ=m), Rayleigh (µ=1
and κ→0) and one sided Gaussian distribution (µ=0.5 and κ→0) fading distributions as
special cases. This distribution is better suited for line-of-sight propagation.
18
3.2 The η-µ Distribution
The probability density function (PDF) of η-µ distributed RV is given as [17]
√ 1 ( )
4 πµµ+ 2 hµ αi2µ − 2µh α2i 2µH 2
pαη−µ (αi ) = e Ωi
Iµ− 1 α , (3.2)
1 µ+ 1
Γ(µ)H µ− 2 Ωi 2
2 Ωi i
where Γ(·) is the gamma function, and h and H are functions of the parameter η defined
for two formats in next subsections. µ denotes the number of multipath clusters. This
fading distribution includes Hoyt (η=q 2 ,µ=0.5), Nakagami-m (η=1,µ=m/2), Rayleigh
and one sided Gaussian distribution as special cases.
In this format 0 < η < ∞ is the power ratio of the in-phase and quadrature compo-
2+η −1 +η η −1 −η
nents of the scattered-waves of each multipath. In this case, h = 4
and H = 4
.
Within 0 < η ≤ 1, we have H ≥ 0, on the other hand, within 0 < η −1 ≤ 1, we have H ≤ 0.
Because Iv (−z) = (−1)v Iv (z), the distribution is symmetrical around η=1. Therefore, as
far as the envelope(or power) distribution is concerned, it is sufficient to consider η only
within one of these ranges. In Format 1, H/h = (1 − η)/(1 + η) [17].
In this format −1 < η < 1 is the correlation coefficient between the in-phase and
1
quadrature components of the scattered waves of each cluster. In this case, h = 1−η 2
and
H= η
1−η 2
. Within 0 ≤ η < 1, we have H ≥ 0, on the other hand, within −1 < η ≤ 0, we
have H ≤ 0. Because Iv (−z) = (−1)v Iv (z), the distribution is symmetrical around η=0.
Therefore, as far as the envelope(or power) distribution is concerned, it is sufficient to
consider η only within one of these ranges. In Format 2, H/h = η [17].
19
where Mγ (·) is the MGF of the receiver output SNR, ψ is the modulation parameter i.e.,
ψ=1 for BPSK, ψ=0.5 for BFSK, and ψ=0.715 for BFSK with minimum correlation [22].
Thus, for the BER analysis of the SSC, SEC and SECps receivers under consideration in
κ-µ / η-µ fading channels applying Equation 3.3, we need to obtain the MGF of γSSC ,
γSEC and γSECps i.e. MγSSC (s), MγSEC (s) and MγSECps (s). Expressions for the MGFs are
derived in the following subsection.
where, p(x) is the PDF of RV x. The MGF of the output SNR of SSC, SEC and SECps
combining schemes is derived in following subsections.
If γSSC denotes the SNR per symbol of the combiner output and γT denotes the
predetermined switching threshold, to derive the MGF of SSC output SNR, we first
derive the CDF of the output SNR, PγSSC (γ) in terms of CDF of individual branch SNR,
Pγ (γ) as [2]
P (γ )P (γ), γ < γT
γ T γ
PγSSC (γ) = (3.4)
P (γ) − P (γ ) + P (γ)P (γ ), γ ≥ γ
γ γ T γ γ T T
Differentiating PγSSC (γ) with respect to γ, we get the PDF of SSC output SNR in terms
of CDF {Pγ (γ)} and the PDF {pγ (γ)} of the individual branch SNR as [2]
P (γ )p (γ), γ < γT
γ T γ
pγSSC (γ) = (3.5)
(1 + P (γ ))p (γ), γ ≥ γT
γ T γ
Using the above PDF, an expression for the MGF of γSSC in a fading channel can be
obtained from the formula [2]
∫∞
MγSSC (s) = Pγ (γT )Mγ (s) + pγ (γ)e−sγ dγ , (3.6)
γT
| {z }
I(s)
20
where Pγ (γT ) is the cumulative distribution function (CDF) of κ-µ / η-µ fading distribu-
Es 2
tion and pγ (γ) is the probability density function (PDF) of the SNR γ = N0
α .
The CDF of the output SNR of SEC receiver for L i.i.d branches, is given as [2]
[P (γ )]L−1 P (γ), γ < γT
γ T γ
PγSEC (γ) = ∑ (3.7)
L−1 [P (γ )]j [P (γ) − P (γ )] + [P (γ )]L , γ ≥ γT
j=0 γ T γ γ t γ T
Differentiating PγSEC (γ) with respect to γ, we get the PDF of SEC output SNR in
terms of CDF {Pγ (γ)} and the PDF {pγ (γ)} of the individual branch SNR as [2]
[P (γ )]L−1 p (γ), γ < γT
γ T γ
pγSEC (γ) = ∑ (3.8)
L−1 [P (γ )]j p (γ), γ ≥ γT
j=0 γ T γ
using the above PDF, an expression for the MGF of γSEC for L-branch SEC in a fading
channel can be obtained from the formula [2]
∑
L−2 ∫ ∞
L−1
MγSEC (s) = [Pγ (γT )] Mγ (s) + j
[Pγ (γT )] pγ (γ)e−sγ dγ , (3.9)
γT
j=0 | {z }
I(s)
where L is the number of diversity branches, Pγ (γT ) is the cumulative distribution func-
tion (CDF) of κ-µ / η-µ fading distribution and pγ (γ) is the probability density function
Es 2
(PDF) of the SNR γ = N0
α .
The CDF of the output SNR of L-branch SECps receiver in terms of PDF of individual
branch SNR is given as [2]
[P (γ)]L , γ < γT
γ
PγSECps (γ) = ∑ (3.10)
1 − L−1 [P (γ )]j [1 − P (γ)], γ ≥ γ
j=0 γ T γ T
Differentiating Equation 3.10 with respect to γ, we can obtain the PDF of SECps output
SNR in terms of CDF {Pγ (γ)} and the PDF {pγ (γ)} of the individual branch SNR as [2]
21
L[P (γ)]L−1 p (γ), γ < γT
γ γ
pγSECps (γ) = ∑ (3.11)
L−1 [P (γ )]j p (γ), γ ≥ γ
j=0 γ T γ T
Using the above PDF, an expression for the MGF of γSECps for L-branch SECps in a
fading channel can be obtained
∫ γT ∑
L−1 ∫ ∞
−sγ
MγSECps (s) = L L−1
[Pγ (γ)] pγ (γ)e dγ + [Pγ (γT )]j
pγ (γ)e−sγ dγ , (3.12)
|0 {z } j=0 | γT {z }
I1 (s) I(s)
where L is the number of diversity branches, Pγ (γ) is the cumulative distribution func-
tion (CDF) of κ-µ / η-µ fading distribution and pγ (γ) is the probability density function
Es 2
(PDF) of the SNR γ = N0
α .
In order to findout the MGF of ouput SNR for SSC, SEC and SECps diversity com-
bining we need to find the Pγ (γT ), Mγ (s) and I(s) in Equations 3.6 and 3.9. These
quantities are derived below:
For κ-µ and η-µ fading scenarios an expression for pγ (γ) is given as [18]
( ) µ+1 ( √ )
µ 1+κ 2 ( ) µ−1
γ 2 − µ(1+κ) κ(1 + κ)γ
pγκ−µ (γ) = µκ e γ̄ γ Iµ−1 2µ (3.13)
e γ̄ κ γ̄
√ ( )µ+ 12 ( ) 1 ( )
2 πhµ µ γ µ− 2 − 2µhγ 2µHγ
pγη−µ (γ) = e γ̄ Iµ− 1 (3.14)
Γ(µ) γ̄ H 2 γ̄
Further, an expression for Pγ (γT ) (CDF) of κ-µ distribution is given as [17]
( √ )
√ 2(1 + κ)µγT
Pγκ−µ (γT ) = 1 − Qµ 2κµ, , (3.15)
γ̄
Yv (a, b) = v− 12
x e Iv− 1 (ax2 )dx.
a Γ(v) b
2
22
with −1 < a < 1 and b ≥ 0.
The MGF of individual branch SNR, Mγ (s) for κ-µ and η-µ distribution is given as
[18]
( )µ ( )
µ(1 + κ) µ2 κ(1 + κ)
Mγκ−µ (s) = exp − µκ (3.17)
µ(1 + κ) + sγ̄ µ(1 + κ) + sγ̄
( )µ
4µ2 h
Mγη−µ (s) = (3.18)
(2(h − H)µ + sγ̄)(2(h + H)µ + sγ̄)
To obtain an expression for MγSSC (s) in Equation 3.6, MγSEC (s) in Equation 3.9 and
MγSECps (s) in Equation 3.12 we need to solve the second term integral I(s). A solution
for I(s) has been obtained in Appendix A, which is given as below.
∑∞ ( ( ) )
µ2n+µ κn (1 + κ)n+µ µ(1 + κ)
Iκ−µ (s) = κµ [µ(1 + κ) + sγ̄]n+µ
Γ n + µ, + s γT , (3.19)
n=0
n!Γ(n + µ)e γ̄
√ ( ( ) )
2 πhµ ∑
∞
µ2n+2µ H 2n 2µh
Iη−µ (s) = Γ 2n + 2µ, + s γT ,
Γ(µ) n=0 n!Γ(n + µ + 0.5)(γ̄s + 2µh)2n+2µ γ̄
(3.20)
where, Gm,n
p,q [·] is the Meijer’s G-function [24].
23
3.3.2 Probability of Bit Error Analysis
The probability of error can be obtained by using the derived MγSSC (s), MγSEC (s)
and MγSECps (s) in Equation 3.3.
The expressions of probability of error, Pe for SSC and SEC combining schemes are
given below.
∫π/2 ( ) ∫π/2 ( )
Pγ (γT ) ψ 1 ψ
Pe |SSC = Mγ dθ + I dθ (3.22)
π sin2 θ π sin2 θ
0 0
∫π/2 ( ) ∑
L−2 ∫π/2 ( )
[Pγ (γT )]L−1 ψ 1 ψ
Pe |SEC = Mγ dθ + [Pγ (γT )] ·
j
I dθ (3.23)
π sin2 θ j=0
π sin2
θ
0 0
1
∫ π/2 ( ψ
)
The first term in Equations 3.22 and 3.23, i.e. π 0
Mγ sin2 θ
dθ represents the proba-
bility of error with no-diversity and can be solved in a closed form as [19]
√
Γ(µ + 21 )(µ(1 + κ))µ ψγ̄
peκ−µ = √ 1
2 πeµκ Γ(µ + 1)[µ(1 + κ) + ψγ̄]µ+ 2
( )
1 µ(1 + κ) µ2 κ(1 + κ)
× Φ1 µ + , 1; µ + 1; , (3.24)
2 µ(1 + κ) + ψγ̄ µ(1 + κ) + ψγ̄
where Φ1 (·) is a confluent hypergeometric function of two variables, which can be imple-
mented by using its finite integral representation [25]
∫ 1
Γ(c)
Φ1 (a, b; c; x1 , x2 ) = eux2 ua−1 (1 − u)c−a−1 (1 − ux1 )−b du
Γ(a)Γ(c − a) 0
and
( )µ
Γ(2µ + 21 ) 4µ2 h(2(h + H)µ + ψγ̄)−1
peη−µ = √
2 πΓ(2µ + 1) (2(h − H)µ + ψγ̄)
( )
1 2(h − H)µ 2(h + H)µ
× F1 , µ, µ; 2µ + 1; , (3.25)
2 2(h − H)µ + ψγ̄ 2(h + H)µ + ψγ̄
where F1 (·) is an Appell’s hypergeometric function [25] that has a finite integral repre-
sentation
∫ 1 ∏
2
Γ(c)
F1 (a, b1 , b2 ; c; x1 , x2 ) = u a−1
(1 − u) c−a−1
(1 − uxi )−bi du
Γ(a)Γ(c − a) 0 i=1
while the evaluation of the second term of Equations 3.22 and 3.23 requires a numerical
integration.
24
3.3.2.2 SECps combining
The expressions of probability of error, Pe for SECps combining scheme is given below.
∫ π/2 ( ) ∑
L−1 ∫ π/2 ( )
L ψ 1 ψ
Pe |SECps = I1 dθ + [Pγ (γT )] ·
j
I dθ (3.26)
π 0 sin2 θ j=0
π 0 sin2 θ
The numerical integration of the above equation is quite complex. So, simulation results
of BER analysis of SECps over κ-µ and η-µ fading channels are obtained and plotted in
next chapter.
25
Chapter 4
The derived expression for Pe in Equation 3.22 is numerically evaluated for SSC
combining and is plotted in Figure 4.1 for BPSK (ψ=1) and BFSK (ψ=0.5) signaling
over κ-µ fading channels. The graph is plotted with respect to average SNR per input
branch (γ̄) for γT =10 dB and different values of κ and µ. In the figure, the curves for
κ=0 correspond to Nakagami-m fading with µ=m [17]. Thus, the curve for κ=0 and µ=1
is for the well-known Rayleigh fading distribution [17]. Also, the curves for µ=1 is for
Rice distribution with κ as rice parameter K [17]. It can be observed from the figure that
for a given µ, Pe improves with increase in κ. It is because of the increase in the power
of the dominant components of the waves over the scattered components of waves of the
fading model. Further, comparing the BER for BPSK and BFSK it can be observed that
BFSK has a poor performance (≃3 dB) as expected.
In Figure 4.2, BER is numerically evaluated for SSC combining and is plotted for
BPSK and BFSK signalling over η-µ fading channels. In the figure, the curves for η=1
correspond to Nakagami-m fading with µ=m/2 [17]. Thus, the curve for η=1 and µ=0.5
is for the well-known Rayleigh fading distribution [17]. Also, curves for µ=0.5 correspond
to Hoyt distribution with Hoyt parameter q 2 =η [17]. For a given µ, Pe improves with
increase in η.
Figures 4.3 and 4.4 show the BER performance of SEC combining for κ-µ and η-
µ distributions, respectively, for L=3. It can be observed that BER performance has
improved over dual-SSC receiver as three branches are involved for receiving the signal.
From the Figures 4.1 and 4.3, it can be observed that for BPSK over Rayleigh fading,
to have an BER of ≈ 10−2 , an average SNR of 11.3dB is required in SSC receiver, while
26
γT = 10 dB κ=0
κ=1
BPSK (ψ=1)
κ=2
Simulation
e
−1
10
Probability of error, P
Rayleigh
µ=1
−2
10
µ=2
µ=4
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
γ = 10 dB κ=0
T κ=1
BFSK (ψ=0.5)
κ=2
Simulation
−1
e
10
Probability of error, P
Rayleigh
µ=1
−2
10
µ=2
µ=4
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
Figure 4.1: BER performance of SSC receiver over κ-µ fading channel for BPSK and
BFSK modulations
27
γT = 10 dB η=0.1
η=0.3
BPSK (ψ=1) η=1
Simulation
e
−1
10
Probability of error, P
Rayleigh
µ = 0.5
−2
10
µ=1
µ=2
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
γT = 10 dB η=0.1
η=0.3
BFSK (ψ=0.5)
η=1
Simulation
e
−1
10
Probability of error, P
Rayleigh
µ = 0.5
µ=1
−2
10
µ=2
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
Figure 4.2: BER performance of SSC receiver over η-µ fading channel for BPSK and
BFSK modulations
28
γ = 10 dB κ=0
T
BPSK (ψ=1) κ=1
κ=2
Simulation
Probability of error, Pe
−1
10
Rayleigh
µ=1
−2
10
µ=2
µ=4
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
γT = 10 dB κ=0
κ=1
BFSK (ψ=0.5)
κ=2
Simulation
Probability of error, Pe
−1
10
Rayleigh
µ=1
−2
10
µ=2
µ=4
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
Figure 4.3: BER performance of SEC receiver over κ-µ fading channel for L=3 for BPSK
and BFSK modulations
29
γ = 10 dB η=0.1
T
η=0.3
BPSK (ψ=1)
η=1
Simulation
e
−1
10
Probability of error, P
Rayleigh
µ = 0.5
−2
µ=1
10
µ=2
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
γT = 10 dB η=0.1
η=0.3
BFSK (ψ=0.5)
η=1
Simulation
e
−1
Probability of error, P
10
Rayleigh
µ = 0.5
−2
µ=1
10
µ=2
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
Figure 4.4: BER performance of SEC receiver over η-µ fading channel for L=3 for BPSK
and BFSK modulations
30
γ = 10 dB κ=0
T
BPSK (ψ=1) κ=1
κ=2
Simulation
e
−1
Probability of error, P
10
Rayleigh
µ=1
−2
10
µ=2
µ=4
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
γ = 10 dB κ=0
T
BFSK (ψ=0.5) κ=1
κ=2
Simulation
e
−1
Probability of error, P
10
Rayleigh
−2
10
µ=1
µ=2
µ=4
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
Figure 4.5: BER performance of SEC receiver over κ-µ fading channel for L=5 for BPSK
and BFSK modulations
31
γT = 10 dB η=0.1
BPSK (ψ=1) η=0.3
η=1
Simulation
e
−1
Probability of error, P
10
Rayleigh
µ = 0.5
−2
10
µ=1
µ=2
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
γT = 10 dB η=0.1
BFSK (ψ=0.5) η=0.3
η=1
Simulation
e
−1
10
Probability of error, P
Rayleigh
µ = 0.5
−2
10
µ=1
µ=2
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
Figure 4.6: BER performance of SEC receiver over η-µ fading channel for L=5 for BPSK
and BFSK modulations
32
0
10
γ = 10 dB
T
−2
10
γT = 5 dB
−3
10
−4
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
10dB is required in SEC receiver with L=3, i.e. SEC receiver with L=3 has better
performance (≃1.3dB) than dual-SSC receiver. Similarly, Figures 4.5 and 4.6 show the
BER performance of SEC receiver over generalized fading channels for L=5. Again,
improvement can be observed for SECL=5 over SECL=3 and SSC. It can be observed from
the Figure 4.5, for BPSK over rayleigh fading the required SNR to have an BER of ≈ 10−2
is now reduced to 8.5dB, which is 1.5dB below than that of the SECL=3 and 2.8dB below
than that of the dual-SSC receiver. Similarly, for η-µ fading channels, improvement in
BER can be observed with increasing number of branches. These results are also verified
for special cases of κ-µ and η-µ distributions using simulation.
In Figure 4.7, comparison of BER improvement is done for different combining schemes
over Rayleigh fading channels for different values of γT . It can be observed that, in SSC
receiver for BPSK over Rayleigh fading with γT =5dB, BER of 10−2 can be observed
at γ̄=9dB, while with γT =10dB, same BER can be observed at γ̄=11.3dB. Similarly, for
SECL=5 receiver with γT =5dB, γ̄=5dB is required for BER ≈ 10−2 , while with γT =10dB,
γ̄=8.5dB is required for the same BER. This is because for higher γT probability of
switching to the next branch reduces and most of the time switch will be connected to
33
the first branch and for lower values of γT , most of the time switch will rest on the last
branch.
Similarly, Figures 4.8 and 4.9 show the BER performance of SECps combining for
κ-µ and η-µ distributions, respectively, for L=2 and γT =2dB. From the figures, it is
observed that the performance is better than SSC and SEC|L=2 . For γT =2dB, to have an
BER of 10−2 , the SNR required for SECps is ≈ 7.5dB, whereas, ≈ 8.5dB is required for
conventional SEC|L=2 , i.e. SECps has 1dB better performance than that of conventional
SEC|L=2 or SSC over Rayleigh fading channel. Further, it can be observed that BFSK has
3dB poor performance as compared to BPSK as expected. Again, the effect of channel
parameters on BER performance can be observed, i.e. for fixed µ, increasing κ or η gives
better BER performance as expected. Also, for fixed κ or η, increasing µ improves the
BER performance.
In Figure 4.10, it is shown that for γ̄ << γT , as both the diversity branches will
be below threshold, the SECps performs same as conventional selection combining (SC),
whereas for γ̄ >> γT , the average SNR of the current branch will fall below the threshold
very occasionally, so in such case it performs as no-diversity has been used. For medium
range of average SNR, the SECps has almost same performance as that of conventional
SEC.
34
0
10
κ=0
γT = 2 dB κ=1
BPSK
Probability of error, Pe
−1
10
Rayleigh
−2
10
µ=1
µ=2
−3
µ=4
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
0
10
κ=0
γ = 2 dB
T κ=1
BFSK
Probability of error, Pe
−1
10
Rayleigh
−2
10
µ=1
µ=2
µ=4
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
Figure 4.8: BER performance of SECps receiver over κ-µ fading channel for L=2 for
BPSK and BFSK modulations
35
0
10
η=0.1
γ = 2 dB η=0.3
T
BPSK
Probability of error, Pe
−1
10
−2
10
µ=2 µ = 0.5
µ=1
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
0
10
η=0.1
γT = 2 dB η=0.3
BFSK
Probability of error, Pe
−1
10
µ = 0.5
−2
10
µ=1
µ=2
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
Figure 4.9: BER performance of SECps receiver over η-µ fading channel for L=2 for
BPSK and BFSK modulations
36
SEC−ps(L=2)
BPSK (ψ = 1) SEC(L=2)/SSC
Rayleigh Channel SC(L=2)
γT = 2 dB
−1
Probability of error, Pe
10
−2
10
−3
10
0 2 4 6 8 10 12 14 16 18 20
γ̄ in dB
Figure 4.10: Comparison of BER performance of SC, SEC|L=2 /SSC and SECps combining
schemes over Rayleigh fading
37
Chapter 5
The BER performance of switch and stay combining (SSC), switch and examine com-
bining (SEC) and switch and examine combining with post-examining selection (SECps)
over κ-µ and η-µ fading channels for BPSK and BFSK modulation schemes using MGF
based approach has been presented. Mathematical expressions for the MGF of SSC and
SEC output SNR has been derived for both fading distributions and MGF of SECps
output SNR has been derived for κ-µ fading channels, and using it numerical results for
the BER has been obtained.
The numerically evaluated BER results are plotted for different parameters, and the
results are compared with simulation results for the special cases of κ-µ and η-µ distri-
butions for the system under consideration.
The presented work can be extended for generalized switch and examine combin-
ing (G-SEC) and other antenna selection techniques over other non-homogeneous fading
channels too. The G-SEC combining schemes have been studied extensively for Rayleigh,
Ricean and Nakagami fading channels. However, these schemes have not been analyzed
over non-homogeneous fading channels such as κ-µ and η-µ fading channels.
38
Appendix A
where
µ+1 µ−1
µ(1 + κ) 2 γ 2
β= µ−1 µ+1 .
κ 2 exp(µκ)γ̄ 2
Using [24, (8.445)] for modified Bessel function in Equation A.2, we get
(√ )2n+µ−1 ∫∞ ( )
∑∞
µ2n+µ−1 κ(1 + κ) µ(1 + κ)γ
I(s) = β γ n+µ−1
exp − − sγ dγ
n=0
n!Γ(n + µ) γ̄ γ̄
γT
the integration term in above equation can be represented as upper incomplete gamma
function Γ(·, ·) and thus, I(s) can be expressed as in Equation 3.19.
39
Appendix B
where, Pγ (γ) is the CDF of individual branch SNR, and for κ-µ distribution, is given by
[17]
( √ )
√ 2(1 + κ)µγ
Pγ (γ) = 1 − Qµ 2κµ, (B.2)
γ̄
and, pγ (γ) is the PDF of the SNR of κ-µ distributed channel, and is given by Equa-
tion 3.13. Further, on substituting Pγ (γ) in Equation B.1, I1 (s) can be written as
[∫ ∫ γT ( √ ) ]
γT
−sγ
√ 2(1 + κ)µγ −sγ
I1 (s) = 2 · pγ (γ)e dγ − Qµ 2κµ, pγ (γ)e dγ
0 0 γ̄
or
∫ ∞ ∫ γT ( √ )
√ 2(1 + κ)µγ
I1 (s) = 2·Mγ (s) − −sγ
pγ (γ)e dγ − Qµ 2κµ, pγ (γ)e dγ
−sγ
γ̄
γ
| T {z } |
0
{z }
I(s) I2 (s)
(B.3)
where, I(s) has already been calculated in Appendix A, and Mγ (s) is given in Equa-
tion 3.17. Now, I2 (s) can be written as
∫ γT ( √ ) ( √ )
√ 2(1 + κ)µγ −
µ(1+κ) κ(1 + κ)µγ
γ 2 e ( γ̄ ) I
µ−1
I2 (s) = θ · Qµ 2κµ, +s γ
µ−1 2µ dγ
0 γ̄ γ̄
40
where,
( ) µ+1 ( ) µ−1
µ 1+κ 2 1 2
θ = µκ
e γ̄ κ
∑∞ ( )j
−(α2 +β 2 )/2 α
QM (α, β) = e Ij (αβ)
j=1−M
β
∞ ( )−j/2 ∫ γT ( √ )
∑ 1+κ 2µ(1+κ) κ(1 + κ)µγ
I2 (s) = θe−κµ γ (µ−j−1)/2 e−( γ̄ +s)γ Ij 2µ
j=1−µ
κγ̄ 0 γ̄
( √ )
κ(1 + κ)µγ
×Iµ−1 2µ dγ
γ̄
−µ − v − 1 ∈
/N
γ
also, using the series representation of exponential function and putting γT
= u, I2 (s)
can now be written as
∑
∞ ∑
∞ ( )n µ−j+1
(−1)n 2µ(1 + κ) ( +n)
I2 (s) = λ g(j) + s γT 2
1−µ n=0
n! γ̄
∫ 1 [ ]
1 1
2 κ(1 + κ)γT u
µ−j−1 0, ,
× u( 2 +n) G1,2
3,5 4µ j+µ−1 j+µ−1 2 j−µ+1
4
du (B.4)
0 γ̄ 2 , − 2 , 2 , µ−1−j 2
, 14
| {z }
I3 (s)
where,
√ ( ) µ+1 ( ) µ−1
µ π 1+κ 2 1 2
λ = 2µκ
e γ̄ κ
and ( )−j/2
1+κ (π )
g(j) = csc (2j + 2µ + 1)
κγ̄ 4
using the identity of Meijer-G function [28]
∫ 1 [ ] [ ]
a α, a
−α p p
x (1 − x) Gp,q zx
α−β−1 m,n
dx = Γ(α − β)Gp+1,q+1 z
m,n+1
0 bq bq , β
41
I3 (s) can be solved as
[ ( µ−j−1 ) ]
− + n , 0, 21 , 14
κ(1 + κ)γT
1,3
I3 (s) = G4,6 4µ2 2
( ) (B.5)
γ̄ j+µ−1
2
, − j+µ−1
2
, j−µ+1
2
, µ−1−j
2
, 14 , − µ−j+1
2
+n
now, putting Equation B.5, λ and g(j) in Equation B.4, I2 (s) can be found. Further, on
putting I(s), I2 (s) and Mγ (s), I1 (s) can be found.
Using the derived I1 (s), the MGF of output SNR of SECps scheme, MγSECps (s) can be
found and is given in Equation 3.21.
42
References
[4] D. G. Brennan, “Linear diversity combining techniques,” Proc. IRE, vol. 47,
pp. 1075–1102, June 1959.
43
[11] M. K. Simon and M. S. Alouini, “A unified performance analysis of digital com-
munications with dual selective combining diversity over correlated Rayleigh and
Nakagami-m fading channels,” IEEE Trans. Commun., vol. COM-47, pp. 33–43,
January 1999.
[17] M. D. Yacoub, “The κ-µ distribution and the η-µ distribution,” IEEE Antennas and
Propagation Mag., vol. 49, pp. 68–81, February 2007.
[18] N. Y. Ermolova, “Moment generating functions of the generalized η-µ and κ-µ dis-
tributions and their applications to performance evaluations of communication sys-
tems,” IEEE Commun. Lett., vol. 12, pp. 502–504, July 2008.
44
[21] D. Dixit and P. R. Sahu, “Performance of L-branch MRC receiver in η-µ and κ-µ
fading channels for QAM signals,” IEEE Wireless Commun. Lett., vol. 1, pp. 316–
319, August 2012.
[22] H. Shin and J. H. Lee, “On the error probability of binary and M-ary signals in
Nakagami-m fading channels,” IEEE Trans. Commun., vol. 52, pp. 536–539, April
2004.
[24] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and Products. New
York: Academic, 6th ed., 2000.
[26] http://mathworld.wolfram.com/MarcumQ-Function.html/.
[27] http://functions.wolfram.com/Bessel-TypeFunctions/BesselI/.
[28] http://functions.wolfram.com/HypergeometricFunctions/MeijerG/.
45