1
Asymptotically Tight Error Rate Bounds for
Diversity Receptions Over Arbitrarily Correlated
Rician Channels
Bingcheng Zhu
∗†
, Julian Cheng
†
, Naofal Al-Dhahir
‡
and Lenan Wu
∗
∗
School of Information Science and Engineering, Southeast University
Nanjing, Jiangsu, China, Emails:
{230109122, wuln}@seu.edu.cn
†
School of Engineering, The University of British Columbia
Kelowna, BC, Canada Email: julian.cheng@ubc.ca
‡
Electrical Engineering Department, The University of Texas at Dallas
Email: aldhahir@utdallas.edu
Abstract—Prior work on exact error rate of diversity
receptions over arbitrarily correlated Rician channels involves
intractable integrals. Although asymptotic error rates provide
closed-form approximations which are accurate at high signal-
to-noise ratio, it is still unknown at what signal-to-noise ratio the
approximation becomes accurate. In this paper, asymptotically
tight closed-form upper and lower bounds are derived for
maximal-ratio combining and selection combining over arbi-
trarily correlated Rician channels. Using these bounds, we can
show at what signal-to-noise ratio that the asymptotic error rates
are accurate without resorting to time-consuming Monte Carlo
simulation.
I. INTRODUCTION
Multi-branch receiver combining, also known as diversity
reception, is an important technique to combat fading in
wireless communications [1]. Diversity reception increases
the robustness of the system and provides larger diversity
order than a single-input single-output system. Among all
the diversity reception techniques, maximal-ratio combining
(MRC), selection combining (SC) and equal-gain combining
(EGC) are the three most fundamental linear combinings and
they have been widely studied. Existing work has provided
error rate analysis for diversity receptions over independent
fading channels [2]. However, channel coefficients can be
correlated due to the lack of spacing between antennas in
practical systems. Analyzing the error rates of diversity recep-
tions over correlated channels is challenging and usually leads
to intractable multilevel integrals [3], [4]. The computational
complexity also grows exponentially with the number of anten-
nas for correlated channels. To the best of our knowledge, no
closed-form exact error rate expressions have been presented
for the three diversity receptions over arbitrarily correlated
Rician channels.
Since analyzing the exact error rates is analytically in-
tractable, asymptotic analysis is proposed to provide closed-
form and accurate approximations at large signal-to-noise ratio
(SNR) [5]–[7]. However, it is still unknown at what SNR value
the asymptotic error rate can provide accurate approximation
to the exact error rate. To bound the discrepancy between the
exact error rate and the asymptotic approximation, the authors
in [8] derived upper and lower bounds for the error rates of
SC and EGC over correlated Rician channels, but the assumed
underlying correlation structure was not arbitrary and such a
correlation has its severe limitations.
Monte Carlo simulation can be accurate at low to medium
SNR region, but when the error rate is below 10
−10
, accurate
Monte Carlo simulation is time-consuming, and requires large
number of simulation trials. In this paper, we derive novel up-
per and lower closed-form bounds for the error rates of MRC
and SC over arbitrarily correlated Rician channels. The key
idea behind our derivation is to bound the probability density
function (pdf) of the coefficients of the Rician fading channels,
and to exploit rapidly decreasing characteristic of Gaussian
Q-function. Both the upper and lower bounds for the error
rates of MRC and SC converge to the asymptotic error rate in
large SNR region. With these new analytical bounds, one can
bound the discrepancy between the asymptotic error rates and
the exact error rates without using the time-consuming Monte
Carlo simulation.
II. S
YSTEM MODEL AND ASYMPTOTIC ERROR RATE S
We consider diversity reception with N branches over slow
frequency-nonselective Rician fading channels. After coherent
demodulator, the complex output vector can be expressed as
r = cx + n (1)
where x is the transmitted complex signal; c =[c
1
,c
2
,
...,c
N
]
T
where c
n
is the complex Rician channel coefficient
of the nth branch and where [·]
T
denotes the transpose
operation; n is an N × 1 vector representing the additive
Gaussian white noise. The pdf of the channel coefficient vector
c is given by
f
c
(c)=
1
π
N
det (R)
exp
−(c − c
L
)
H
A (c − c
L
)
(2)
where c
L
=[c
L,1
,...,c
L,N
]
T
is a complex line-of-sight
(LOS) vector, which is also the expectation of c; R =
E[c
s
c
s
H
] is the correlation matrix of the scattering compo-
nents defined as c
s
= c − c
L
and (·)
H
denotes the conjugate
2015 IEEE Wireless Communications and Networking Conference (WCNC 2015) - Track 1: PHY and Fundamentals
978-1-4799-8406-0/15/$31.00 ©2015 IEEE 699