2016 International Symposium on Signal, Image, Video and Communications (ISIVC)

Performance analysis of PPM modulation in

multihop FSO system over gamma gamma channel
Hechmi Saidi Kamel Tourki Noureddine Hamdi
SYSCOM Research Lab, ENIT, Mathematical and Algorithmic Sciences Lab, SYSCOM Research Lab, ENIT,
Manar University France Research Center, Manar University
Email: hechmi.saidi@rns.tn Huawei Technologies Co. Ltd. Email: noureddine.hamdi@ept.rnu.tn
Email: Kamel.tourki@huawei.com

AbstractAtmospheric turbulence induced fading is one of the
main impairments affecting the operation of free-space optical
(FSO) communication systems. In this work, the average symbol
error rate (SER) of M-ary pulse position modulation (PPM)
is analyzed under Gamma-Gamma fading channel. Using the
characteristics of Meijer G function, Closed-form expression
for average SER is derived. In this paper, we analyze the
performance of detect and forward relay or/and amplify and for-
ward three-hop FSO systems over strong atmospheric turbulence
channels. In our analysis, turbulence channels are modeled by
Gamma-Gamma distribution and the AF transmission is based
on optical amplify-and forward (OAF) technique. The DF relay
refers to detect and forward protocol. We look for symbol-error
probability (SEP) of dual-hop PPM modulated FSO systems
taking into account the effects of various noises as well as path
loss and geometric spreading of optical beam. The numerical
results show that strong turbulence has a severe effect on the
performance of FSO systems. we employ PPM modulation to
improve the systems performance.
Index Terms- FSO, DF, AF, multihop, PPM modulation,
Gamma-Gamma fading, SNR, SER
I. INTRODUCTION
Wireless Optical Communication or Free Space Optic (FSO)
is a complementary technology to microwave and optical
ber to meet the growing needs in telecommunications at
high data rates. In 2008, the rst FSO link with 10 Gbps
was introduced on the market, making it the fastest wireless
technology available in the market.
A. Related Works
FSO systems have these days received much attention
in rstmile access environments owing to their advantages
including unregulated bandwidth, high security, and ease of
installation [1] . FSO systems are, however, conned to short-
haul applications since their reliability is degraded due to the
distance dependent atmospheric turbulence and channel loss
[2]
Wireless relaying systems have drawn a signicant attention
in radio frequency (RF) wireless communications due to their
strong potential to increase the coverage area and quality-of-
service (QoS). Relays can be deployed to extend the cov-
erage of networks to areas where it is either technically or
economically unfeasible to set up fully edged base stations
(desert, offshore, etc.). Furthermore, relaying can help to
decrease interference and thereby improve the overall net-
work performance: If the number of available relays is large
compared to the number of base stations, mobile users will
most probably be closer to one of the relays than to a base
station and therefore require less transmit power, leading to
less interference. As a result, relaying techniques have already
found application in several wireless standards.
In reality, multi-hop transmission has been introduced as
a promising technique to improve FSO links coverage and
reliability [3] , [7] , [5] , [6] and [4] , but also to treat parallel
and serial cooperative communication using OOK modulation
in strong turbulence channel [19] or BPSK technique focusing
either in log-normal or gamma gamma channel [20] . By
subsequently deploying of smaller multiple hops (i.e., is the
distance between two relays), more reliable FSO transmission
over longer distances can be achieved. Moreover, multi-hop
techniques can support an optical connection between two
buildings which do not have a line-of-sight. Most of the
existing studies focused on conventional electrical relaying
[21] ,[20] and [19] , where amplifying or decoding process
is done in electrical domain [7] . With the aim of allow-
ing efcient high-speed transmission without the need for
complex optoelectronics and electronic processing at each
relay, alloptical relaying was rst proposed by S. Kazemlou
et al [8] . Recently, in [21] , the authors had focused in the
BER performance of multiple-input multiple-output (MIMO)
FSO systems using OOK modulation technique over Double
Generalized Gamma distribution channel.
B. Contributions
In this work, we will study serial relaying for single input
single output FSO systems. More specically, we aim to ana-
lyze the performance of hybrid OAF or/and DF three-hop FSO
systems over strong atmospheric turbulence channels. In our
analysis, turbulence channels are modeled by Gamma-Gamma
distribution and three-hop transmission. We investigate in
symbol error probability (SEP) of pulse position modulation
(PPM) three-hop FSO systems taking into account the effects
of various noises as well as path loss and geometric spreading
of optical beam. The numerical results show the comparison of

978-1-5090-3611-0/16/$31.00 2016 IEEE 207 ISIVC2016

the performance of FSO systems schemes. In addition, back- optical AF or DF. For the DF relay, it just detects the signal
ground noise and thermal noise at the destination are dominant and retransmits it. For the Optical AF one, Optical signals at
compared to other terms of noise from the source and the relay. its input include not only data signal but also background light
Thanks to OAF technique, the required transmitted power is with the power of Pb . At the relay, these signals are amplied
reduced accordingly to the amplier gain thus the transmission by an optical amplier with the gain of G and then transmitted
distance can be extended. For DF relaying, we focus just to the destination together with ASE noise generated by the
on "Detect and Forward" (no decoding and encoding) for optical amplier. ASE noise can be modeled as an additive
simplicity and because of the space limitations. We will only zero-mean white Gaussian noise with the optical power given
consider the indirect links through relays and neglect direct by [9] :
links between the source and the relay 2, the relay 1 and
PA = hp f (G 1)nsp B0 (1)
the destination but also between the source and destination
because the source is considered far from the destination and
where hp is Plancks constant, f is the frequency, nsp is
the transmitted power of the source signal small enough that
the amplier spontaneous emission parameter, and B0 is the
it barely reaches the relay 1. Therefore, there is no diversity
optical bandwidth.
at the reception: The destination receives only the signal from
At the destination, the total electric eld is converted to the
the relay 2. We also consider that each Relay is equiped with
photocurrent thanks to a photodetector (PD). Next, integrated
one antenna while having full knowledge of the channel. The
photocurrents over M time slots are compared at the PPM
noise is assumed to be centered Gaussian one. Performance
demodulator to nd the position of the slot with the highest
evaluation will be reected in the calculation of the symbol
current, which determines the transmitted symbol. Finally,
error probability for the PPM modulation which is equivalent
detected symbol is converted to the binary data.
on its side to SEP (Symbol Error probability) since we have a
Furthermore, we present the mathematical model of FSO
discrete transmission that is to say a transmission of a frame
channel with the state of h . In our model, h arises due to two
of bits. For the BER for PPM frame, it can be derived as
factors: channel loss (hl ) and atmospheric turbulence (ha ).
BER = 0.5M Pe
(M 1) , denoting Pe as the symbol error probability. The channel coefcient hence can be described as h = hl ha .
For this reason, we will only consider SER which means the
same thing in our case.
In order to model a channel with strong turbulence, Gamma- A. Turbulence channel model
Gamma distribution has been found to be a suitable choice
Several channel models have been assumed to model atmo-
since they are capable of modeling both small and large scale
spheric turbulence in literature.
uctuations.
As a generalization, Gamma-Gamma distribution is the best
In this paper, we are interested to deal with a three hops
one : It is in fact suitable in both strong and weak turbulence.
FSO system using PPM modulation technique and DF or/and
So, we will use it in our analysis and its PDF is given by [11]
AF relaying protocol as a solution to provide high bit rate
:
with best QoS. We will rstly present our system model   (+)
(+) 2 1
and secondly study each scheme performance and then plot
f (Ks ) = 2()
2 Ks
our simulations results comparing them by theoretical ones. ()()
 

 (2)
In fact the performance analysis will be done by choosing
K 2 K s
the appropriate sceme and calculate symbol error probability
(SEP). We close this paper by concluding.
where
II. S YSTEM MODEL    1
2
0.49R
= exp 12 7
1 (3)
(1 + 1.11R5 ) 6

   1
2
0.51R
= exp 12 5
1 (4)
(1 + 0.69R5 ) 6

Fig. 1: Three hop FSO system where and are the scintillation parameters that are de-
2 2
pendent on R , R is unitless Rytov variance, (.) is the
gamma function, and Kc (.) denotes the cth order modied
At the source, look at our model in gure 1, input data is Bessel function of the second kind. The scintillation index is
rst modulated at a PPM modulator, where each block of b related to and as
bits is mapped to one of M possible symbols (s0 , s1 , ..., sM )
converted into an optical pulse with constant power of Pt by 1 1 1
SC = + + (5)
a laser diode. In our work, we will use two types of relays:

208
B. Geometric Spreading and Channel Loss
It is easy to obtain the value of loss due to geometric spread- where c > 1 denotes the number of terms; and for any
ing as a function of the receiver area A and beam divergence i 1, 2, ...., c , xi is the it h root of Laguerre polynomial Lc (x)
angle . In addition, the path loss of laser power through the with degree c, wi is the corresponding weighting coefcient
atmosphere can be described by the exponential BeersLambert [10] , f (xi ) is the gamma-gamma atmospheric turbulence
Law. As a result, the total of geometric spreading and path loss distribution, M is the number of slots per frame, Kmin is
can be formulated as [11] : the minimum photon count in symbol signal slots and
A KbKmin Kb
hl = exp(d) (6) p0 (Kmin ) = e (8)
(d)2 Kmin !
where is attenuation coefcient, d is the distance.
(Ks + Kb )Kmin (Ks +Kb )
III. P ERFORMANCE ANALYSIS p1 (Kmin ) = e (9)
Kmin !
Our performance evaluation concerns the model of a sym-
metrical network ; we studied the system performance for 
K min
(Ks + Kb )j (Ks +Kb )
P1 (Kmin ) = e (10)
relatively remote positions of relays between source and j!
j=0
destination and thus considering only the relayed signals. We
considered a symmetrical network where the best position of 
K min
Kbj Kb
the relay 1 is when it is midway between the source and the P0 (Kmin ) = e (11)
j!
relay 2 and where the best position of the relay 2 is when it j=0
is midway between relay 1 and the destination. Here, Kb and Kb + Ks denote the mean values of photon
A. Serial relaying DF-DF scheme count in non-signal and signal slots, respectively.
We can consider that PSR e
1
= PRe 1 R2 = PRe 2 D = Pe
because we consider the same characteristics in each hop.
Subsequently: Pte = 1 (1 Pe )3

B. Serial hybrid relaying DF-AF scheme

Fig. 2: Three hop DF-DF FSO system

The three-hop DF FSO channel in gure 2 is modelled as

a concatenation of 3 links. Since each channel is assumed to
be independent, the end-to-end error probability and hence Fig. 3: Three hop DF-AF FSO system
the SEP for the three-hop DF relaying system is given by:
Pte = 1 (1 PSR
e
1
) (1 PRe 1 R2 ) (1 PRe 2 D )
Using the gure 3, we have Pte = 1 (1 PSR e
1
) (1
At the receiver side, we assume that the received photon PRe 1 R2 D ). e
Denoting PR1 R2 D as the symbol error probability
count per PPM slot follows a Poisson distribution. Also, a (or SEP) and IM = {1, . . . , M 1}, we assume that the
single hop FSO link can be modelled as a Binary Symmetric instantaneous symbol error probability can be expressed as
Channel (BSC) with an error probability Pe. We assume also
PRe 1 R2 D 1 P r{Is > In /n IM , s = s0 } 
that we work in a clear atmosphere conditions and also both
  

the transmitter and receiver are xed and perfectly aligned. M21 0 0 f (hra1 r2 )f (hra2 d )erf c SN R
2 (12)
Scintillation or channel fading, however, is considered in our
FSO performance derivation. In that case, the SEP is given by dhar2 d dhra1 r2
[14] after slight modications: 1) PPM average SEP analysis in atmospheric turbulence
 
M 1 M 1 channel optical AF dual-hop FSO system : The destination
Pe = Kmin =0 l=1 employs a photodetector to convert the received signal to the
l
c eKb photocurrent, which can be written as


i=1 wi Kmin ! (xi + Kb )
Kmin
f (xi )
P0 (Kmin )M 2 (1 P0 (Kmin ))l + Is = Pt Hsr Ghrd (13)
(7)

where  is the responsivity of the photodetector. The receiver

noises including shot noise and thermal noise are additive
p0 (Kmin )l P0 (Kmin 1)M 2
1
 1 
l+1 Gaussian noise with zero mean. Assuming that Be is the
1 electrical bandwidth of the photo detector, the total variance

209
of noises in the case of signal slot (s2 ) and non-signal slot Neglecting noise terms induced by the source and the relay,
(n2 ) can be expressed as SNR can be written as [16] :
4kB T Be
s2 = 2e(Pt hsr Ghrd + (Pb G + PA )hrd + Pb )Be + RL , (Pt hsr Ghrd )2
n2 = 2e[(Pb G + PA )hrd + Pb ]Be + 4kBRTLBe SNR = (17)
4ePb Be + 8k bT
RL B e
(14)
where e is the electron charge; kB is Boltzmann constant; T By substituting Eq. (17) in Eq. (16), expressing erfc(.) in-
is the absolute temperature; RL is the load resistance. Hence, tegrands as Meijers G-functions [12] , and based on [13]
the electrical signal to noise ratio (SNR) at the destination that (Eq.(07.34.21.0013.01)), a closed-form expression for the
uses PPM can be dened that symbol error probability is derived as Eq. (18):
Is2
SNR = (15)
s2 + n2 (M 1)2( 1 +1 +2 +2 5)
Pre = 5 G2,8
9,2
In this paragraph, we present the method to calculate the
1 )(1 )(2 )(2 )
2 (
32(Pt Ghsr l hrd l )2

symbol error probability of the all-optical AF dual-hop FSO 2k T B

system using PPM. Denoting Per as the symbol error prob- (ePb Be + bR e
)(1 1 2 2 )2
L
12 22 12 22 11 21 11 21

ability. We assume that the transmitted data is large enough 2 , 2 , 2 , 2 2 , 2 , 2 , 2 ,1
that the probabilities of sending any symbols are the same. 0, 12
Without the loss of generality, we also assume that symbol (18)
s0 is transmitted with received current of Is . By using union As an approximation to simplify following works , we can
bound technique, the upper bound to the instantaneous symbol take case of equality. Subsequently, we can consider R1 like
error probability can be expressed as an intermediate source for R1 R2 D which make PRe 1 R2 D = Pre
e
calculated in (18).For the same thing, PSR it can be appoxi-
Pre 1 P r{Is > In /n {1, . . 
. , M 1},
 s = s0 }
1
mated to Pe calculated in (7). Thus, Pt = 1(1Pre )(1Pe )
e
  

M21 0 0 f (hasr )f (hard )erf c SN R
2 dhard dhasr
(16) C. Serial hybrid relaying AF-DF scheme
where s represents the transmitted symbol. In is the It is the same thing for DF-AF scheme, we have just to
current corresponding to non-signal slot and erfc (. ) is the reverse roles.
complementary error function.

The derivation of closed-form expression for SEP taking
into account all noises is very complicated. In many previous
studies, such as [15] and [4] , noise is assumed as additive
white Gaussian noise (AWGN) for the sake of simplicity. In
fact,the performance of OAF relaying FSO systems over strong
atmospheric turbulence channels has been recently studied [20]
,[16] ,[19] ,[18] and [17] . Nevertheless, due to the analysis
complexity, these studies are limited, case of PPM modulation,
to the special case of dual-hop or parallel OAF relaying
systems. To the best of authors knowledge, the performance
of generalized case of multihop OAF relaying FSO systems
with an arbitrary number of nodes under the impact of strong
atmospheric turbulence has not yet been investigated in the
literature.
In our analysis we assume that we work in a clear atmo-
sphere conditions and also both the transmitter and receiver
are xed and perfectly aligned. Scintillation or channel fading,
however, is considered in our FSO performance derivation.
In what follows, the closed-form expression can be obtained
by considering only the effect of noises induced by the
destination: we only consider background noise and thermal
at the destination. In fact, we can prove that the impact
of background noise and thermal noise at the destination
is dominant in comparison with other noise terms from the
source and the relay. This explains why the performance of
systems in general case and our approximation are the same.
D. Serial hybrid relaying AF-AF scheme
We should calculate the SNR from start to nish and then
calculate the distribution function and thus extract the symbol
error rate for PPM modulation. In amplied relaying, each
intermediate terminal simply amplies the received signal
from the immediately preceeding terminal. Due to this, the
noise also gets amplied by each intermediate terminal and
hence is propagated end to end. Any error due to decoding
thus is present only at the end receiver and the delay due to
relaying by the intermediate terminals is minimized. At each
hop, the received signal plus noise is amplied. Hence, the
received SNR is same as the transmitted SNR. Noise gets
added to the signal at each hop by No. For N hops, the
received signal can be expressed as: SN Rith Hop = (i.N SN R
0)
for any i {1, 2, 3}, Hence, the instantaneous symbol error
probability can be expressed as
e
PSR1 R2 D
1  P r{I {1, . . . , M 1}, s = s0 }
 s > In /n sr
M 1
2 0 0 0 f (ha 1 )f (hra1 r2 )f (hra2 d )

SN R
erf c 6 dhar2 d dhra1 r2 dhsr
a
1
(19)
where s represents the transmitted symbol. In is the current
corresponding to non-signal slot and erfc (. ) is the comple-
mentary error function.

210
 IV. S IMULATIONS RESULTS
In this section, we numerically investigate the SER of the
DF or/and AF serial relaying FSO system using PPM under
gamma gamma atmospheric turbulence distribution. For a fair
comparison with other systems, BER is considered as the
function of the transmitted power per bit, which is given by
Ps = (M 1ogPt
. the power gain is 9 dB. For that reason, we will adopt for
2M )
We also assume that each hop is equidistant. For amplify and
forward protocol, we xed amplier gain at G = 10dB with
Kb = 1. For gamma gamma turbulence,
R {0.25, 0.75}(respectively weak and stron turbulence).
The whole distance between source and destination is L =
3Km
Fig. 4: Strong VS weak turbulence for 8PPM FSO system
under gamma gamma turbulence channel
We see rstly due to the gure 4 that simulation values
and theoretical ones are approximately the same. Besides,
when the SNR(or transmitted power per bit) is growing, the
SER is deacreasing. These two results remain valid for all
plots illustrated on the following gures. Also, we deduce that
gamma gamma turbulence describes better the atmospheric
turbulence channel which are valid either for weak and strong
turbulence. For that reason, we are adopting gamma gamma
turbulence in our study to generalization and optimization.
Fig. 5: Comparison of M-ary PPM for three-hop DF relaying
FSO system
It is shown in the gure 5 that when the number of time
per frame M for M-ary PPM is increasing, the average SER
is decreasing. We make the comparison between the 8PPM
211
and BPPM ones in term of the required transmitted power
per bit. It is seen that the required transmitted-power per bit
of BPPM system is about 17 dBm at BER of 101 . By using
8PPM( the transmitted power per bit is equal to 11dBm when
SER = 101 ), higher M-ary PPM one(8 instead of 2) can
reduce the required transmitted-power per bit to -6 dBm, i.e.,
8PPM in what follows for optimization.
The gure 6 below demonstrates in fact that when the
number of hops increases, the average SER decreases. We
make the comparison between the single-hop FSO system
and multi-hop(dual and three-hop) FSO ones in term of the
required transmitted power per bit. It is seen that the required
transmitted-power per bit of single-hop FSO system is about
16 dBm at SER of 101 . By using 2 relays or 3 hops(
the transmitted power per bit is equal to 11 dBm when
SER = 101 ), multi-hop FSO one can reduce the required
transmitted-power per bit to -5 dBm, i.e., the power gain is
7.5 dB.
Fig. 6: Comparison of N hops DF relaying FSO system where
N from 1 to 3
We conclude Finally (look at the gure 7) that DF/DF
scheme is better than the DF/AF scheme in terms of SER.In
fact, AF is amlifying either signal and noise.
Fig. 7: DF/DF versus DF/AF three-hop relaying FSO system
V. CONCLUSION
In this paper, we have considered a multihop FSO commu-
nications and investigated the optimal DF or AF relays types
for serial relaying. We have changed relays types as AF or/and [17] Liang Yang et al., "Performance analysis of relay-assisted all-optical
DF ones and compared between obtained schemes networks FSO networks over strong atmospheric turbulence channels with pointing
errors," IEEE/OSA J. Lightw. Technol., vol. 32, no. 23, pp. 4011- 4018,
performances. Dec. 2014.
For performance analysis, we have calculated the average [18] Jin-Yuan Wang et al., "Performance analysis for free-space optical com-
symbol error probability function of signal-to-noise ratio. We munications using parallel all-optical relays over composite channels,"
IET Commun., vol. 8, Iss. 9, pp. 1437-1446, Jun. 2014.
have further discussed in simulations results the effect of relays [19] Theodoros A. Tsiftsis, Harilaos G. Sandalidis, George K. Karagiannidis,
types, placed equidistant along the path from the source to and Nikos C. Sagias "Multihop Free-Space Optical Communications Over
the destination, on the overall performance: DF/DF scheme Strong Turbulence Channels,"IEEE ICC 2006 proceedings., pp. 2755-
2759, Dec. 2006
communication proved more reliable than the DF/AFs one. [20] Thanh V. Pham and Anh T. Pham, "Performance Analysis of Amplify-
The biggest challenge for us was the choice of Gamma- Decode-and-Forward Multihop BPSK/FSO Systems using APD Receivers
Gamma channel for FSO link and M-ary pulse position over Atmospheric Turbulence Channels," IET Commun., pp. 1-28, Dec.
2013.
modulation to improve the whole systems performance which [21] Mohammadreza A.Kashani, Murat Uysal, and Mohsen Kavehrad "On
made our calculation harder. The great problem that represents the Performance of MIMO FSO Communications over Double General-
always a challenge for our multihop FSO system is the ized Gamma Fading Channels,"IEEE Commun., pp. 1-6, Feb. 2015
complexity of serial AF multihop mathematics analysis. The
research on FSO and relay technology is still at the beginning
and there are open questions in numerous directions. It is now
required to go deeper into the analysis and into the study of
the boost of its performance and we are looking forward that
this technology will have a bright future.
R EFERENCES
[1] D. J. Heatley et aI., "Optical wireless: The story so far," IEEE Commun.
Mag., vol. 36, no. 12, pp. 72-82, Dec. 1998.
[2] X. Zhu and J. M. Khan, "Free-space optical communication through
atmospheric turbulence channels," IEEE Trans. Commun., vol. 50, pp.
1293-1300, Aug. 2002.
[3] Akella J., Yuksel M., and Kalyanaraman S. - Error Analysis of Multi-
Hop Free-Space Optical Communication, Proc. of IEEE International
Conference on Commun., 2005. pp. 1777-1781.
[4] Safari M. and Uysal M. - Relay-assisted free-space optical communica-
tion, IEEE Trans. Wireless Comm. 7 (2008) 5441-5449.
[5] Datsikas C. K., Peppas K. P., Sagias N. C., and Tombras G. S. -
Serial free-space optical relaying communications over Gamma-Gamma
atmospheric turbulence channels, J. Opt. Commun. Netw. 2 (2010) 576-
586.
[6] Safari M., Rad M. M., and Uysal M. - Multi-hop relaying over the
atmospheric poisson channel: outage analysis and optimization," IEEE
Trans. Comm. 60 (2012) 817-829.
[7] E. Bayaki, D.S. Michalopoulos, R. Schober, "EDFA-based all-optical
relaying in free-space optical systems," IEEE Trans. Commun., vol. 60,
no. 12, pp. 3797-3807, Dec. 2012.
[8] S. Kazemlou, S. Hranilovic, and S. Kumar, "All-optical multihop
freespace optical communication systems," J. Lightw. Technol., vol. 29,
no. 18, pp. 2663-2669, Sept. 2011.
[9] Agrawal GP. Fiber-optic communication systems. A John Wiley and Sons,
2002.
[10] M. Abramowitz and I. A. Stegun, Eds., Handbook of mathematical
functions, 10th ed. Dover Publications, Dec. 1972.
[11] R. L. Al-Habash, M. A. Andrews, and L. C. Phillips, "Mathematical
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.
[12] LS. Gradshteyn and I.M. Ryzhik, Table of Integrals, Series, and Prod-
ucts, 7th ed. New York: Academic, 2007.
[13] Wolfram. The Wolfram function site. Internet. [Online]. Available:
http://functions.wolfram.com, 2001.
[14] J. Hamkins and B. Moision, "Multipulse pulse-position modulation
on discrete memoryless channels," The Interplanetary Network Progress
Report, vol. 42, no. 161, pp. 1-13, May 2005.
[15] M. Feng et al., "Outage performance for parallel relay-assisted free-
space optical communications in strong turbulence with pointing errors,"
Proc. of International Conference on Wireless Commun. and Signal
Processing (WCSP), pp. 1-5, 201l.
[16] P. V. Trinh et al., "BER analysis of all-optical AF dual-hop FSO
systems over Gamma-Gamma channels," Proc. of IEEE 4th International
Conference on Photonics, pp. 175-177, Oct. 2013.

212