Kashani2015

This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/JLT.2015.2410695, Journal of Lightwave Technology
1
A Novel Statistical Channel Model for

Turbulence-Induced Fading in Free-Space Optical
Systems
Mohammadreza A. Kashani, Student Member, IEEE, Murat Uysal, Senior Member, IEEE,
and Mohsen Kavehrad, Fellow, IEEE
Abstract—In this paper, we propose a new probability distri- of the atmosphere result in variations of the refractive index
bution function which accurately describes turbulence-induced and cause atmospheric turbulence. This manifests itself as ran-
fading under a wide range of turbulence conditions. The proposed dom fluctuations in the received signal and severely degrades
model, termed Double Generalized Gamma (Double GG), is
based on a doubly stochastic theory of scintillation and developed the FSO system performance particularly over long ranges.
via the product of two Generalized Gamma (GG) distributions. In the literature, several statistical models have been pro-
The proposed Double GG distribution generalizes many existing posed in an effort to model this random phenomenon. His-
turbulence channel models and provides an excellent fit to the torically, log-normal distribution has been the most widely
published plane and spherical waves simulation data. Using this used model for the probability density function (pdf) of the
new statistical channel model, we derive closed form expressions
for the outage probability and the average bit error as well as random irradiance over atmospheric channels [5]–[7]. This
corresponding asymptotic expressions of free-space optical com- pdf model is however only applicable to weak turbulence
munication systems over turbulence channels. We demonstrate conditions. As the strength of turbulence increases, lognormal
that our derived expressions cover many existing results in the statistics exhibit large deviations compared to experimental
literature earlier reported for Gamma-Gamma, Double-Weibull data. Moreover, lognormal pdf underestimates the behavior
and K channels as special cases.
in the tails as compared with measurement results. Since the
Index Terms—Free-space optical systems, fading channels, calculation of detection probabilities for a communication sys-
Double GG distribution, propagation, irradiance, optical wireless, tem is primarily based on the tails of the pdf, underestimating
spatial diversity.
this region significantly affects the accuracy of performance
analysis.
I. I NTRODUCTION In an effort to address the shortcomings of the lognormal
REE-SPACE optical (FSO) communication enables wire- distribution, other statistical models have been further pro-
F less connectivity through atmosphere using laser trans-
mitters at infrared bands. These systems provide high data
posed to describe atmospheric turbulence channels under a
wide range of turbulence conditions. These include the Nega-
rates comparable to fiber optics while they offer much more tive Exponential/Gamma model (also known widely as the K
flexibility in (re)deployment. Since they operate in unregulated channel) [8], I-K distribution [9], log-normal Rician channel
spectrum, no licensing fee is required making them also a (also known as Beckman) [10], Gamma-Gamma [11], M
cost-effective solution [1]–[3]. With their unique features and distribution [12] and Double-Weibull [13]. Particularly worth
advantages, FSO systems have attracted attention initially as mentioning is the Gamma-Gamma model [11], [14] which has
a last mile solution and can be used in a wide array of been widely used in the literature for the performance analysis
applications including cellular backhaul, inter-building con- of FSO systems, see e.g., [15], [16], along with the log-normal
nections in enterprise/campus environments, video surveil- model. This model builds upon a two-parameter distribution
lance/monitoring, fiber back-up, redundant link in disaster and considers irradiance fluctuations as the product of small-
recovery and relief efforts among others. scale and large-scale fluctuations, where both are governed by
A major performance limiting factor in FSO systems is independent gamma distributions. In a more recent work by
atmospheric turbulence-induced fading (also called as scintil- Chatzidiamantis et al. in [13], the Double-Weibull distribution
lation) [4]. Inhomogenities in the temperature and the pressure was proposed as a new model for atmospheric turbulence
channels. Similar to the Gamma-Gamma model, it is based
Mohammadreza A. Kashani and Mohsen Kavehrad are with Pennsyl- on the theory of doubly stochastic scintillation and considers
vania State University, University Park, PA 16802 (email: mza159@psu.edu,
mkavehrad@psu.edu). irradiance fluctuations as the product of small-scale and large-
Murat Uysal is with Department of Electrical and Electronics scale fluctuations which are both Weibull distributed. It is
Engineering, Özyeǧin University, Istanbul, Turkey, 34794 (email: mu- shown in [13] that Double-Weibull is more accurate than the
rat.uysal@ozyegin.edu.tr).
This paper was presented in part at the IEEE international Conference Gamma-Gamma particularly for the cases of moderate and
on Transparent Optical Networks (ICTON), Cartagena, Spain, June 2013. The strong turbulence.
work of M. Uysal is supported by TUBITAK Research Grant 111E143. In this paper, we propose a new and unifying statistical
Copyright (c) 2015 IEEE. Personal use of this material is permitted.
However, permission to use this material for any other purposes must be model, named Double Generalized Gamma (Double GG), for
obtained from the IEEE by sending a request to pubs-permissions@ieee.org. the irradiance fluctuations. The proposed model is valid under
0733-8724 (c) 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
2
all range of turbulence conditions (weak to strong) and con- where fIx (I|Iy ) is obtained as
tains most of the existing statistical models for the irradiance m γ −1 ( ( )γ1 )
fluctuations in the literature as special cases. Furthermore, γ1 (I/Iy ) 1 1 m1 I
fIx (I|Iy ) = m exp −
we provide comparison of the proposed model with Gamma- Iy (Ω1 /m1 ) 1 Γ (m1 ) Ω 1 Iy
Gamma and Double-Weibull models. For this purpose, we (4)
use the set of simulation data from [17], [18] for plane and The integration in (3) yields
spherical waves1 . Our model demonstrates an excellent match γ2 ppm2 −1/2 q m1 −1/2 (2π) I −1
1−(p+q)/2
to the simulation data and is clearly superior over the other fI (I) = (5)
Γ (m1 ) Γ (m2 )
two models which show discrepancy from the simulation data [( )p p q q ]
Ω2 p q Ω1 ∆ (q : 1 − m1 ) , ∆ (p : 1 − m2 )
in some cases. In the second part of the paper, we use this × G0,p+q |
new channel model to derive closed form expressions for the
p+q,0
I γ2 mq1 mp2 −
BER and the outage probability of single-input single-output where Gm,n 2
p,q [.] is the Meijers G-function defined in [25,
(SISO) and single-input multiple-output (SIMO) FSO systems Eq.(9.301)], p and q are positive integer numbers that satisfy
with intensity modulation and direct detection (IM/DD). Our p/q = γ1 /γ2 and ∆(j; x) , x/j , ..., (x + j − 1)/j .
performance results can be seen as a generalization of the We name this new distribution as Double GG. Employing
results in [20]–[23]. [26, Eq. (10)] and after some simplifications, the cumulative
The rest of the paper is organized as follows: In Section II, distribution function (cdf) of Double GG distribution can be
we propose Double GG distribution to characterize turbulence- obtained as
induced fading. In Section III, we confirm the accuracy of our
pm2 −1/2 q m1 −1/2 (2π)
1−(p+q)/2
model through comparisons with simulation data for plane and FI (I) = (6)
spherical waves. In Section IV, we present the derivation of bit Γ (m ) Γ (m )
[( γ2 )p 1 q p 2 ]
error rate (BER) and outage probability for SISO FSO system I m1 m2 1
× Gp+q,1 |
over Double GG channel. In Section V, we present BER 1,p+q+1
Ω2 pp q q Ωq1 ∆ (q : m1 ) , ∆ (p : m2 ) , 0
expressions for FSO links with multiple receiver apertures.
Finally, Section VI concludes the paper. The distribution parameters γi and Ωi i = 1, 2 of the Double
GG model can be identified using the first and second order
moments of small and large scale irradiance fluctuations. The
II. D OUBLE GG D ISTRIBUTION latter are directly tied to the atmospheric parameters. Without
loss of generality, we assume E (I) = 1 or equivalently
The irradiance of the received optical wave can be modeled
E (Ix ) = 1 and E (Iy ) = 1. The second moment of irradiance
as [11], [13] I = Ix Iy , where Ix and Iy are statistically
is expressed as
independent random processes arising respectively from large-
( ) ( ) ( ) ( )( )
scale and small scale turbulent eddies. We assume that both E I 2 = E Ix2 E Iy2 = 1 + σx2 1 + σy2 (7)
large-scale and small-scale irradiance fluctuations are gov-
erned by Generalized Gamma (GG) distributions [24, Eq. (1)]. where σx2 and σy2 are respectively normalized variances of Ix
The pdfs of Ix ∼ GG (γ1 , m1 , Ω1 ) and Iy ∼ GG (γ2 , m2 , Ω2 ) and Iy . The nth moment of Ix (similarlyIy ) is given by
are given by ( )n/γ1
Ω1 Γ (m1 + n/γ1 )
( ) E (Ixn ) = (8)
γ1 Ixm1 γ1 −1 m1 γ 1 m1 Γ (m1 )
fIx (Ix ) = m exp − I (1)
(Ω1 /m1 ) 1 Γ (m1 ) Ω1 x Therefore, by inserting the second order moment obtained
( ) from (8) in (7), and considering that E (I) = 1, we have
γ2 Iym2 γ2 −1 m2 γ2
fIy (Iy ) = m exp − I (2) Γ (m1 + 2/γ1 ) Γ (m1 )
(Ω2 /m2 ) 2 Γ (m2 ) Ω2 y σx2 = −1 (9a)
Γ2 (m1 + 1/γ1 )
where γi > 0 , mi ≥ 0.5 and Ωi i = 1, 2 are the GG Γ (m2 + 2/γ2 ) Γ (m2 )
parameters. The pdf of I can be written as σy2 = −1 (9b)
Γ2 (m2 + 1/γ2 )
( )γi
∫∞ Γ (mi )
Ωi = mi , i = 1, 2 (10)
fI (I) = fIx (I|Iy ) fIy (Iy ) dIy (3) Γ (mi + 1/γi )
0
where mi is a distribution shaping parameter and found using
1 The simulation data in [17], [18] was obtained through phase screen ap-
curve fitting on the simulated/measured channel data. Note
proach which consists of approximating a three-dimensional random medium that in (9a) and (9b), the variances of small- and large-scale
as a collection of equally spaced, two-dimensional, random phase screens that uctuations (i.e., σx2 and σy2 ) are directly tied to the atmospheric
are transverse to the direction of wave propagation. In [17], it was discussed conditions. Particularly, assuming a plane wave when inner
in detail that such a numerical simulation approach contains all the essential
physics for accurately predicting the pdf of irradiance (or, equivalently, log- scale effects are considered, the variances for the large- and
normal irradiance), and shown that the simulation results provide an excellent
match to known experimental measurements reported in [19] for both plane 2 Meijers G-function is a standard built-in function in mathematical
and spherical waves. The same set of simulation data was also used in [11] software packages such as MATLAB, MAPLE and MATHEMATICA. If
and [13] which respectively introduced Gamma-Gamma and Double-Weibull required, this function can be also expressed in terms of the generalized
distributions as turbulence channel models. hypergeometric functions using [25, Eqs.(9.303-304)].
3
Fig. 1. Pdfs of the scaled log-irradiance for a plane wave assuming weak Fig. 2. Pdfs of the scaled log-irradiance for a plane wave assuming moderate
irradiance fluctuations. irradiance fluctuations.
the small-scale scintillations are given by [4, Eqs. 9.46 and In (15), σ̃ 2 (l0 /R0 ) is defined as
9.55] [ ( ( )
( ) 11 −1 ηl
2 ∼
σ̃ (l0 /R0 ) = 3.86 1 + 9/ηl 2 11/12
sin tan
 ( )7/6 6 3
2.61η ( )
σx2 = exp 0.16σRytov 2.61 4 ηl
2 l
2 7/6 + sin tan−1
2.61 + ηl + 0.45σRytov ηl 2
(9 + ηl )
1/4 3 3
 ( )1/2 ( )) ]
2.61 0.518 5 −1 ηl −5/6
× 1 + 1.753 − 7/24
sin tan − 8.75ηl (16)
2
2.61 + ηl + 0.45σRytov ηl
7/6 (9 + η 2 ) 4 3
l
( )7/12 
2.61 In the presence of a nite inner scale, the small-scale scintil-
− 0.252  − 1 (11)
2
2.61 + ηl + 0.45σRytov
7/6
ηl lation is again described by (14) and the large-scale variance
is given by [4, Eq. 78]
 ( )7/6
 
2 ∼ 8.56η
σx = exp 0.04β0 2 l
2
 0.51σRytov 
σy2 ∼
7/6
= exp  ( )5/6  − 1 (12) 8.56 + ηl + 0.195β02 ηl
12/5
1 + 0.69σRytov  ( )1/2
 8.56
× 1 + 1.753 7/6
where ηl = 10.89 (R0 /l0 ), and R0 /l0 denotes the ratio of 8.56 + ηl + 0.195β02 ηl
Fresnel zone to finite inner scale. ( )7/12 
For spherical waves in the absence of inner scale, σx2 and 8.56  − 1
−0.252 7/6
(17)
2
σy are given by [4, Eqs. 9.63 and 9.70] 8.56 + ηl + 0.195β02 ηl
  Therefore, the parameters of the Double GG distribution are
 0.49β02  readily deduced from these expressions using only values
σx2 ∼
= exp  ( )7/6  − 1 (13)
12/5 of the refractive index structure parameter and inner scale
1 + 0.56β0 according to the atmospheric conditions. The scintillation
 
index can be further calculated as
 0.51β02  ( )
σy2 ∼
= exp  ( )5/6  − 1 (14) E I2 ( )( )
σI2 = 2 − 1 = 1 + σx 1 + σy2 − 1
2
1+
12/5
0.69β0 (18)
E(I)
We should emphasize that this distribution is very generic
where β02 is the Rytov scintillation index of a spherical wave
since it includes some commonly used fading models as
given by
special cases. For γi → 0, mi → ∞, Double GG pdf coincides
β02 = σRytov
2
/σ̃ 2 (l0 /R0 ) (15) with the log-normal pdf. For γi = 1, Ωi = 1, it reduces
4
Fig. 3. Pdfs of the scaled log-irradiance for a plane wave assuming strong Fig. 4. Pdfs of the scaled log-irradiance for a spherical wave assuming weak
irradiance fluctuations. irradiance fluctuations.
TABLE I
NRMSE FOR DIFFERENT STATISTICAL MODELS AND TURBULENCE logarithm of irradiance was particularly chosen to illustrate
CONDITIONS DEFINED IN F IGS . 1-6 the high and low irradiance tails [17]. Thus, sensitivity to the
small irradiance fades is increased, while sensitivity to large
Gamma-Gamma [11] Double-Weibull [13] Double GG (Proposed) irradiance peaks is decreased. The pdf plots were also scaled
Fig. 1 2% 7% 0.6%
Fig. 2 2.8% 1.2% 0.8% by subtracting the mean value to center all distributions on
Fig. 3 1.2% 1% 0.8% zero and dividing by the square root of variance.
Fig. 4 0.3% 10% 0.3%
Fig. 5 19% 8.7% 1.5%
In Fig. 1, we assume weak turbulence conditions which are
2
Fig. 6 4.8% 2.4% 1.7% characterized by σRytov = 0.1 and l0 /R0 = 0.5. The values
of the variances of small and large scale fluctuations, (σx2 and
σy2 ) are calculated from (11) and (12). Using (9a), (9b) and
to Gamma-Gamma while for mi = 1, it becomes Double- (10), the Double-GG parameters are obtained as γ1 = 2.1,
Weibull. For γi = 1, Ωi = 1, m2 = 1, it coincides with the K γ2 = 2.1, m1 = 4, m2 = 4.5, Ω1 = 1.0676 and Ω2 = 1.06
channel. where p = q = 1. We further employ normalized root-mean-
square error (NRMSE) as a statistical goodness of fit test.
III. V ERIFICATION OF THE P ROPOSED C HANNEL M ODEL Table I provides the NRMSE results for different statistical
models. According to Table I and Fig. 1, both Gamma-Gamma
In this section, we compare the Double GG distribution
and Double-Weibull distributions fail to match the simulation
model with simulation data of plane and spherical waves pro-
data. On the other hand, the proposed Double GG distribution
vided respectively in [17] and [18]. In [17], Flatté et al. carried
follows closely the simulation data.
out exhaustive numerical simulations and published the results
In Fig.2, we assume moderate irradiance fluctuations which
assuming plane wave propagation through homogeneous and 2
are characterized by σRytov = 2 and l0 /R0 = 0.5. The
isotropic Kolmogorov turbulence. In [18], Hill and Frehlich
parameters of the Double GG distribution for this case are
presented the simulation data for the propagation of a spher-
obtained as γ1 = 2.1690, γ2 = 0.8530, m1 = 0.55, m2 =
ical wave through homogeneous and isotropic atmospheric
2.35, Ω1 = 1.5793 and Ω2 = 0.9671. In the calculations,
turbulence. The turbulence severity is characterized by Rytov
2 p and q are chosen as p = 28 and q = 11 in order to
variance (σRytov ) which is proportional to the scintillation index
satisfy p/q = γ1 /γ2 . Among the three distributions under
[4]. We emphasize that the same data set was also employed
consideration, the proposed Double GG model provides the
in [11] and [13] which have introduced the Gamma-Gamma
best fit to the simulation data. It is apparent that Gamma-
and Double-Weibull fading models.
Gamma fails to match the simulation data particularly in the
tails. As Table I demonstrates, the accuracy of Double Weibull
A. Plane Wave is better than that of Gamma-Gamma, but slightly inferior to
Figs. 1-3 compare the Gamma-Gamma, Double-Weibull and our proposed distribution.
Double GG models under a wide range of turbulence condi- In Fig. 3, we assume strong irradiance fluctuations which are
2
tions (weak to strong) assuming plane wave propagation. In characterized by σRytove = 25 and I0 /R0 = 1. The parameters
these figures, the vertical axis of the figure represents the log- of the Double GG distribution are calculated as γ1 = 1.8621,
irradiance pdf multiplied by the square root of variance. The γ2 = 0.7638, m1 = 0.5, m2 = 1.8, Ω1 = 1.5074 and Ω2 =
5
(a)
Fig. 5. Pdfs of the scaled log-irradiance for a spherical wave assuming

moderate irradiance fluctuations.
(b)
(c)
Fig. 6. Pdfs of the scaled log-irradiance for a spherical wave assuming strong
irradiance fluctuations.
0.9280 where p and q are chosen as 17 and 7 respectively.

The Double GG model again provides an excellent match to
the simulation data and as it is clear from Table I, its accuracy
is better than Gamma-Gamma and Double Weibull.
B. Spherical Wave
Figs 4-6 compare the Gamma-Gamma, Double-Weibull
and Double GG models under weak, moderate and strong
turbulence conditions assuming spherical wave propagation.
(d)
These pdfs are plotted as a function of (ln I + 0.5σ 2 )/σ [18],
where σ is the square root of the variance of ln I. The y-axis Fig. 7. Outage probability as a function of γ̄/γth for a) Plane wave -
2 2
depicts the log-irradiance pdf multiplied by σ. σRytov = 2, l0 /R0 = 0.5, b) Plane wave-σRytov = 25, l0 /R0 = 1, c)
2
Spherical wave - σRytov 2
= 2, l0 /R0 = 0, d) Spherical wave - σRytov = 5,
In Fig. 4, we consider spherical wave propagation and
2 l0 /R0 = 1
assume weak turbulence which are characterized by σRytov =
6
0.06 and l0 /R0 = 0. The parameters of Double GG are evalu-

ated using the variances of small and large scale fluctuations,
(σy2 and σx2 ) for spherical waves. The values of these variances
are given by (13) , (14) and (17). Therefore, employing (9)
and (10) we obtain m1 = 34.24, m2 = 32.79, γ1 = γ2 =
Ω1 = Ω2 = 1 where p and q are equal to 1. It can be noted
that in this case, the Double GG coincides with the Gamma-
Gamma distribution. It is apparent that both Gamma-Gamma
and Double GG distributions provide an excellent match to the
simulation data while the Double-Weibull distribution fails to
match the simulation data.
In Fig. 5, we assume moderate irradiance fluctuations, which
2
are characterized by σRytov = 2 and l0 /R0 = 0. The param-
(a)
eters of the Double GG model for this case are calculated
as γ1 = 0.9135, γ2 = 1.4385, m1 = 2.65, m2 = 0.85,
Ω1 = 0.9836 and Ω2 = 1.1745 where p and q are selected as
7 and 11 respectively. It is clearly observed that the Double
GG model provides a better fit with simulation data, especially
for small irradiance values.
In Fig. 6, we assume strong irradiance fluctuations which are
2
characterized by σRytov = 5 and l0 /R0 = 1. The parameters
of the Double GG model are calculated as γ1 = 0.4205,
γ2 = 0.6643, m1 = 3.2, m2 = 2.8, Ω1 = 0.8336 and
Ω2 = 0.9224 where p and q are chosen as 7 and 11
respectively. It is apparent from this figure and Table I that
both Gamma-Gamma and Double-Weibull distributions fail to
match the simulation data. On the other hand, the proposed (b)
Double GG distribution follows closely the simulation data.
IV. P ERFORMANCE E VALUATION

A. Outage Probability Analysis of SISO FSO System
Denote Rt as a targeted transmission rate and assume
γth = C −1 (Rt ) as the corresponding signal-to-noise ratio
(SNR) threshold in terms of the instantaneous channel capacity
for a particular channel realization. Therefore, the outage
probability is calculated by Pout (Rt ) = P r (γ < γth ) [27],
[28]. If SNR exceeds γth , no outage happens and the receiver
can decode the signal with arbitrarily low error probability. For
the system under consideration, the instantaneous electrical
2 (c)
SNR can be defined as γ = (ηI) /N0 , while the average
electrical SNR is obtained as γ̄ = η 2 /N0 √ since E (I) = 1.
Therefore, I can be expressed as I = γ/γ̄. After the
transformation of the random variable, I, the cdf of γ can
be easily derived from (6) and setting γ = γth therein, we
obtain the outage probability as in (19) at the top of the next
page.
(19) can be seen as a generalization of earlier outage
analysis results in the literature. Specifically, if we insert
γi = 1 and Ωi = 1 in (19), we obtain the outage probability
expression reported in [20, Eq. (15)] for Gamma-Gamma
channel. Setting mi = 1 in (19), we recover Eq (16) of [13]
derived for Double-Weibull channel. Similarly, for γi = 1,
Ωi = 1 and m2 = 1, (19) reduces to (3) of [29] reported for (d)
the K channel. 2
Fig. 8. Average BER as a function of γ̄ a) Plane wave - σRytov = 2,
Based on the derived expression in (19), Figs. 7.a-d present 2
l0 /R0 = 0.5, b) Plane wave-σRytov = 25, l0 /R0 = 1, c) Spherical wave -
the outage probabilities of a SISO FSO system for different 2
σRytov 2
= 2, l0 /R0 = 0, d) Spherical wave - σRytov = 5, l0 /R0 = 1
degrees of turbulence severity. We adopt the same parameters
7
[( )pγ2 /2 ]
pm2 −1/2 q m1 −1/2 (2π)
1−(p+q)/2
γth mq1 mp2 1
Pout = Fγ (γth ) = Gp+q,1
1,p+q+1 p q| (19)
Γ (m1 ) Γ (m2 ) γ̄ (Ω2 p) (Ω1 q) ∆ (q : m1 ) , ∆ (p : m2 ) , 0
used in Figs 2-3 and 5-6 and consider the following four In (24) k and l are positive integer numbers that satisfy
2
cases: a) Plane wave with σRytov = 2 and l0 /R0 = 0.5, b) pγ2 /2 = l/k and Jξ (y, x) is defined as
2
Plane wave with σRytov = 25 and l0 /R0 = 1, c) Spherical
2
wave with σRytov = 2 and l0 /R0 = 0, d) Spherical wave with Jξ (y, x) (25)
( ) ( ) ( )
2
σRytov = 5 and l0 /R0 = 1 . It is observed from Fig. 7.a y−x y−1−x 1−x
= ∆ ξ, , ∆ ξ, , . . . , ∆ ξ,
that an SNR of 37.8 dB is required to achieve a targeted y y y
outage probability of 10−2 . As the turbulence strength gets
The derived BER expression in (24) can be seen as a
stronger (see Fig. 7.b), the required SNR to maintain the same
generalization of earlier BER results in the literature. If we
performance climbs up to 50.5 dB. Similarly, for spherical
insert γi = 1 and Ωi = 1 in (24), we obtain the BER
waves, SNRs of 36.8 dB and 50.9 dB are respectively required
expression derived in [21, Eq. (9)] under the assumption of
for moderate and strong turbulence conditions. In these figures,
Gamma-Gamma channel. Setting mi = 1 in (24), we obtain
we further include the outage results for Double-Weibull and
Eq (15) of [13] derived for Double-Weibull channel. On the
Gamma-Gamma for comparison purposes. As expected from
other hand, for γi = 1, Ωi = 1 and m2 = 1, (24) reduces to
the earlier comparisons of their pdfs, the outage performance
(12) of [22] reported for the K-channel.
over Double-Weibull and Double-GG for plane wave (See
In an effort to have some further insights into system per-
Figs. 7.a and 7.b) are similar while the Gamma-Gamma model
formance, we investigate the asymptotical BER performance
overestimates the outage performance. On the other hand,
in the following. For large SNR values, the asymptotic BER
the superiority of Double GG is more obvious for spherical
behavior is dominated by the behavior of the pdf near the
wave (See Figs 7.c and 7.d), particularly for strong turbulence
origin, i.e. fI (I) at I → 0 [32]. Thus, employing series
conditions, where the outage performance plots of Double-
expansion corresponding to the Meijers G-function [33, Eq.
Weibull and Gamma Gamma significantly deviate.
(07.34.06.0006.01)], the Double GG distribution given in (5)
can be approximated by a single polynomial term as
B. BER Analysis of SISO FSO System
∏
p+q
Γ (bj − bk )I pγ2 min{ }−1
m1 m2
In this section, we present the BER performance analysis of fI (I) ≈ A q , p (26)
an FSO system with on-off keying (OOK) over the proposed j=1
Double GG channel. The received optical signal is written as j̸=k
where A is obtained as
y = ηIx + n (20)
γ2 ppm2 −1/2 q m1 −1/2 (2π)
1−(p+q)/2
where x represents the information bits and can be either 0 A=
or 1, n is the Additive White Gaussian noise (AWGN) term Γ (m1 ) Γ (m2 )
( q p )min{ mq1 , mp2 }
with zero mean and variance σn2 = N0 /2 , η is the optical-to- m m
electrical conversion coefficient and I is the normalized irra- × 1
q
2
p (27)
(qΩ1 ) (pΩ2 )
diance whose pdf follows (5). Conditioned on the irradiance,
the instantaneous BER for OOK is given by [22] In (26), bk and bj are defined as
( ) { }
ηI m1 m2
Pe,ins = 0.5 erfc √ (21) bk = min
q
,
p
(28)
2 N0
where erfc (.) stands for the complementary error function
bj ∈ (29)
defined as ∫ ∞ { }
2 m1 m2
e−t dt.
2
erfc(x) = √ (22) {1 − ∆ (q : 1 − m1 ) , 1 − ∆ (p : 1 − m2 )} \ min ,
π x q p
The average BER can be then calculated by averaging (21) Therefore, based on (23), the average BER can be well
over the distribution of I, i.e., approximated by
∫ ∞ [ ( )]
ηI ∏
p+q ( )pγ2 bk
Pe = fI (I) 0.5 erfc √ dI (23) 2 Γ ((1 + pγ2 bk ) /2)
0 2 N0 PSISO ≈ A Γ (bj − bk ) √ √
j=1
γ̄ 2 πpγ2 bk
The above integral can be evaluated in closed form by ex- j̸=k
pressing the erfc (.) integrand via a Meijers G-function using (30)
[30, Eq. (8.4.14.2)], [30, Eq. (8.2.2.14)] and [31, Eq. (21)]. From (30), it can be readily deduced that the diversity order
Thus, a closed-form solution is obtained as in (24) at the top of SISO FSO system is given by 0.5pγ2 min {m1 /q , m2 /p}.
of the next page.
8
[( )k ( ) ]
γ2 k m1 +m2 pm2 +1/2 q m1 −1/2
l
k(p+q),2l mq1 mp2 (4l) ∆ (l : 1) , ∆ l : 12
PSISO = G |
3
2 2 lΓ (m1 ) Γ (m2 ) (2π)
l+k(p+q)
2 −1 2l,k(p+q)+l pP Ωp2 q q Ωq1 γ̄ l k k(p+q) Jk (q : 1 − m1 ) , Jk (p : 1 − m2 ) , ∆ (l : 0)
(24)
Fig. 9. Comparison of the average BER between SISO and SIMO with Fig. 10. Comparison of the average BER between SISO and SIMO with
optimal combing for plane wave as defined in Fig. 3. optimal combing for spherical wave as defined in Fig. 5.
It is observed from Fig.8.a that an SNR of 51.1 dB is The optimum decision metric for OOK is given by [23]
required to achieve a BER of 10−3 for a plane wave in moder-
ate turbulence conditions. For stronger turbulence conditions, on
the required SNR to achieve the same BER performance is P (r|on,In ) ≶ P (r|off,In ) (31)
off
68.2 dB as seen from Fig. 8.b. For spherical waves, SNRs of
49.8 dB and 63.8 are respectively required for moderate and
strong turbulence conditions. Comparison with the expressions where r = (r1 , r2 , ..., rN ) is the received signal vector and In
presented for other channel models reveals that the Gamma- is the fading channel coefcient which models the channel from
Gamma model significantly overestimates the performance. the transmit aperture to the nth receive aperture. Following
Similar to earlier observations on the outage analysis, the the same approach as [22], [23] the conditional bit error
superiority of Double GG is more obvious for spherical wave. probabilities are given by
As observed from Figs 8.c and 8.d, the performance plots v
u N  
of Double-Weibull and Gamma Gamma considerably diverge u γ̄ ∑
1
particularly for strong turbulence conditions. Furthermore, it Pe (off |In ) = Pe (on |In ) = Q  t I2 (32)
can be clearly seen that the asymptotic results are in excellent N 2 n=1 n
agreement with exact analytical results within a wide range
of SNR showing the accuracy and usefulness of the derived The average BER can be then obtained as
asymptotic expression given in (30).
v
u 
∫ u γ̄ ∑N
V. FSO L INKS WITH R ECEIVE D IVERSITY PSIMO,OC = fI (I) Q t I 2  dI (33)
2N n=1 n
As it can be noticed from Section IV, the performance of I
a SISO FSO link over moderate and strong atmospheric tur-

bulence is quite poor. To address this issue, multiple transmit where fI (I) is the joint pdf of vector I = (I1 , I2 , . . . , IN ).
and/or receive apertures can be employed and the performance The factor N in (33) is used to ensure that sum of the N
can be improved via diversity gains. In the following, we receive aperture areas is the same as the area of the receive
assume that multiple receive apertures are employed and aperture of the SISO link.
present the BER derivations under the assumption that optimal Eq. (33) does not yield a closed-form solution and requires
gain combining is used. N-dimensional integration. Nevertheless, the Q-function can
9
[ ]
αn ln−0.5 kn 1,n (υN ) n ωn−kn lnln
m +m2,n l
k (p +q ),l ∆ (ln , 1)
Λ (n, υ) = G n n n n k (p +q )
(34)
γ̄ ln knn n n Jkn (qn , 1 − m1,n ) , Jkn (pn , 1 − m2,n )
0.5(ln −1+(kn −1)(pn +qn )) ln ,kn (pn +qn )
2(2π)
x2 2x2
be well-approximated as Q(x) ≈ e− 2 /12 + e− 3 /4 [34] excellent match to simulation results. As a benchmark, the
and thus the average BER can be obtained as BER of SISO FSO link is also included in these figures.
N ∫ ( ) It is observed that receive diversity signicantly improve the
1 ∏ ∞ −γ̄ 2 performance. For instance, at a target bit error rate of 10−3 ,
PSIMO,OC ≈ fIn (In ) exp I dIn
12 n=1 0 4N n we observe performance improvements of 26.8 dB and 39.6
N ∫ ( ) dB respectively for with N = 2 and 3 with respect to the SISO
1 ∏ ∞ −γ̄ 2
+ fIn (In ) exp In dIn (35) transmission over Double GG turbulence channels defined in
4 n=1 0 3N Fig.3. Similarly, for Double GG channels defined in Fig. 5,
The above integral can be evaluated by first expressing the ex- at a BER of 10−3 , performance improvements of 19 dB and
ponential function in terms of the Meijer G-function presented 25.1 dB are achieved for SIMO links with N = 2 and 3
in [31, eq. (11)] as compared to the SISO deployment. It can be further observed
[ − ] that asymptotic bounds on the BER become tighter at high
exp (−x) = G1,0
0,1 x 0 (36) enough SNRs confirming the accuracy and usefulness of the
asymptotic expression given in (40).
Then, a closed-form expression for (35) is obtained using [31,
Eq. (21)] as VI. C ONCLUSIONS
1 ∏
N
1 ∏
N In this paper, we have introduced a new channel model,
PSIMO,OC ≈ Λ (n, 4) + Λ (n, 3) (37) so called Double GG, which accurately describes irradiance
12 n=1 4 n=1
fluctuations over atmospheric channels under a wide range
where Λ (n, υ) is defined in (34) at the top of the page. of turbulence conditions. It is based on the theory of doubly
In (34), ln and kn are positive integer numbers that satisfy stochastic scintillation and considers irradiance fluctuations as
pn γ2,n /2 = ln /kn , and αn and ωn n ∈ {1, 2, . . . , N } are the product of small-scale and large-scale fluctuations which
defined as are both Generalized Gamma distributed. We have obtained
m +1/2 m1,n −1/2 1−(pn +qn )/2 closed-form expressions for the pdf and cdf in terms of
γ2,n pn 2,n qn (2π)
αn = (38) Meijers G-function. Comparisons with the Gamma Gamma
Γ (m1,n ) Γ (m2,n ) and Double-Weibull have shown that the new model pro-
( ) (
−1 pn
)qn
ωn = Ω2,n pn m2,n qn Ω1,n m−1
1,n (39) vides an accurate fit with numerical simulation data for both
plane and spherical waves. Using the new channel model,
The derived expression in (37) includes the previously reported
we have obtained closed-form expressions for the BER and
result in [22] for K channel as a special case.
the outage probability of SISO and SIMO FSO systems. We
Based on the approximation in (26), the corresponding
have demonstrated that our derived expressions cover many
closed-form asymptotic solution for (35) can be obtained as
existing results in the literature earlier reported for Gamma-
1 ∏ 1 ∏
N N Gamma, Double-Weibull and K channels as special cases.
PSIMO,OC asy ≈ Λasy (n, 4) + Λasy (n, 3) (40) Based on the asymptotical performance analysis, we have
12 n=1 4 n=1
further derived diversity gains for SISO and SIMO FSO
where Λasy (n, υ) is defined as systems under consideration.
pn∏
+qn
R EFERENCES
Λasy (n, υ) = αn Γ (bj,n − bk,n )
[1] D. Kedar and S. Arnon, “Urban optical wireless communication net-
j=1
j̸=k
works: the main challenges and possible solutions,” IEEE Communica-
(√ )pn γ2,n bk,n tions Magazine, vol. 42, no. 5, pp. S2–S7, 2004.
[2] S. Arnon, J. R. Barry, G. K. Karagiannidis, R. Schober, and
υN M. Uysal (Eds.), Advanced Optical Wireless Communication. Cambridge
× Γ (pn γ2,n bk,n ) √ pn γ2,n bk,n (41) university press, 2012.
2( γ̄) [3] M. A. Kashani and M. Uysal, “Outage performance of fso multi-
hop parallel relaying,” in IEEE Signal Processing and Communications
Therefore, the diversity order of FSO links with N receive Applications Conference, pp. 1–4, IEEE, 2012.
apertures employing optimal gain combining is obtained as [4] L. C. Andrews and R. L. Phillips, Laser beam propagation through
∑
N { } random media, vol. 152. Society of Photo Optical, 2005.
0.5 pn γ2,n min m1,n /qn , m2,n /pn . [5] M. A. Kashani, M. Safari, and M. Uysal, “Optimal relay placement
n=1 and diversity analysis of relay-assisted free-space optical communication
Figs. 9-10 illustrate the BER performance of the SIMO systems,” IEEE/OSA Journal of Optical Communications and Network-
FSO system under consideration. We present approximate ing, vol. 5, no. 1, pp. 37–47, 2013.
analytical results which have been obtained through (37) and [6] M. A. Kashani and M. Uysal, “Outage performance and diversity gain
analysis of free-space optical multi-hop parallel relaying,” IEEE/OSA
(40) along with the Monte-Carlo simulation of (33). As clearly Journal of Optical Communications and Networking, vol. 5, no. 8,
seen from Figs. 9-10, our approximate expressions provide an pp. 901–909, 2013.
10
[7] M. Karimi and M. Nasiri-Kenari, “Ber analysis of cooperative systems in [21] Z. Wang, W. Zhong, S. Fu, and C. Lin, “Performance comparison of
free-space optical networks,” Journal of Lightwave Technology, vol. 27, different modulation formats over free-space optical (fso) turbulence
no. 24, pp. 5639–5647, 2009. links with space diversity reception technique,” IEEE Photonics Journal,
[8] E. Jakeman and P. Pusey, “Significance of k distributions in scattering vol. 1, no. 6, pp. 277–285, 2009.
experiments,” Physical Review Letters, vol. 40, no. 9, pp. 546–550, 1978. [22] T. A. Tsiftsis, H. G. Sandalidis, G. K. Karagiannidis, and M. Uysal,
[9] L. Andrews and R. Phillips, “Mathematical genesis of the ik distribution “Optical wireless links with spatial diversity over strong atmospheric
for random optical fields,” Journal of Optical Society of America A, turbulence channels,” IEEE Transactions on Wireless Communications,
vol. 3, no. 11, pp. 1912–1919, 1986. vol. 8, no. 2, pp. 951–957, 2009.
[10] J. H. Churnside and S. F. Clifford, “Log-normal rician probability- [23] S. M. Navidpour, M. Uysal, and M. Kavehrad, “Ber performance of free-
density function of optical scintillations in the turbulent atmosphere,” space optical transmission with spatial diversity,” IEEE Transactions on
Journal of Optical Society of America A, vol. 4, no. 10, pp. 1923–1930, Wireless Communications, vol. 6, no. 8, pp. 2813–2819, 2007.
1987. [24] E. Stacy, “A generalization of the gamma distribution,” The Annals of
[11] M. Al-Habash, L. C. Andrews, and R. L. Phillips, “Mathematical Mathematical Statistics, vol. 33, no. 3, pp. 1187–1192, 1962.
model for the irradiance probability density function of a laser beam [25] A. Jeffrey and D. Zwillinger, Table of integrals, series, and products.
propagating through turbulent media,” Optical Engineering, vol. 40, Academic Press, 2007.
no. 8, pp. 1554–1562, 2001. [26] N. C. Sagias, G. K. Karagiannidis, P. T. Mathiopoulos, and T. A. Tsiftsis,
[12] A. Jurado-Navas, J. M. Garrido-Balsells, J. F. Paris, and A. Puerta- “On the performance analysis of equal-gain diversity receivers over
Notario, “A unifying statistical model for atmospheric optical scintilla- generalized gamma fading channels,” IEEE Transactions on Wireless
tion,” arXiv preprint arXiv:1102.1915, 2011. Communications, vol. 5, no. 10, pp. 2967–2975, 2006.
[13] N. D. Chatzidiamantis, H. G. Sandalidis, G. K. Karagiannidis, S. A. [27] M. A. Kashani, M. M. Rad, M. Safari, and M. Uysal, “All-optical
Kotsopoulos, and M. Matthaiou, “New results on turbulence modeling amplify-and-forward relaying system for atmospheric channels,” IEEE
for free-space optical systems,” in IEEE International Conference on Communications Letters, vol. 16, no. 10, pp. 1684–1687, 2012.
Telecommunications (ICT), pp. 487–492, 2010. [28] M. A. Kashani, M. Safari, and M. Uysal, “Optimal relay placement
[14] L. C. Andrews, R. L. Phillips, C. Y. Hopen, and M. Al-Habash, “Theory in cooperative free-space optical communication systems.,” in IEEE
of optical scintillation,” Journal of Optical Society of America A, vol. 16, WCNC, pp. 995–999, 2012.
no. 6, pp. 1417–1429, 1999. [29] G. Karagiannidis, T. Tsiftsis, and H. Sandalidis, “Outage probability of
[15] M. Uysal, J. Li, and M. Yu, “Error rate performance analysis of coded relayed free space optical communication systems,” Electronics Letters,
free-space optical links over gamma-gamma atmospheric turbulence vol. 42, no. 17, pp. 994–996, 2006.
channels,” IEEE Transactions on Wireless Communications, vol. 5, no. 6, [30] A. Prudnikov, Y. A. Brychkov, and O. Marichev, Integrals and series.
pp. 1229–1233, 2006. Volume 3: More special functions. Transl. from the Russian by GG
[16] E. Bayaki, R. Schober, and R. K. Mallik, “Performance analysis of mimo Gould. New York: Gordon and Breach Science Publishers, 1990.
free-space optical systems in gamma-gamma fading,” IEEE Transactions [31] V. Adamchik and O. Marichev, “The algorithm for calculating integrals
on Communications, vol. 57, no. 11, pp. 3415–3424, 2009. of hypergeometric type functions and its realization in reduce system,” in
[17] S. M. Flatté, C. Bracher, and G.-Y. Wang, “Probability-density functions Proceedings of the international symposium on Symbolic and algebraic
of irradiance for waves in atmospheric turbulence calculated by numeri- computation, pp. 212–224, 1990.
cal simulation,” Journal of Optical Society of America A, vol. 11, no. 7, [32] A. Garcı́a-Zambrana, B. Castillo-Vázquez, and C. Castillo-Vázquez,
pp. 2080–2092, 1994. “Asymptotic error-rate analysis of fso links using transmit laser selection
[18] R. J. Hill and R. G. Frehlich, “Probability distribution of irradiance for over gamma-gamma atmospheric turbulence channels with pointing
the onset of strong scintillation,” Journal of Optical Society of America errors,” Optics express, vol. 20, no. 3, pp. 2096–2109, 2012.
A, vol. 14, no. 7, pp. 1530–1540, 1997. [33] Wolfram Research Inc., “The wolfram functions site.”
[19] J. H. Churnside and R. Hill, “Probability density of irradiance scintilla- http://functions.wolfram.com.
tions for strong path-integrated refractive turbulence,” Journal of Optical [34] M. Chiani, D. Dardari, and M. K. Simon, “New exponential bounds
Society of America A, vol. 4, no. 4, pp. 727–733, 1987. and approximations for the computation of error probability in fading
[20] H. E. Nistazakis, T. A. Tsiftsis, and G. S. Tombras, “Performance channels,” IEEE Transactions on Wireless Communications, vol. 2, no. 4,
analysis of free-space optical communication systems over atmospheric pp. 840–845, 2003.
turbulence channels,” IET communications, vol. 3, no. 8, pp. 1402–1409,
2009.

Kashani2015

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Kashani2015

Uploaded by

Copyright:

Available Formats

This article has been accepted for publication in a future issue of this journal, but has not been

A Novel Statistical Channel Model for

Fig. 5. Pdfs of the scaled log-irradiance for a spherical wave assuming

0.9280 where p and q are chosen as 17 and 7 respectively.

0.06 and l0 /R0 = 0. The parameters of Double GG are evalu-

IV. P ERFORMANCE E VALUATION

a SISO FSO link over moderate and strong atmospheric tur-

You might also like