6234 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 67, NO.

9, SEPTEMBER 2019

Level Crossing Rate and Average Outage Duration

of Free Space Optical Links
Chaouki Ben Issaid and Mohamed-Slim Alouini , Fellow, IEEE

Abstract— The level crossing rate (LCR) and the average the packet error rate (PER) [4]– [5]. Second-order statistics
outage duration (AOD) are two important second order statistics are also more meaningful when analyzing the performance
that allow a deeper understanding of the behavior of the channel. of a communication system in a cellular environment with
In this paper, we study these metrics in order to assess the
performance of free space optical (FSO) communication links in co-channel interference where the AOD is more important
the presence of weak atmospheric turbulence and rice-induced than the probability of outage itself [6]. In [4], the authors
pointing errors. More specifically, we derive an integral and a derived the expression of the LCR for a multi-antenna system
Gauss–Laguerre quadrature representation for both the LCR employing maximum ratio combining (MRC) in the presence
and the AOD in the single hop case and for their respective of additive white Gaussian noise and co-channel interferers
bounds in the multihop case. Selected numerical simulations are
presented to show the accuracy of the derived results and to study with unequal received powers and Doppler shifts. The LCR
the effect of certain system parameters on these two performance expressions were obtained for both spatially uncorrelated and
metrics. correlated MRC systems. The authors have investigated, in [7],
Index Terms— Level crossing rate, average outage duration, the PER of interference-limited mobile packet systems using
free-space optical communication, Gauss–Laguerre quadrature, the Markov model for infinitesimal interval and finite-time
multihop. interval. The state and state transition probabilities of the
Markov chains, used to model the time-varying signal-to-
interference ratio (SIR) over frequency-selective fading chan-
I. I NTRODUCTION
nels, are obtained from both the LCR and AOD of the SIR.

U LTRA-RELIABLE low latency communication (uRLLC)

services will play a key role in the fifth generation (5G)
networks and beyond [1]. In this context, it is important to
The PER analysis can be carried out for both the cellular
communication systems and ad-hoc mobile networks. In [8],
the authors derived expressions for the LCR and the AOD for
accurately characterize the average outage duration (AOD) of well-known multipath fading models such as Rayleigh, Rice,
wireless links. In fact, the temporal variations of the fading and Nakagami. They showed that these theoretical expressions
channels are characterized by the second order statistics, present a reasonable fit compared to the measured data. Using
in particular, the level crossing rate (LCR) and the AOD. Laplace approximation method, Krstić et al. determined the
These two important statistics are useful for the Markov LCR expression of the ratio of product of two κ − μ random
modeling of wireless channels [2], [3] and the design of error variables and Nakagami-m random variables [9]. They also
control codes and the type of diversity to be used in wireless investigated the effect of the fading parameters on the LCR
communication systems to combat the effect of fading, and of a wireless relay communication system with two sections
therefore they have been the topic of many works in the operating over κ − μ multipath fading environment in the
literature. Since it can change with time because of user presence of cochannel interference subjected to Nakagami-m
mobility, fading can be represented as a random process. fading. In [10], the authors derived the second order statistics,
Examining the statistical characteristics of fading time inter- i.e. the LCR and the AOD, of the received signal enve-
vals is very critical for designing wireless communications lope and the channel capacity under Inverse Gaussian and
systems. In fact, studying the LCR and the AOD is of major Nakagami-Inverse Gaussian fading distributions. They have
importance when it comes to evaluating adaptive communi- studied the impact of the distribution parameters on the derived
cation systems in terms of the optimum packet length and metrics in the context of radio frequency as well as free space
optical (FSO) communication systems. Novel exact integral
Manuscript received November 26, 2018; revised April 1, 2019; accepted expressions for the LCR and the AOD of κ − μ shadowed fad-
May 14, 2019. Date of publication May 22, 2019; date of current ver-
sion September 16, 2019. This work was supported by the KAUST ing channels are derived in [11]. The derived expressions pro-
Office of Sponsored Research. The Associate Editor for this paper was vided an accurate representation of the second order statistics
C. Olaverri-Monreal. (Corresponding author: Chaouki Ben Issaid.) of the shadowed fading of empirical data collected for cellular
The authors are with the Computer, Electrical, and Mathematical Sciences
and Engineering Division, King Abdullah University of Science and Technol- D2D and bodycentric fading channels. Analytical expressions
ogy, Thuwal 23955, Saudi Arabia (e-mail: [email protected]; of the LCR and the AOD for the maximal ratio combining
[email protected]). (MRC) diversity operating on: (i) non-identical Rayleigh and
Color versions of one or more of the figures in this article are available
online at http://ieeexplore.ieee.org. (ii) independent and identical Ricean were derived in [12].
Digital Object Identifier 10.1109/TCOMM.2019.2918324 The authors also presented accurate approximations for both
0090-6778 © 2019 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.

Authorized licensed use limited to: Universidad de Malaga. Downloaded on October 30,2020 at 11:26:57 UTC from IEEE Xplore. Restrictions apply.
BEN ISSAID AND ALOUINI: LCR AND AOD OF FSO LINKS 6235

cases in the equal gain combining (EGC) diversity scheme.

Yang et al. provided closed-form expressions for the AOD of
multihop regenerative communication systems over Rayleigh,
Nakagami, or Rician fading in [13]. Two type of systems were
studied: (i) noise-limited, and (ii) interference-limited systems.
In [14], Velkov et al. studied the second order statistics of the
amplify-and-forward multihop Rayleigh fading channel. Using
the multivariate Laplace approximation theorem, they derived
closed-form approximations for the LCR and the AOD. In this
work, we derive integral and Gauss Laguerre expressions for
the LCR and the AOD of FSO systems in the presence of
weak turbulence and under the assumption that the pointing
error has non-zero boresight. We also provide bounds for the
LCR and AOD in the multihop case.
The reminder of this paper is organized as follows. We start
by describing the system model in Section II. We then derive,
in Section III, exact integral expressions for the LCR and
AOD in our particular set-up. We also give a Gauss-Laguerre
quadrature representation for the computation of the LCR.
In Section IV, we show some selected numerical simulations
to show the accuracy of our approach compared to Monte
Fig. 1. Illustration of laser beam in the detector plan in the presence of
Carlo (MC) simulations and to investigate the effect of some pointing errors.
of the system parameters on both metrics. Finally, the paper
ends with a summary of the main results.

II. S YSTEM M ODEL where σẋ2 is the variance of the time derivative of x.
From [17, 6.6-21], we have E[ẋ] = 0 and the variance of ẋ is
In this work, we consider a composite fading model where given by [17, 6.6-20]
the fading h can be expressed as h = hl ha hp . The path loss
hl is assumed to be equal to unity, the random processes ha V[ẋ] = η 2 = |R̈(0)|, (4)
and hp represent the turbulence fading and the pointing errors,
respectively, and are assumed to be independent. where R̈(·) is the second derivative with respect to (w.r.t)
time of the autocorrelation function R(·) of the process x.
A. Pointing Errors Similarly, we have E[ẏ] = 0 and V[ẏ] = η 2 . In the remainder
Pointing errors are errors due to the misalignment between of this paper, we consider the following generic autocorrelation
the transmitter and the receiver. Let z be the distance between function [17, Sec 6.5]
the transmitter and the detector, and we assume that the  
|τ |
detector is circular with an aperture radius a, as shown R(τ ) = 2σ 2 exp −Γ cos(Ωτ ), (5)
2
in Fig. 1.
We consider the complex representation of the displacement where Γ and Ω are the damping coefficient and the resonance
2
as frequency, respectively. In this case, η 2 = 2σ 2 |Ω2 − Γ4 |.
In this work, we assume that the envelope r and the
d(t) = x(t) + jy(t) = r(t) exp(jφ(t)), (1)
phase φ are slow variable functions of the time. Following the
where x(t) and y(t) are respectively, the horizontal and derivations of [18, Sec. 6.4], the joint distribution of r and ṙ
vertical displacement of the beam in the detector plane. is given by [18, Eq. (6-62)]
We assume that x and y are independent, and that they have  2   2   
a normal distribution with the same variance σ 2 and different r r + s2 ṙ rs
frṙ (r, ṙ) =  exp − 2
exp 2
I0
means μx and μy , respectively [15]. Therefore, for a fixed t, 2
2πη σ 2 2σ 2η σ2
the radial displacement r(t) = x(t)2 + y(t)2 follows a Rice (6)
distribution
 2    Assuming that the transmitter and the receiver plans are
r r + s2 rs
fr (r) = 2 exp − I0 , (2) parallel and that the laser beam is perpendicular to the receiver
σ 2σ 2 σ2
area, the expression of the pointing error loss factor can be
where s2 = μ2x + μ2y and I0 (·) is the 0th order mod- approximated by [19] as
ified Bessel function of the first kind [16, Sec. 8.431.1].  
From [17, Sec. 6.6], we know that ẋ is also Gaussian with 2r2
hp  A0 exp − 2 , (7)
probability density function (PDF) wzeq
 
1 ẋ as long as waz > 6, where wz is the beamwidth, wzeq is the
fẋ (ẋ) =  exp − 2 , (3)
2πη 2 2η equivalent beam width and A0 is the fraction of the collected

Authorized licensed use limited to: Universidad de Malaga. Downloaded on October 30,2020 at 11:26:57 UTC from IEEE Xplore. Restrictions apply.
6236 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 67, NO. 9, SEPTEMBER 2019

power at r = 0. In this case, the distribution of hp is [19] III. LCR AND AOD C OMPUTATION
 γ 2 The LCR LH (·) and the AOD TH (·) are two important
γ2 hp
fhp (hp ) = × , 0 ≤ hp ≤ A0 , (8) second order statistics that describe the random process behav-
hp A0 ior. In fact, the LCR represents the average number of cross-
wzeq
where γ = 2σ .
ings of the signal above a certain threshold h0 per unit of time
and is defined as [18, Eq. (2-73)]
 +∞
B. Atmospheric Turbulence
LH (h0 ) = ḢfH Ḣ (h0 , Ḣ)dḢ. (14)
Several models are proposed in the literature to describe the 0
atmospheric turbulence. These models are based on the PDF of The AOD is the average duration during which the signal
the fluctuation of light intensity or irradiance. The lognormal stays below a certain threshold h0 and its expression is given
probability density function is the most common distribution, by [18, Eq. (2-75)]
used in the case of the weak turbulence regime [20]. There-
fore, the PDF of ha is given by [19] FH (h0 )
TH (h0 ) = , (15)
  2  LH (h0 )
1 1 log(ha ) − μ
fha (ha ) = √ exp − , (9) where FH (h0 ) = P (H ≤ h0 ) is the CDF of H evaluated
2πha σR 2 σR
at h0 .
where σR2
is the Rytov variance for a plane wave and the mean In the remainder of this section, we will be interested
σ2 in determining the LCR and the AOD of H. The CDF of
is given by μ = − 2R .
H presents a lower bound for the outage probability of the
multihop system (see [23] and Section IV of [24] for more
C. Multihop Relaying details on how to obtain this lower bound). Since {ha,i }N
i=1
The coverage distance of FSO links is quite limited since are lognormal RVs, then Ha is also a lognormal RV with PDF
these links are severely harmed by a path loss. To increase the   2 
1 1 log(Ha ) − μ
coverage of these links and enhance their reliability, multihop fHa (Ha ) = √ exp − ,
2πHa σR 2 σR
relaying, where the signal is transmitted by means of inter-
mediate terminals, can reduce the effect of shadowing [21] (16)
and enable high data rate transmission [22]. In this section, 
N 
N 
N
we consider amplify and forward multihop relaying, which where σR = σR,i and μ = μi = − 12 2
σR,i .
i=1 i=1 i=1
involves the harmonic mean of the instantaneous signal-to- As for Hp , we can re-write its expression as
noise ratio of the hops. More specifically, We consider a  
multihop with N FSO links in cascade for which ha,i is N 2σ 2
Hp = A0 exp − 2 R , (17)
modeled as a lognormal RV, for i = 1, . . . , N , with PDF wzeq
  2 
1 1 log(ha,i ) − μi N 2
 N 2
 N 2

ri xi yi
fha,i (ha,i ) = √ exp − , where R = σ = σ + σ . Since { xσi }N
i=1
2πha,i σR,i 2 σR,i i=1 i=1 i=1
(10) and { yσi }N
i=1 are 2N independent Gaussian RVs with mean
2
μx,i or μy,i and with unit variance, then the RV R has a
σR,i
where μi = − 2 and for i = 1, . . . , N , we have non-central chi-squared distribution
   2  2  N2−1 s√ 
2r2 1 σ R + s2 σ R
hp,i  A0 exp − 2 i , (11) fR (R) = exp − IN −1 R ,
wzeq 2 2σ 2 s2 σ
 (18)
where for i = 1, . . . , N , ri (t) = xi (t)2 + yi (t)2 follows the
Rician PDF 
N
 2   where s2 = s2i and IN −1 (·) is the (N − 1)th order
ri r + s2 ri si  i=1
fri (ri ) = 2 exp − i 2 i I0 , (12) modified Bessel function of the first kind [16, Sec. 8.431.1].
σ 2σ σ2 Differentiating R w.r.t time, we get
where xi ∼ N (μx,i , σ 2 ), yi ∼ N (μy,i , σ 2 ) and s2i = μ2x,i + N
μ2y,i , for i = 1, . . . , N . 2 
Ṙ = ri r˙i . (19)
Now, we assume that each fading channel hi can be written σ 2 i=1
as hi = ha,i hp,i , i = 1, . . . , N , and let H be defined as
Given {ri }Ni=1 , Ṙ is a linear combination of N independent
N
Gaussian RVs and thus it is also a Gaussian RV. From [11]
H= h i = Ha H p , (13) and [25], we know that ri and r˙i are independent and r˙i is a
i=1
zero mean Gaussian RV. Since we are assuming that xi and yi

N are independent and both have the same variance σ 2 , then we

where Ha = ha,i and Hp = hp,i . have σx2˙i = σy2˙i . Thus, the variance of r˙i is given by σr2˙i = η 2 ,
i=1 i=1

Authorized licensed use limited to: Universidad de Malaga. Downloaded on October 30,2020 at 11:26:57 UTC from IEEE Xplore. Restrictions apply.
BEN ISSAID AND ALOUINI: LCR AND AOD OF FSO LINKS 6237

for i = 1, . . . , N . Given {ri }N

i=1 , the mean and variance of Ṙ
is given by
N
2 
E[Ṙ] = ri E[r˙i ] = 0, (20)
σ 2 i=1
N
4  2 4η 2
V[Ṙ] = 4
ri V[r˙i ] = 2 R. (21)
σ i=1 σ
Therefore, we can write
 
σ σ 2 Ṙ2
f (Ṙ|R) = √ exp − 2 . (22)
2η 2πR 8η R

Since fRṘ (R, Ṙ) = f (Ṙ|R)fR (R), then using (18) and (22)
we have
   
N s2
σ N R 2 −1 exp − 2σ 2 σ 2 Ṙ2 Fig. 2. LCR as function of the level h0 for Ω = 4 and different values of Γ.
fRṘ (R, Ṙ) = √ exp − 2
4η 2πsN −1 8η R
  
R s√ 
× exp − IN −1 R . (23)
2 σ
Proposition 1: The joint distribution of H and its time
derivative Ḣ is given by (24), as shown at the bottom of the
next page.
Proof: See Appendix A. 
With the expression of fH Ḣ (H, Ḣ) at hand, the expression
of the LCR is given by Proposition 2.
Proposition 2: The LCR is given by (25), as shown at the
bottom of the next page.
Proof: See Appendix B. 
The expression of FH (h0 ) is given by the following
corollary
Corollary 1: The CDF of H is given by (26), as shown at
the bottom of the next page. Fig. 3. LCR as function of the level h0 for Γ = 1.5 and different values
Proof: See Appendix C.  of Ω.
Using Gauss-Laguerre quadrature [26, Eq. (22.2.13)],
Corollary 2: For the single hop case (N = 1), the joint
we can approximate the value of LCR in (25) and (15) by
  distribution of h and its time derivative ḣ is given by (30), as
N s2 shown at the bottom of the page 6, the LCR is given by (31),
wzeq exp − 2σ 2
n

LH (h0 ) ≈ N +5 wi g(yi ), (27) as shown at the bottom of the page 6, and the CDF of h is
2 2 πσR σ 2 sN −1 i=1 given by (32), as shown at the bottom of the page 6.
where yi is the ith root of Laguerre polynomial Ln (y) [26,
Eq. (25.4.45)], the function g(·) is defined as IV. N UMERICAL S IMULATIONS
N −1 2
 In this section, we investigate the accuracy of the derived
g(y) = y 2 e(1−γ )y 8η 2 y + βσR 2 w2
zeq
⎛ results for both the LCR and the AOD by means of MC
⎛   ⎞2 ⎞ simulations. A number of samples M = 107 has been
h0
⎜ 1 y + log − μ
⎠ ⎟
A0N
× exp ⎝− ⎝ ⎠ used to obtain the numerical results. Unless stated otherwise,
2 σR we consider a dual-hop system, i.e. N = 2.
  We investigate the behavior of the LCR as function of the
wzeq s √ level h0 and for different values of the damping coefficient and
× IN −1 √ y , (28)
2σ 2 resonance frequency in Figures 2 and 3. In fact, we consider a
and the weight wi is given by [26, Eq. (25.4.45)] fixed value of the resonance frequency Ω = 4 and we plot the
yi LCR for different values of the damping coefficient Γ in Fig. 2.
wi = . (29) The case Γ = 0 correspond to an oscillatory autocorelation
(n + 1)2 [Ln+1 (yi )]2
model. The effect of the resonance frequency on the LCR
A special case is when N = 1 which corresponds to a single for a fixed damping coefficient Γ = 1.5 is plotted in Fig 3.
hop FSO system. The following corollary give the expressions An exponential decay autocorrelation model is assumed when
of fhḣ (·), the LCR as well as the CDF of h. Ω = 0. We can see that in both plots, the LCR reaches its

Authorized licensed use limited to: Universidad de Malaga. Downloaded on October 30,2020 at 11:26:57 UTC from IEEE Xplore. Restrictions apply.
6238 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 67, NO. 9, SEPTEMBER 2019

Fig. 4. AOD as function of the level h0 for Ω = 4 and different values

of Γ.
Fig. 6. LCR as function of the level h0 for Ω = 2 and Γ = 1.5 and different
values of n.

AOD in Figures 4 and 5. In all these plots, we can see that

the derived results match the numerical result obtained using
MC simulations. In Fig. 6, we plotted the LCR for Ω = 2 and
Γ = 1.5 while varying the number of quadrature points in the
Gauss-Laguerre representation. As we can see, as n increases,
the representation becomes more accurate and we can see that
a relatively small number n = 40 is sufficient to give a good
approximate for the LCR.
In Fig.7 (respectively Fig. 8), we plot the LCR (respectively
the AOD) as function of the level h0 for different fading
environments (i.e. different values of the parameter s). We can
Fig. 5. AOD as function of the level h0 for Γ = 1.5 and different values see that for low levels h0 , the LCR is higher as the value of s
of Ω. is larger, whereas for high levels h0 , the LCR increases as s
maximum around the mean value of the process as observed in becomes smaller. For high levels of h0 , the AOD of channels
similar studies done in different context. We studied the effect with higher values of s is higher than the channels with smaller
of the resonance frequency and the damping coefficient on the values of s.

 
N +2 s2
wzeq exp − 2σ 2
fH Ḣ (H, Ḣ) = N +5
2 2 πσR σ 2 sN −1 H 2 ⎛ ⎛  ⎞2 ⎞
  y+log H
−μ  
⎜ ⎠ ⎟ wzeq s √
2
wzeq Ḣ 2 AN
exp ⎝−12 ⎝
N −1
−γ 2 y
y 2 e exp − H 2 0
⎠ IN −1 √2σ2 y
 [ 2
16η y+2βwzeq 2 σ2
R ] σR
+∞
×  dy (24)
0 2 w2
8η 2 y + βσR zeq
 
N s2
wzeq exp − 2σ 2
LH (h0 ) = N +5
2 2 πσR σ 2 sN −1
⎛ ⎛   ⎞2 ⎞
 +∞  y + log H
− μ  
N −1
−γ 2
y 2 ⎜ 1⎝ AN
⎠ ⎟ wzeq s √
× 2 2
8η y + βσR wzeq exp ⎝− ⎠IN −1 √ 2 y dy
0
y 2 e (25)
0 2 σR 2σ
  ⎛ ⎛   ⎞2 ⎞
N +1
wzeq exp − 2σ s2  h0  +∞ −γ 2 y N −1 y + log H
− μ  
2 e y 2
⎜ 1 A0N
⎠ ⎟ wzeq s √
FH (h0 ) = N +2 √ exp ⎝− ⎝ ⎠ IN −1 √ y dydH
22 πsN −1 σ 2 σR 0 0 H 2 σR 2σ 2

(26)

Authorized licensed use limited to: Universidad de Malaga. Downloaded on October 30,2020 at 11:26:57 UTC from IEEE Xplore. Restrictions apply.
BEN ISSAID AND ALOUINI: LCR AND AOD OF FSO LINKS 6239

Fig. 9. LCR as function of the level h0 for Ω = 2 and Γ = 1.5 and different
Fig. 7. LCR as function of the level h0 for Ω = 2 and Γ = 1.5 and different values of N .
values of s.

Fig. 10. AOD as function of the level h0 for Ω = 2 and Γ = 1.5 and
Fig. 8. AOD as function of the level h0 for Ω = 2 and Γ = 1.5 and different values of N .
different values of s.
can see that the shape of the LCR is widened for larger values
of N . Similarly, the AOD of the channels, using the same
The effect of increasing the number of hops on the LCR parameters as in Fig. 9, is depicted in Fig. 10. We note that
is studied in Fig. 9. We can see that as the number of hops the AOD curves are monotonically increasing with the level
N increases, the LCR curves are shifted to the right. We also h0 . Also, the AOD tends to increase with the number of hops.

 
3 s2
wzeq exp − 2σ 2
fhḣ (h, ḣ) =
8πσ 2 σR h2
    h 2   
y+log −μ wzeq s √
2
2 wzeq ḣ2
e−γ y
exp − 12
A0
 exp − h2 I0 √ y
+∞ ][ 2 σ2
16η 2 y+2βwzeq R
σR 2σ2
×  dy (30)
0 2 w2
8η 2 y + βσR zeq
⎛ ⎛   ⎞2 ⎞
 +∞  y + log h0
− μ  
wzeq −γ 2 y 2 ⎜ 1⎝ A0
⎠ ⎟ wzeq s √
Lh (h0 ) = e 2 2
8η y + βσR wzeq exp ⎝− ⎠ I0 √ 2 y dy (31)
8πσ 2 σR 0 2 σR 2σ
  ⎛ ⎛   ⎞2 ⎞
2
wzeq s2
exp − 2σ2  h 0
 +∞ 2
e−γ y
H
⎜ 1 ⎝ y + log AN − μ ⎟ w s√ 
FH (h0 ) = √ exp ⎝ − 0 ⎠ ⎠ I0 √zeq y dydH (32)
5
2 2 πσ 2 σR 0 0 H 2 σR 2σ 2

Authorized licensed use limited to: Universidad de Malaga. Downloaded on October 30,2020 at 11:26:57 UTC from IEEE Xplore. Restrictions apply.
6240 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 67, NO. 9, SEPTEMBER 2019

V. C ONCLUSION where β = (2πσc )2 and σcis related to the 3-dB cutoff

In this work, we have introduced an integral and a Gauss- frequency according to fc = 2 log(2)σc .
Laguerre representation for the LCR and the AOD of FSO The joint PDF of H and Ḣ is derived using the following
links subject to weak turbulence and in the presence of Rice expression [29]
induced pointing errors. The proposed approach simplified the fH Ḣ (H, Ḣ)
triple integral expression of the LCR to a finite sum, resulting  +∞  +∞  
in a simpler and a faster evaluation of the LCR. We have also 1 H Ḣ Ḣa H
= f
2 Hp Ḣp
, −
provided bounds for the LCR and AOD of the mutihop case. H
N −∞ Ha Ha Ha Ha2
A0
Using these results which were validated with MC simulations,
we have investigated the effect of the resonance frequency, × fHa Ḣa (Ha , Ḣa )dḢa dHa
     N2 −1
the damping coefficient of the radial displacement as well as N +2 s2  +∞  AN
0 Ha
wzeq exp − 2σ 2 +∞ log H
the number of hops on both the LCR and the AOD. = N +9 √
2
2 2 N −1 H 2
2 π βπησ σR s H
0 Ha2
A PPENDIX A AN
⎛ 0
2 ⎞
P ROOF OF P ROPOSITION 1  −γ 2
AN0 Ha 1 Ḣa
Proof: First, we start by recalling the expression of R as × exp ⎝− ⎠
H 2β Ha σR
function of Hp 
 N  2 
2
wzeq A0 1 log(Ha ) − μ
R= log = g(Hp ). (A.1) × exp −
2σ 2 Hp 2 σR
⎛  2 ⎞
Differentiating w.r.t t, we get 2 Ḣ Ḣa
⎜ wzeq H − Ha  ⎟
2
wzeq Ḣp × exp ⎝− ⎠
Ṙ = g  (Hp )Ḣp = − 2
. (A.2) AN H
16η 2 log 0H a
2σ Hp
   
Using the Jacobian, the joint PDF of Hp and Ḣp can swzeq AN 0 Ha
be expressed as function of the joint PDF of R and Ṙ × IN −1 √ log dḢa dHa . (A.7)
2σ 2 H
[27, Chap. 5]
 
 ∂R ∂R  We start by integrating w.r.t Ḣa
  ⎛  2 ⎞
 ∂Hp ∂ Ḣp   2

fHp Ḣp (Hp , Ḣp ) =   f (R, Ṙ)  +∞ w 2 Ḣ
− Ḣa
 ∂ Ṙ ∂ Ṙ  RṘ ⎜ 1
exp ⎝−
Ḣa
−
zeq H
 N ⎟
Ha
 ∂H ⎠ dḢa
∂ Ḣp  2β H σ 2 A Ha
p −∞ a R 16η log 0
H
= [g  (Hp )]2 fRṘ (R, Ṙ). (A.3) ⎛ ⎞
2  +∞
wzeq Ḣ 2 2
Therefore, we can write = exp ⎝−  N ⎠ e−pḢa +2qḢa dḢa ,
A H
 2 16H 2 η 2 log 0H a
−∞
2
wzeq
fHp ,H˙p (Hp , Ḣp ) = 2
fR,Ṙ (R, Ṙ). (A.4) (A.8)
2σ Hp
2
where p = 1
2
2βHa2 σR
+
wzeq
AN Ha
 and q =
With the expression of the joint PDF of R and Ṙ in (23), 16η 2 Ha2 log 0
H
we can derive the joint PDF of Hp and Ḣp 2
wzeq Ḣ
 . Using [16, Eq. (3.462.2)], we get
fHp ,H˙p (Hp , Ḣp ) 16η 2 HHa log
AN Ha
0
  H

N +2
wzeq exp − 2σ s2   N  N2 −1 ⎛  2 ⎞
2 A0   2
= N +7 √ log +∞
⎜ 1
2
wzeq Ḣ
− Ḣa

 N ⎟
2 N −1 2 Ḣa H Ha
2 2 πησ s Hp Hp exp ⎝− − ⎠ dḢa
⎛ ⎞ −∞ 2β Ha σR 16η 2 log
A0 Ha
  N  2 2 H
A w zeq Ḣ p ⎛ ⎞
× exp −γ 2 log 0
exp ⎝−  N ⎠   
Hp A
16η 2 Hp2 log H0p log 0H a exp ⎝− 
AN H
2
wzeq Ḣ 2
 ⎠
    h2 16η 2 log
AN
0 H a 2 σ2
+2βwzeq
H R
swzeq AN0 =  .
× IN −1 √ log . (A.5) √ −1  N 
2σ 2 Hp 4 βπHa σR η 2 βσ 2 +8η 2 log A0 Ha
wzeq R H
The joint distribution of Ha and Ḣa is [28] (A.9)
⎛  2 ⎞
1 1 Ḣ a Therefore, we get
fHa Ḣa (Ha , Ḣa ) = √ exp ⎝− ⎠
2π β(Ha σR )2 2β Ha σR
fH Ḣ (H, Ḣ)
  2   
1 log(Ha ) − μ N +2 s2
wzeq exp − 2σ 2
× exp − , (A.6) =
2 σR N +5
2 2 πσR σ 2 sN −1 H 2

Authorized licensed use limited to: Universidad de Malaga. Downloaded on October 30,2020 at 11:26:57 UTC from IEEE Xplore. Restrictions apply.
BEN ISSAID AND ALOUINI: LCR AND AOD OF FSO LINKS 6241

  N  N2−1   2  √  
2 C0 η y N −γ 2 y wzeq s √
 +∞ log A0 Ha
exp − 2 1 log(Ha )−μ • φ(y) ∼ ζ(y) = √ √√ y 2 e exp √ y
H σR y→+∞ π y 2σ2
×    (where C0 is a constant independent of y and
H
AN Ha wzeq
N
2 βσ 2 + 8η 2 log A0 Ha I0 (x) ∼ exp(x)
√
2πx
[26, (9.7.1)]),
0 R H x→+∞
   The function y → φ(y) has a finite integral over [0, +∞[

•
 N −γ 2
A0 Ha swzeq AN0 H a  +∞
× IN −1 √ log dHa . φ(y)dy
H 2σ 2 H
0
(A.10)
 N   +∞  
A H 4ηC0 1 2 2 wzeq s
Performing the change of variable y = log 0H a , we get = √ z N + 2 e−γ z
exp √ z dz
the desired result. 
π 0 2σ 2
  N2 +34    s
A PPENDIX B 4ηC0 2σ 2 s2 3
= √ 2
e 4σ2 Γ N + D−N−32 −
P ROOF OF P ROPOSITION 2 π wzeq 2 σ
Proof: We recall the expression of the average LCR
Lh (H0 ) < +∞ (using [16, 3.462.1]),

LH (h0 ) where Dη (·) is the parabolic cylinder

  function [16, Eq. (9.240)].
 +∞ N +2
wzeq exp − 2σs2
2
Therefore, the integral I is convergent and we can interchange
= ḢfH Ḣ (h0 , Ḣ)dḢ = N +5
0 2 2 πσR σ s 2 N −1 h20 the integration order to obtain
   
N −1
−γ 2
y
2
wzeq Ḣ 2 N
wzeq exp − 2σ s2  +∞
 +∞  +∞ Ḣy 2 e exp −H 2 16η2 y+2βw2 σ2 2 N −1 2
[ zeq R ] LH (h0 ) = N +5 y 2 e−γ y
×  2 2 πσR σ s 2 N −1 0
0 0 8η 2 y +βσR2 w2
zeq
⎛ 
⎛   ⎞2 ⎞ 2 w2
h0   × 8η 2 y + βσR
⎜ 1 y +log AN −μ ⎟ zeq
⎠ ⎠ IN −1 w zeq s √
× exp ⎝− ⎝ 0
√ y
2 σR 2σ 2 ⎛ ⎛   ⎞2 ⎞
H
⎜ 1 y + log −μ
(B.1) × exp ⎝− ⎝
AN
0 ⎠ ⎟⎠
2 σR
In order to be able to interchange the order of integration,
we need to show that the integral w.r.t Ḣ then w.r.t y is  
convergent. We start by computing the integral w.r.t Ḣ to get wzeq s √
× IN −1 √ y dy. (B.5)
 +∞   2σ 2
2
wzeq Ḣ 2
Ḣ exp − 2  2 σ2
 dḢ 
0 h0 16η 2 y + 2βwzeq R
 
h20 8η 2 y + βwzeq
2 2
σR A PPENDIX C
= 2
. (B.2) P ROOF OF C OROLLARY 1
wzeq
Therefore, we need to show that the integral Proof: We start by deriving the PDF of Hp from the PDF
 +∞  of R. To this end, we recall the expression of R
N −1 2  N
I = y 2 e−γ y 8η 2 y + βσR 2 w2
zeq
2
wzeq A0
0 R= log = g(Hp ). (C.1)
⎛ ⎛   ⎞2 ⎞ 2σ 2 Hp
H
⎜ 1 y + log − μ
⎠ ⎟
A0N
The PDF of Hp is therefore given by
× exp ⎝− ⎝ ⎠
2 σR
fHp (Hp ) = fR (g(Hp ))|g  (Hp )|
 
wzeq s √  
× IN −1 √ y dy (B.3) N +1 s2  N −γ 2   N  N2−1
2σ 2 wzeq exp − 2σ 2 A0 A0
= N +3 log
is convergent. 2 2 sN −1 σ 2 Hp Hp Hp
To this  end, we introduce the function φ(y) =
N −1
−γ 2 y 2 w2 wzeq s √    
y 2 e 8η 2 y + βσR zeq IN −1
√
2σ2
y and we swzeq AN
0
can write × IN −1 √ log . (C.2)
 +∞ 2σ 2 Hp
I≤ φ(y)dy. (B.4) Using [19, Eq. (12)], the PDF of H is
0  +∞  
Since, we know that: 1 H
fH (H) = fH fHa (Ha )dHa . (C.3)
• y → φ(y) is a continuous function on [0, +∞[,
H
N
Ha p Ha
A0

Authorized licensed use limited to: Universidad de Malaga. Downloaded on October 30,2020 at 11:26:57 UTC from IEEE Xplore. Restrictions apply.
6242 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 67, NO. 9, SEPTEMBER 2019

Thus, we can write [15] F. Yang, J. Cheng, and T. A. Tsiftsis, “Free-space optical communication
  with nonzero boresight pointing errors,” IEEE Trans. Commun., vol. 62,
N +1
wzeq s2
exp − 2σ  +∞  N −γ 2 no. 2, pp. 713–725, Feb. 2014.
2 1 A0 Ha [16] I. S. Gradshteyn, I. M. Ryzhik, and A. Jeffrey, Table of Integrals, Series
fH (H) = N √ and Products, 7th ed. Amsterdam, The Netherlands: Elsevier, 2007.
2 2 +2 sN −1 πσR σ 2 H AHN Ha H
[17] P. Beckmann, Probability in Communication Engineering. San Diego,
0
  N  N2−1 CA, USA: Hartcourt Brace & World, 1967.
A0 Ha [18] W. C. Y. Lee, Mobile Communications Engineering. New York, NY,
× log USA: McGraw-Hill, 1982.
H
⎛ ⎛   ⎞2 ⎞
[19] A. A. Farid and S. Hranilovic, “Outage capacity optimization for free-
H space optical links with pointing errors,” J. Lightw. Technol., vol. 25,
⎜ 1 y + log − μ
⎠ ⎟
N
A0 no. 7, pp. 1702–1710, Jul. 2007.
× exp ⎝− ⎝ ⎠ [20] M. A. Al-Habash, L. C. Andrews, and R. L. Phillips, “Mathemati-
2 σR cal model for the irradiance probability density function of a laser
   
beam propagating through turbulent media,” Opt. Eng., vol. 40, no. 8,
pp. 1554–1562, Aug. 2001.
swzeq AN
0 Ha
× IN −1 √ log dHa . (C.4) [21] J. Boyer, D. D. Falconer, and H. Yanikomeroglu, “Multihop diversity
2σ 2 H in wireless relaying channels,” IEEE Trans. Commun., vol. 52, no. 10,
pp. 1820–1830, Oct. 2004.
  [22] A. K. Dinnis and J. S. Thompson, “Increasing high data rate coverage
AN
0 Ha
Performing the change of variable y = log H , in cellular systems using relaying,” in Proc. 60th Veh. Technol. Conf.,
the desired result is obtained.  vol. 5, Sep. 2004, pp. 3424–3428.
[23] C. B. Issaid, K.-H. Park, M.-S. Alouini, and R. Tempone, “Fast out-
age probability simulation for FSO links with a generalized pointing
error model,” in Proc. IEEE Global Commun. Conf. (GLOBECOM),
R EFERENCES Dec. 2016, pp. 1–7.
[24] C. B. Issaid, K.-H. Park, and M.-S. Alouini, “A generic simulation
[1] M. Bennis, M. Debbah, and H. V. Poor, “Ultra-reliable and low-latency approach for the fast and accurate estimation of the outage probability
wireless communication: Tail, risk and scale,” 2018, arXiv:1801.01270. of single hop and multihop FSO links subject to generalized pointing
[Online]. Available: https://arxiv.org/abs/1801.01270 errors,” IEEE Trans. Wireless Commun., vol. 16, no. 10, pp. 6822–6837,
[2] P. Sadeghi, R. A. Kennedy, P. B. Rapajic, and R. Shams, “Finite- Oct. 2017.
state Markov modeling of fading channels—A survey of principles and [25] W. C.-Y. Lee, “Statistical analysis of the level crossings and duration of
applications,” IEEE Signal Process. Mag., vol. 25, no. 5, pp. 57–80, fades of the signal from an energy density mobile radio antenna,” Bell
Sep. 2008. Syst. Tech. J., vol. 46, no. 2, pp. 417–448, Feb. 1967.
[3] W.-J. Wang, H.-C. Yang, and M.-S. Alouini, “Wireless transmission of [26] M. Abramowitz and I. A. Stegun, Handbook of Mathematical Functions:
big data: A transmission time analysis over fading channel,” IEEE Trans. With Formulas, Graphs, and Mathematical Tables. New York, NY, USA:
Wireless Commun., vol. 17, no. 7, pp. 4315–4325, Jul. 2018. Dover, 1972.
[4] A. O. D. Ali, C. M. Yetis, and M. Torlak, “Novel expressions and [27] H. Kobayashi, B. L. Mark, and W. Turin, Probability, Random Processes,
applications for the level crossing rate of maximal ratio combining in the and Statistical Analysis: Applications to Communications, Signal
presence of cochannel interferers,” IEEE Trans. Veh. Technol., vol. 66, Processing, Queueing Theory and Mathematical Finance. Cambridge,
no. 11, pp. 9793–9808, Nov. 2017. U.K.: Cambridge Univ. Press, 2011.
[5] A. O. D. Ali, C. M. Yetis, and M. Torlak, “Optimal packet length for [28] T. T. Tjhung and C. C. Chai, “Fade statistics in Nakagami-lognormal
throughput maximization in correlated multi-user channels,” in Proc. channels,” IEEE Trans. Commun., vol. 47, no. 12, pp. 1769–1772,
IEEE Int. Conf. Commun. (ICC), May 2016, pp. 1–6. Dec. 1999.
[6] R. Annavajjala and J. Zhang, “Level crossing rates and average out- [29] A. Krantzik and D. Wolf, “Distribution of the fading-intervals of
age durations of SINR with multiple co-channel interferers,” in Proc. modified Suzuki processes,” in Signal Processing V: Theories and Appli-
MILCOM Mil. Commun. Conf., Oct./Nov. 2010, pp. 1233–1238. cations, L. Torres, E. Masgrau, and M. A. Lagunas, Eds. Amsterdam,
[7] K. Fukawa, H. Suzuki, and Y. Tateishi, “Packet-error-rate analysis using The Netherlands: Elsevier, 1990, pp. 361–364.
Markov models of the signal-to-interference ratio for mobile packet
systems,” IEEE Trans. Veh. Technol., vol. 61, no. 6, pp. 2517–2530,
Jul. 2012.
[8] A. Abdi, K. Wills, H. A. Barger, M.-S. Alouini, and M. Kaveh, Chaouki Ben Issaid was born in Sfax, Tunisia.
“Comparison of the level crossing rate and average fade duration of He received the Diplôme d’Ingénieur degree from
Rayleigh, Rice and Nakagami fading models with mobile channel data,” l’École Polytechnique de Tunisie, La Marsa, Tunisia,
in Proc. Veh. Technol. Conf. Fall, IEEE VTS Fall, 52nd Veh. Technol. in 2013, and the master’s degree in applied math-
Conf., vol. 4, Aug. 2000, pp. 1850–1857. ematics and computational science from the King
[9] D. Krstie, I. Romdhani, M. M. B. Yassein, S. Minic, G. Petkovic, Abdullah University of Science and Technology
and P. Milacic, “Level crossing rate of ratio of product of two k-μ (KAUST), Thuwal, Saudi Arabia, where he is cur-
random variables and Nakagami-m random variable,” in Proc. IEEE Int. rently pursuing the Ph.D. degree in statistics. His
Conf. Comput. Inf. Technol., Ubiquitous Comput. Commun., Depend- current research interests include efficient Monte
able, Autonomic Secure Comput., Pervasive Intell. Comput., Oct. 2015, Carlo simulations for the performance of wireless
pp. 1620–1625. communication systems.
[10] I. Trigui, A. Laourine, S. Affes, and A. Stephenne, “The inverse
Gaussian distribution in wireless channels: Second-order statistics
and channel capacity,” IEEE Trans. Commun., vol. 60, no. 11, Mohamed-Slim Alouini (S’94–M’98–SM’03–F’09)
pp. 3167–3173, Nov. 2012. was born in Tunis, Tunisia. He received the Ph.D.
[11] S. L. Cotton, “Second-order statistics of κ-μ shadowed fading channels,” degree in electrical engineering from the California
IEEE Trans. Veh. Technol., vol. 65, no. 10, pp. 8715–8720, Oct. 2016. Institute of Technology (Caltech.), Pasadena, CA,
[12] X. Dong and N. C. Beaulieu, Average Level Crossing Rate and Average USA, in 1998. He served as a Faculty Member of the
Fade Duration of Diversity Methods. Boston, MA, USA: Springer, 2003, University of Minnesota, Minneapolis, MN, USA,
pp. 181–203. and Texas A&M University at Qatar, Education
[13] L. Yang, M. O. Hasna, and M.-S. Alouini, “Average outage duration of City, Doha, Qatar, before joining the King Abdullah
multihop communication systems with regenerative relays,” IEEE Trans. University of Science and Technology (KAUST),
Wireless Commun., vol. 4, no. 4, pp. 1366–1371, Jul. 2005. Thuwal, Makkah, Saudi Arabia, as a Professor of
[14] Z. Hadzi-Velkov, N. Zlatanov, and G. K. Karagiannidis, “Level crossing electrical engineering in 2009. His current research
rate and average fade duration of the multihop Rayleigh fading channel,” interests include the modeling, design, and performance analysis of wireless
in Proc. IEEE Int. Conf. Commun., May 2008, pp. 4451–4455. communication systems.

Authorized licensed use limited to: Universidad de Malaga. Downloaded on October 30,2020 at 11:26:57 UTC from IEEE Xplore. Restrictions apply.