IET Optoelectronics

Research Article

ISSN 1751-8768
Performance improvement of FSO satellite Received on 24th September 2015
Revised on 1st December 2015
downlink using aperture averaging and Accepted on 21st December 2015
doi: 10.1049/iet-opt.2015.0102
receiver spatial diversity www.ietdl.org

Pooja Gopal ✉, Virander Kumar Jain, Subrat Kar

Electrical Engineering Department, Bharti School of Telecom Technology and Management, Indian Institute of Technology Delhi, Hauz
Khas, New Delhi 110016, India
✉ E-mail: poojagopal.iitd@gmail.com

Abstract: In this study, the bit error rate performance evaluation of a free space optical (FSO) satellite downlink has been
carried out. Subsequently, improvement in performance using aperture averaging and receiver spatial diversity
techniques have been explored. With a single large aperture receiver or an array of small aperture receivers, the
scintillation index in the downlink can be reduced to a very small value. Moreover, a single receiver gives better
performance than an array of receivers (employing equal gain combining) with the total area same as that of the single
receiver. The expression of channel capacity with outage for the slow fading FSO channel has been derived using the
log-normal and gamma–gamma channel models. Based on maximum acceptable outage probability, the minimum
required signal-to-noise ratio is determined, from which the capacity with outage is calculated. It has been observed
that the maximum achievable capacity per unit bandwidth increases and the corresponding outage probability reduces
with an increase in the aperture diameter for a single receiver or with the size of the array for a receiver with spatial
diversity.

1 Introduction outage probability for the FSO satellite links are not available in
the literature and hence is the subject of study in this paper.
Optical satellite links provide several advantages over the The organisation of this paper is as follows. In Section 2, the bit
conventional radio frequency (RF) satellite links like higher error rate (BER) performance of an FSO satellite downlink is
bandwidth, lower transmit power and high security [1]. Free space studied. The reduction in scintillation effect and the improvement
optical (FSO) inter-satellite links are well researched and working in BER performance by using aperture averaging and receiver
links have been established [2, 3]. However, FSO links between spatial diversity techniques are discussed in Section 3. Link
an Earth station and a satellite are under rigorous investigation [4]. capacity evaluation is discussed in Section 4. The last section
These links fail in the presence of clouds and therefore cannot be gives the conclusions of the study.
used without an RF backup link. However, even in clear sky
conditions, the problem arises because of the atmospheric
turbulence. The major effects of turbulence on an FSO satellite 2 Performance evaluation
downlink are scintillation and angle of arrival fluctuations [1]. The
latter effect results in image dancing in the focal plane of an For this study, we have considered an FSO link from GEO to Earth.
imaging system and is not of much concern in an intensity The GEO satellite is in Earth’s equatorial plane and the zenith angle,
modulated/direct detection (IM/DD) communication system. In this ζ of the link from New Delhi, India (which is at a latitude of 28.61°)
paper, the performance evaluation of IM/DD optical is calculated to be around 30° for satellite altitude, H = 38.5 × 106 m,
communication link has been carried out considering the effects of receiver at ground level and radius of the Earth, RE = 6371 × 103 m.
scintillation. Subsequently, techniques to mitigate scintillation Therefore, we have considered ζ = 30° in our analysis. For most of
effects on the performance of an FSO downlink from a satellite in the path, the signal travels through free space where there is no
geostationary orbit (GEO) to the Earth station are explored. atmospheric turbulence. By the time it reaches the receiver at the
The scintillation effect at the receiver can be reduced by using ground, the signal wave-front would be approximately plane.
multiple apertures, that is, spatial diversity. Aperture averaging is Hence the scintillation is modelled using the plane wave
the simplest form of receiver spatial diversity in which there is a approximation.
reduction in scintillation with an increase in aperture diameter [5,
6]. The Shannon/ergodic capacity for FSO links based on
log-normal and gamma–gamma channel models is available in 2.1 Channel models
literature [7–9]. Practical achievement of this capacity would
require extremely long codewords for FSO links working at Gbps For the description of atmospheric turbulence, several models
data rates since the channel is slow fading. The other definition of namely: double Weibull [17], double generalised gamma [18],
capacity is based on outage which is more suitable for slow fading Malaga [19], gamma–gamma [20], log-normal [20], etc. are
channels. available. The first three models are accurate and also provide
The aperture averaging and spatial diversity techniques applied to closed form expressions for the BER. However, the last two
terrestrial FSO links have been studied in the past [10–16]. The models are relatively easy to tackle mathematically and are as
effect of the above mentioned techniques on FSO satellite links accurate as the first three models in weak scintillations [18] and
using the gamma–gamma channel model which is valid in all the downlink scintillations mostly fall in the weak regime. Further,
turbulence regimes has not been explored. Moreover, the the results on the analysis using the gamma–gamma model for
derivation of expressions for capacity with outage and the study of satellite links are not available in literature. Though this model is
the effect of these techniques on the achievable capacity and valid for all turbulence regimes, at very weak turbulence regime it

IET Optoelectron., 2016, Vol. 10, Iss. 4, pp. 119–127

& The Institution of Engineering and Technology 2016 119

Authorized licensed use limited to: AMITY University. Downloaded on July 11,2020 at 20:08:20 UTC from IEEE Xplore. Restrictions apply.
 




does not give proper results. The log-normal model can be used in m = (RkIlA/ sN2 ), respectively, where R is the responsivity of the
such cases. Therefore, in this paper we have considered both photodetector in AW−1 (R = ηql/hPc where η is the quantum
gamma–gamma and log-normal models. efficiency, q the electronic charge, hP the Planck constant and c the
The atmospheric turbulence in the channel may vary from weak to speed of light), A the area of the receiver aperture of diameter D (A =
strong. In the weak turbulence regime, the channel is modelled using πD 2/4 assuming a circular aperture) and sN2 the total noise variance.
the log-normal distribution. The probability density function (pdf) of The incoming information signal along with the background noise
channel state s (= I/〈I〉 where I and 〈I〉 are the instantaneous and signal falls on the photodetector. Since the photodetector is a
average values of the received intensity, respectively) is given by square-law device, the noise components generated at the
[20] photodetector output are: shot noise due to signal, shot noise due
  to background radiation, signal-background beat noise and
2  background–background beat noise. In addition to these, there will
1 ln(s) + (s2I /2)
pI (s) = 






exp − , s.0 (1) be a thermal noise component at the receiver output. The noise
s 2psI2 2s2I power associated with the above components are ss2 , sb2 , ssb 2
, sbb
2

and s2T , respectively, and are given by

where s2I is the scintillation index. For moderate to strong turbulence
regime, the gamma–gamma distribution is used. The pdf of s in the
gamma–gamma model is given by [20] ss2 = 2qBRIA (7a)

2(ab)((a+b)/2) ((a+b)/2)−1 

pI (s) = (s) Ka−b 4abs , s.0 (2) sb2 = 2qB2 RNb (7b)
G(a)G(b)

where α and β are positive parameters related to the scintillation ssb

2
= 4BR2 Nb IA (7c)
index as [20]

1 1 1 sbb
2
= 2B2 R2 Nb2 (7d)
s2I = + + (3)
a b ab
and
This model can be used in the weak turbulence regime as well.
However, the approximation in numerical computation involving
the Meijer-G function [21] does not give proper results in a very 4kB TB
weak turbulence regime which corresponds to high values of α
sT2 = (7e)
RL
and β. In such cases, the log-normal model is preferred. The
expressions for α and β for the satellite downlink are given by [1]
where B is the receiver bandwidth in Hz, Nb the power spectral

 −1 density (PSD) of the background radiation in WHz−1, kB the
0.49sR2 Boltzmann constant, T the receiver temperature in K and RL the
a = exp −1 (4a)
(1 + 1.11s12/5
7/6 load resistance in Ω.
R )
The PSD of background radiation is obtained from the following
formula [22]
and

 −1
0.51sR2 l4
b = exp −1 (4b) Nb = m R(l) (8)
5/6 c
(1 + 0.69s12/5
R )

where s2R is the Rytov variance and is given by [1] where R(l) is the spectral radiance of the background source in
Wm−3, c is the speed of light in ms−1 and m is the number of
H spatial modes. A single spatial mode corresponds to the light
s2R = 2.25 k7/6 sec11/6 (z) Cn2 (h)(h − h0 )5/6 dh (5) collected from the diffraction limited field of view (FOV) which is
h0 the smallest possible FOV of the lens. The receiver FOV is
defined as the solid angle over which the detector is sensitive to
Here, k is the wave number in m−1 corresponding to wavelength l in incoming light. As shown in Fig. 1, the number of spatial modes,
m (k = 2π/l), h0 the height in m above ground level of the downlink
receiver and H the satellite altitude in m. The variation of the
refractive index structure constant, Cn2 with height, h for a vertical
or slant path is obtained from the Hufnagel Valley Boundary
model given by the following empirical formula [1]
 w 2 
−h
Cn2 (h) = 0.00594 (10−5 h)10 exp + 2.7
27 1000
 
−h −h
× 10−16 exp + Cn2 (0) exp (6)
1500 100

where w is the rms wind speed in ms−1 and Cn2 (0) the ground level
turbulence in m−2/3.

2.2 BER analysis





For a PIN photodiode detector, the instantaneous and average

electrical
signal-to-noise ratios (SNRs) are given by g = (RIA/ s2N ) and Fig. 1 Different fields of view for calculation of number of spatial modes

IET Optoelectron., 2016, Vol. 10, Iss. 4, pp. 119–127

120 & The Institution of Engineering and Technology 2016
Authorized licensed use limited to: AMITY University. Downloaded on July 11,2020 at 20:08:20 UTC from IEEE Xplore. Restrictions apply.
Table 1 Parameters used in the numerical computation (i) Log-normal model: Substituting pI(s) from (1), we get
Parameter Value
1   2  
1 1 ln(s) + (s2I /2) ms
l 1550 nm kBERl = 





exp − erfc √

ds
ζ 30° 2 2ps2I 0 s 2s2I 2
h0 0m
H 38.5 × 106 m
(14)
w 21 ms−1
η 0.8
B 109 Hz
θFOV −6
90 × 10 radians   





R(l) 2.1171× 105 Wm−3 With ln (s) + (s2I /2) / 2s2I = x and change of limits of the
T 300 K integration, we get
RL 50 Ω
1  




 
1   s2 m
kBERl = √

exp −x2 erfc exp 2s2I x − I √

dx
2 p −1 2 2
m is given by [22] (15)
⎧
⎪ VFOV
⎪
⎨V , Vs . VFOV The above equation is similar to
m= DL
(9)
⎪
⎪ Vs 1
⎩ , Vs , VFOV
VDL exp(−x2 )f (x) dx (16a)
−1
where Ωs, ΩFOV and ΩDL are the solid angles subtended by the
source of background radiation, receiver FOV and the diffraction where
limited FOV, respectively. These are given by  




 
s2 m
p D2 f (x) = erfc exp 2s2I x − I √

(16b)
Vs = 2s (10a) 2 2
4L
pu2FOV For easy numerical computation, we can apply the Gauss–Hermite
VFOV = (10b) approximation [24] given by
and 4
1 
  N
4l2 exp −x2 f (x) dx = wi f (xi ) (17)
VDL = (10c) −1
p D2 i=−N

where Ds is the diameter of the radiation source, θFOV the planar where xi s are the roots and wi s are the corresponding weights of the
angle of receiver FOV and D the receiver aperture diameter. When Hermite polynomial [21]. Applying the above approximation to (15),
the source of background radiation is the Earth’s atmosphere we get
extending over the entire hemisphere, clearly Ωs > ΩFOV (see  




 
Fig. 1), implying m = ΩFOV/ΩDL. The value of m and therefore Nb 1  N
s2 m
kBERl = √

wi erfc exp 2s2I xi − I √

(18)
depends on the aperture diameter. For D = 4 cm and other 2 p i=−N 2 2
parameters as given in Table 1, the values of m and Nb are 3.3275
and 1.3554 × 10−26 WHz−1, respectively. Since m is an integer, it
is taken to be 3.
The total noise variance, s2N is given by
(ii) Gamma–gamma model: Substituting pI(s) from (2) in (13), we
get
s2N = s2s + s2b + s2sb + s2bb + s2T (11)
1  






((ab))((a+b)/2) ms
Among all the noises, for the GEO to Earth link considered we kBERl = s((a+b)/2)−1 erfc √

Ka−b 4abs ds
observe that s2b , s2sb and s2bb are negligible. Moreover, except for G(a)G(b) 0 2
high 〈I〉 and large D, the shot noise is not significant and can be (19)
neglected making the receiver thermal-noise-limited.
For on–off keying modulation scheme, the BER conditioned on s
is given by [23]
√

  We can express Kn ( x) and erfc(x) in terms of Meijer-G function
1 g 1 ms as given in equations 03.04.26.0009.01 and 06.27.26.0006.01 in
pe,c = erfc √

= erfc √

(12)
2 2 2 2 [25] as (see (20))
The integral in the above equation is similar to the form given in
since g/μ = I/〈I〉 = s. In order to get the unconditional BER, we have equation 07.34.21.0011.01 in [25] and hence we get
to integrate pe,c over the pdf pI(s) as
 ⎛ ⎞
 1−a 2−a 1−b 2−b
1 a+b−3  1, , , ,
⎜ 8m 
2
2
kBERl = 3/2 G2,4 ⎝  2 2 2 2 ⎟⎠
pe,uc = kBERl = pe,c pI (s) ds (13) (p) G(a)G(b) 5,2 (ab)2  1
0 0,
2
(21)

1      
 2 2 1
(ab)((a+b)/2) ((a+b)/2)−1  a − b−b − a 2,0 m s 
kBERl = √

s 2,0
G0,2 abs G1,2  1 ds (20)
2 pG(a)G(b) 0  , 2  0,
2 2 2

IET Optoelectron., 2016, Vol. 10, Iss. 4, pp. 119–127

& The Institution of Engineering and Technology 2016 121

Authorized licensed use limited to: AMITY University. Downloaded on July 11,2020 at 20:08:20 UTC from IEEE Xplore. Restrictions apply.
Variations of 〈BER〉 with 〈I〉, computed from (18) for Cn2 (0) = 10−14 and The average intensity in the receiver plane at the optical axis is 〈I〉.
5 × 10−14 m−2/3 and (21) for Cn2 (0) = 10−13 , 5 × 10−13 and 10−12 m−2/3, It reduces as exp (−r 2/W 2) where r is the distance from the optical
are shown in Fig. 2. All other parameters used for the numerical axis andW is the
beam radius in the receiver plane given by










computation are given in Table 1. It is observed from this figure that
W = W0 Q20 + L20 where W0 is the beam radius in the transmitter
the degradation in the link performance due to atmospheric turbulence
plane, Θ0 and Λ0 are the beam parameters at the transmitter given
is significant for Cn2 (0) . 10−13 m−2/3 .
by Θ0 = 1 for a collimated beam and L0 = 2L/(kW02 ). For W0 =
2 cm and remaining parameters as in Table 1, we get W = 1096.78 m
and therefore on aperture diameter (≃cm) the intensity remains
3 Performance improvement techniques almost constant. We see that even for r = 10 m, the intensity is still
99.9915% of that at r = 0 m and hence it is safe to consider the
For a point receiver (i.e. receiver with diameter less than the spatial
intensity to be constant over the receiver aperture. With increase in
correlation width of the atmosphere, ρc), we see from Fig. 2 that the D, the BER performance improves due to two reasons: (a)
performance degrades significantly for high Cn2 (0) values, that is,
reduction in scintillation effect and (b) increase in the average
Cn2 (0) . 10−13 m−2/3 . For receiver aperture sizes greater than ρc,
received power, 〈P〉 = 〈I〉A.
aperture averaging takes place. Moreover, multiple receivers Variations of 〈BER〉 vs. 〈I〉 for various D are shown in Fig. 4 for
separated by distances greater than ρc act like independent
receivers and hence can significantly reduce the scintillation effect. Cn2 (0) of (a) 10−13, (b) 5 × 10−13 and (c) 10−12 m−2/3. The values of
In this section, we explore the effect of aperture averaging and 〈I〉 (required to achieve a BER of 10−9) and s2I at different Cn2 (0) for
receiver spatial diversity on the downlink performance. various D are given in Table 2. Lower the required 〈I〉 to achieve a
BER of 10−9, better is the performance. It is observed from Fig. 4
and Table 2 that the performance improvement due to aperture
3.1 Aperture averaging averaging is more at higher Cn2 (0). Further, improvement is very
significant as the diameter increases from 1 to 4 cm. Afterwards,
The empirical formula for ρc at the downlink receiver on the ground with the increase in diameter, the rate of improvement gradually
is given by [1] reduces. This is true for all Cn2 (0) values.
"


















45 × 103 sec (z)
rc = (22) 3.2 Spatial diversity
k
Sometimes it is not feasible to use a single large receiver aperture due
For ζ = 30° and l = 1550 × 10−9 m, we get the value of ρc to be to practical limitations. In such cases, improved performance can be
1.13 cm. Therefore, a receiver with diameter, D = 1 cm will behave obtained by using an array of smaller sized apertures separated by
like a point receiver since D ≤ ρc. When D > ρc, the aperture distances greater than ρc. At the downlink receiver, ρc is in the
averaged scintillation index is derived using the ABCD parameters order of a few centimetres as mentioned in Section 3.1. This
and is given by [1] implies that receiver spatial diversity is a feasible option for the
satellite downlink. A schematic of receiver spatial diversity in FSO
H
satellite downlink is shown in Fig. 5.
s2I = 8.7 k7/6 (H − h0 )5/6 sec11/6 (z) × Re Cn2 (h) Generally, three types of linear combining methods are used to
h0

combine signals from multiple receiver apertures: selection
5/6 2 5/6 
kD2 h − h0 kD combining (SC), maximum ratio combining (MRC) and equal gain
× +i − dh (23a) combining (EGC) [5]. Among these, SC is the simplest but does
16L H − h0 16L
not cause significant improvement in the average SNR. MRC is
the optimal method but requires complex circuitry to get the
where correct weighting factors. In EGC, equal gains are applied to all
the signals. In spite of its simplicity, its performance though
L = (H − h0 ) sec (z) (23b) slightly poorer, is comparable with that of MRC. Therefore in the
following analysis, we have considered only the EGC method.
is the path length. Fig. 3 shows the reduction in s2I with increase in D For the composite signal from N receivers, both the signal power
for different values of Cn2 (0). We observe that for larger aperture and the noise variance are increased by a factor of N. The average
diameters, s2I reduces to a very small value for all ground SNR considering N receivers is given by
turbulence levels. Moreover, aperture averaging is most effective
at high Cn2 (0).
N RkIlA √



mN = 





= N m (24)
N s2N

Fig. 2 kBERl against. kIl for different Cn2 ( 0) for a point receiver Fig. 3 s2I against. D for downlink for different Cn2 ( 0)

IET Optoelectron., 2016, Vol. 10, Iss. 4, pp. 119–127

122 & The Institution of Engineering and Technology 2016
Authorized licensed use limited to: AMITY University. Downloaded on July 11,2020 at 20:08:20 UTC from IEEE Xplore. Restrictions apply.
Fig. 4 kBERl against kIl for various D at different Cn2 ( 0)
a C2n(0) = 10−13m−2/3
b C2n(0) = 5 × 10−13m−2/3
c C2n(0) = 10−12m−2/3

Moreover, the effective scintillation index is given by [1] 4 Link capacity evaluation

We have considered that channel pdf is known at both the transmitter

s2I
s2I,N = (25) and the receiver ends. If there is no channel state information (CSI)
N available at either the transmitter or the receiver end, then we have to
find an input distribution that maximises capacity. This procedure
The variation of 〈BER〉 with 〈I〉 at Cn2 (0) = 10−12 m−2/3 for an array can be very complex for the FSO channel distributions. When the
of receiver apertures of (a) D = 1 cm (point receiver), (b) D = 4 cm CSI is available at the receiver end, there are two definitions of
and (c) D = 8 cm is shown in Fig. 6. The arrangement of the link capacity: Shannon capacity and capacity with outage. The
receivers in the array is not considered in this paper. However, Shannon or ergodic capacity is given as [26]
even in this case the received intensity can be considered to be
constant over the array.
When we consider an array of finite sized apertures, the effects of 1
both aperture averaging and spatial diversity occur. Presuming all CS = Blog2 (1 + g)p(g) dg (26)
receivers in the array to be of same size, first the aperture averaged 0
scintillation for a single receiver is calculated and then it is
reduced by a factor of N to account for the spatial diversity effect.
It is seen from this figure that a very large array of point apertures where p(g) is the pdf of the channel with the instantaneous SNR, g as
is required to match the performance of single large diameter the random variable. Since the transmitter is not aware of the CSI, it
aperture. Further, an array of finite sized apertures instead of point maintains a constant rate. In order to achieve the above capacity, we
apertures considerably improves the performance. Table 3 gives need to send a sufficiently long code word which gets affected by all
the values of 〈I〉 (required to achieve a BER of 10−9) and s2I,N for the fading states. The coherence time of the atmosphere is in the
different N considering an array of 1 cm (point), 4 and 8 cm order of milliseconds [20]. For a signal with data rate in the order
diameter apertures. An interesting and practically useful of Gbps, the transmitter and the receiver would become extremely
observation can be made from Tables 2 and 3. From these tables, complex. Hence achieving this capacity is not feasible for the FSO
we observe that a single aperture of 12 cm diameter achieves a satellite downlink with a slow fading channel.
similar performance as an array of 5 apertures of 8 cm diameter The second definition is the capacity with outage, Co which is
(corresponding entries in both tables are highlighted in grey). The suitable for slow fading channels. The fixed channel capacity will
single aperture has an area of 113.10 cm2 as compared with a total be C = B log2 (1 + gmin ). This implies that all data with g ≥ gmin
area of 251.33 cm2 for the array. It may be noted that the total is decoded with negligible probability of error. However, data at
receiver noise (thermal noise) in the array is N = 5 times greater the receiver with g < gmin results in an outage whose probability
than that of the single aperture. is given by pout = p(g < gmin). The average capacity with outage is

Table 2 Required 〈I〉 to achieve a BER of 10−9 and reduction in s2I with increase in D at different Cn2 (0)
Cn2 (0)(m−2/3 )  10−13 5 × 10−13 10−12

D, cm ↓ 〈I〉, mW/m2 s2I 〈I〉, mW/m2 s2I 〈I〉, mW/m2 s2I

1 (point Rx.) 40.5713 0.1338 108.3243 0.3584 350.1712 0.5727

4 2.0844 0.0634 2.2311 0.0937 2.5141 0.1316
8 0.4994 0.0422 0.5023 0.0489 0.5141 0.0574
12 0.2152 0.0296 0.2205 0.0323 0.2192 0.0356
16 0.1181 0.0210 0.1213 0.0223 0.1238 0.0240

IET Optoelectron., 2016, Vol. 10, Iss. 4, pp. 119–127

& The Institution of Engineering and Technology 2016 123

Authorized licensed use limited to: AMITY University. Downloaded on July 11,2020 at 20:08:20 UTC from IEEE Xplore. Restrictions apply.
 Table 3 Required 〈I〉 (to achieve a BER of 10−9) and reduction in s2I,N at
Cn2 (0) = 10−12 m−2/3 with increase in N for different D

D (cm)  1 (point Rx.) 4 8

N↓ 〈I〉, mW/ s2I,N 〈I〉, mW/ s2I,N 〈I〉, mW/ s2I,N

m2 m2 m2

1 350.1721 0.5727 2.5142 0.1316 0.5144 0.0574

5 17.0315 0.1145 0.9124 0.0263 0.2196 0.0115
10 10.3914 0.0573 0.6222 0.0132 0.1541 0.0058
15 8.2314 0.0382 0.5041 0.0088 0.1252 0.0038
20 7.0459 0.0286 0.4361 0.0066 0.1092 0.0029

the instantaneous SNR, g, we use the transformation

ds
p(g) = pI (s) (28a)
dg


g 1
= pI × (28b)
m m

since relation between g and s is g = μs.

Fig. 5 Schematic of receiver spatial diversity in FSO satellite downlink

4.1 Log-normal model
therefore given by Using (1) and (28), we can write p(g) as

    2 
Co = (1 − pout )Blog2 (1 + gmin ) (27) 1 ln g/m + (s2I /2)
p(g) = 





exp − , g.0 (29)
2ps2I g 2s2I

In this equation, pout is a design parameter based on acceptable

outage. For a given pout, we can determine gmin. The average The outage probability is given by
capacity with outage, Co is calculated using pout and gmin from (27).
In the previous section, we have seen that by using aperture
averaging and receiver spatial diversity techniques, s2I can be pout = p(g , gmin ) (30a)
reduced to a very small value for all turbulence conditions. This
corresponds to very high values of α and β for which the Meijer-G gmin     2 
1 1 ln g/m + (s2I /2)
approximation of the gamma–gamma distribution does not give = 





exp − dg (30b)
proper results. Hence both the log-normal (which is accurate for 2ps2I 0 g 2s2I
low s2I ) and the gamma–gamma models are used here for the
derivation of capacity with outage.     





If we  take ln g/m + (s2I /2) / 2s2I = x, it gives
The pdf of both log-normal and gamma–gamma models given by 





(1) and (2) are in terms of the channel state, s. To get pdf in terms of g = mexp 2s2I x − (s2I /2) . In that case, the outage probability

Fig. 6 kBERl against. kIl for different N and

a D = 1 cm (point receiver)
b D = 4 cm
c D = 8 cm at Cn2 (0) = 10−12 m−2/3

IET Optoelectron., 2016, Vol. 10, Iss. 4, pp. 119–127

124 & The Institution of Engineering and Technology 2016
Authorized licensed use limited to: AMITY University. Downloaded on July 11,2020 at 20:08:20 UTC from IEEE Xplore. Restrictions apply.
Fig. 7 Co/B and pout against. gmin for different D for aperture averaging at Cn2 ( 0) = 10−12 m−2/3
a Co/B against gmin
b pout against gmin

in terms of x will be 4.2 Gamma–gamma model

xmin 




From (2) and (28)
1
pout = 





exp (−x2 ) 2s2I dx (31a)
2ps2I −1 "











2(ab)((a+b)/2) g
1 xmin p(g) = g((a+b)/2))−1 Ka−b 4ab (34)
1 1 2 G(a)G(b)m((a+b)/2) m
= √

exp(−x2 ) dx + √

exp (−x2 ) dx (31b)
p 0 2 p 0
With
    





where xmin = ln gmin /m + (s2I /2) / 2s2I . Using Gaussian
integral and error function as given by 2(ab)((a+b)/2)
= A1 ,
#

G(a)G(b)m((a+b)/2)
1
p
Gaussian integral: exp(−ax2 ) dx = (32a)
−1 a the outage probability can be written as

x 1
2
Error function : erf (x) = √

2
exp(−x ) dx (32b) pout = 1 − p(g) dg = 1 − Int (35)
p 0 gmin

in (31), we get where

    1 "











1 ln gmin /m + (s2I /2) ((a+b)/2)−1 g
pout = 1 + erf 




(33) Int = A1 g Ka−b 4ab dg (36)
2 2s2I gmin m

Let g1 = g/gmin. This implies g = gming1 and dg = gmin dg1.

Fig. 8 Co/B and pout against gmin for different N for receiver spatial diversity with D = 4 cm at Cn2 ( 0) = 10−12 m−2/3
a Co/B against gmin
b pout against gmin

IET Optoelectron., 2016, Vol. 10, Iss. 4, pp. 119–127

& The Institution of Engineering and Technology 2016 125

Authorized licensed use limited to: AMITY University. Downloaded on July 11,2020 at 20:08:20 UTC from IEEE Xplore. Restrictions apply.
Table 4 Values of gmin, Co/B and pout corresponding to the maximum achievable capacity at Cn2 (0) = 10−12 m−2/3 for aperture averaging and receiver
spatial diversity technique with aperture diameter, D = 4 cm

Technique  Aperture averaging Spatial diversity

Parameter  Aperture diameter, D, cm No. of receivers, N

1 4 8 12 16 1 5 10 15 20

Co /B 1.85 2.51 2.77 2.89 2.98 2.51 2.96 3.08 3.14 3.18
gmin, dB 6.25 7.55 8.15 8.45 8.68 7.52 8.60 8.95 9.11 9.21
pout 22.7% 8.3% 4.7% 3.4% 2.9% 8.2% 2.4% 2.1% 1.8% 1.4%

Table 5 Values of gmin and Co/B corresponding to pout = 1% and pout = 10% at Cn2 (0) = 10−12 m−2/3 for aperture averaging and receiver spatial diversity
technique with aperture diameter, D = 4 cm

Technique  Aperture averaging Spatial diversity

Parameter  Aperture diameter, D, cm No. of receivers, N

1 4 8 12 16 1 5 10 15 20

gmin (dB) for pout = 1% 1.29 6.12 7.47 8.04 8.42 6.22 8.36 8.86 9.07 9.19
corresponding Co/B 1.21 2.34 2.69 2.84 2.96 2.34 2.93 3.07 3.13 3.17
gmin (dB) for pout = 10% 4.57 7.71 8.56 8.89 9.09 7.72 9.04 9.33 9.46 9.55
corresponding Co/B 1.75 2.51 2.70 2.81 2.86 2.50 2.86 2.93 2.99 2.99

Changing the variable in Int from g to g1 we get acceptable pout = 1%, then the values of gmin corresponding to pout
= 1% are chosen instead of those corresponding to maximum
1 achievable Co/B. However if the acceptable pout = 10%, then gmin
Int = A1 (gmin )((a+b)/2) g1((a+b)/2)−1 H(g1 corresponding to maximum achievable Co/B can be chosen for all
0
"

















cases except the case of D = 1 cm which has pout = 22.7% as seen
g1 gmin from Table 4. In this case gmin corresponding to pout = 10% is
− 1)Ka−b 4ab dg1 (37) chosen.
m

where H(x) is the Heaviside function.

√
The Meijer-G function
equivalents of x vH(x − 1) and Kn ( x) are 5 Conclusions
   For a point receiver in optical satellite downlink, the link
v+1 1 2,0 x n − n performance degrades significantly for Cn2 (0) . 10−13 m−2/3 . The
0,1
G1,1 x| G0,2 |
v 2 4 2 −2 improvement in performance using aperture averaging and receiver
spatial diversity techniques have been explored in this work. The
respectively [21]. Substituting this in (35) and using the Meijer-G scintillation index can be reduced to a very small value for all
integration formula given by equation 07.34.21.0011.01 in [25], Cn2 (0) for a single large aperture diameter due to aperture
we get averaging or with an array of small fixed diameter receivers due to
spatial diversity.
In receiver spatial diversity, a very large array of point receivers
pout = 1
⎛  ⎞ are required to match the performance of a single large diameter
 (a+b/2) a+b receiver. However considering an array of finite sized receivers
1−
abgmin /m ⎜
3,0 ⎜abgmin 2 ⎟ instead of point receivers considerably improves the performance.
⎟
−
G(a)G(b)
G1,3 ⎝ m | a − b b − a a + b ⎠ It has been observed that a single receiver gives better
− performance than an array of receivers whose total area is same as
2 2 2
that of the single receiver. This is because the total receiver noise
(38) in the array is N times greater than that in the single receiver, but
the signal power in both the cases remains almost same.
In Figs. 7a and b, variations of Co/B vs. gmin and pout vs. gmin, The capacity with outage per Hz, that is, Co/B increases with the
respectively, are given for different D, for the case of aperture increase in the aperture diameter for a single receiver or with size of
averaging. The maximum achievable Co/B is higher for larger D as the array in the receiver with spatial diversity. Choice of either of the
seen from Fig. 7a. From Fig. 7b, it is observed that the rate of schemes for the realisation of a given Co/B is influenced by
increase in pout with an increase in gmin is higher for larger D. complexity and cost considerations.
Similar behaviour is observed in Figs. 8a and b for the receiver
spatial diversity where maximum achievable Co/B is higher for
larger N. For spatial diversity, the diameter of each receiver is
taken to be D = 4 cm. In both cases, the average received SNR, μ 6 References
is taken to be 10 dB and Cn2 (0) = 10−12 m−2/3 .
1 Andrews, L.C., Phillips, R.L.: ‘Laser beam propagation through random media’
In Table 4, the values of gmin and pout corresponding to maximum (SPIE Press, Washington, 2005)
achievable Co/B are given. When pout is taken to be the design 2 Gregory, M., Heine, F., Kampfner, H., et al.: ‘Commercial optical inter-satellite
parameter, in Table 5 the values of gmin and Co/B corresponding communication at high data rates’, SPIE Opt. Eng., 2012, 51, (3), pp. 1–7
to two values of pout = 1% and pout = 10% are given for both the 3 Panahi, A.S., Kazemi, A.A.: ‘Inter-satellite communications using laser based
optical links’, Proc. SPIE, 2007, 6758, pp. 1–12
techniques. From Table 4, we observe that for all cases of aperture 4 Toyoshima, M., Takayama, Y., Takahushi, T., et al.: ‘Ground-to-satellite laser
averaging and spatial diversity, the value of pout corresponding to communication experiments’, IEEE Aerosp. Electron. Syst. Mag., 2008, 23, (8),
maximum achievable Co/B is >1%. Hence, if the maximum pp. 10–18

IET Optoelectron., 2016, Vol. 10, Iss. 4, pp. 119–127

126 & The Institution of Engineering and Technology 2016
Authorized licensed use limited to: AMITY University. Downloaded on July 11,2020 at 20:08:20 UTC from IEEE Xplore. Restrictions apply.
 5 Andrews, L.C., Phillips, R.L., Hopen, C.Y.: ‘Laser beam scintillation with 16 Mehta, N., Joseph, T., Gopal, P., et al.: ‘Performance improvement techniques for
applications’ (SPIE Press, Washington, 2001) optical wireless link in the presence of atmospheric turbulence’, Proc. SPIE, 2013,
6 Viswanath, A., Gopal, P., Jain, V.K., et al.: ‘Performance enhancement by aperture 87684S, pp. 1–7
averaging in terrestrial and satellite free space optical links’ (Accepted for 17 Chatzidiamantis, N.D., Sandalidis, H.G., Karagiannidis, G.K., et al.: ‘New results
publication in IET Optoelectron., 2015) doi: 10.1049/ietopt.2015.0042 on turbulence modeling for free-space optical systems’. IEEE Int. Conf.
7 Nistazakis, H.E., Karagianni, E.A., Tsigopoulos, A.D., et al.: ‘Average capacity of Telecommunications (ICT), Doha, Qatar, April 2010, pp. 487–492
optical wireless communication systems over atmospheric turbulence channels’, 18 Kashani, M.A., Uysal, M., Kavehrad, M.: ‘A novel statistical model for
J. Lightwave Technol., 2009, 27, (8), pp. 974–979 turbulence-induced fading in free-space optical systems’, IEEE/OSA
8 Nistazakis, H.E., Tombras, G.S., Tsigopoulos, A.D., et al.: ‘Average capacity of J. Lightwave Technol., 2015, 33, (11), pp. 2303–2312
wireless optical communication systems over gamma-gamma atmospheric 19 Jurado-Navas, A., Garrido-Balsellss, J.M., Paris, J.F., et al.: ‘A unifying statistical
turbulence channels’. Proc. IEEE MTT-S, Georgia, USA, June 2008, pp. 1561–1564 model for atmospheric optical scintillation’, in Jan, A. (Ed.): ‘Numerical
9 Gappmair, W.: ‘Further results on the capacity of free space optical channels in simulations of physical and engineering processes’ (InTech, Denmark, 2011),
turbulent atmosphere’, IET Commun., 2011, 5, (9), pp. 1262–1267
pp. 181–206
10 Navidpour, S.M., Uysal, M., Kavehrad, M.: ‘BER performance of free space
20 Majumdar, A.K., Ricklin, J.C.: ‘Free space communications: principles and
optical transmission with spatial diversity’, IEEE Trans. Wireless Commun.,
advances’ (Springer, New York, 2008)
2007, 6, (8), pp. 2813–2819
11 Yuksel, H., Milner, S., Davis, C.: ‘Aperture averaging for optimizing receiver 21 Prudnikov, A.P., Brychkov, Y.A., Marichev, O.I.: ‘Integrals and series volume 3:
design and system performance on free space optical communication links’, more special functions’ (Gordon and Breach Science Publishers, New York, 1986)
J. Opt. Netw., 2005, 4, (8), pp. 462–475 22 Fidler, F.: ‘Optical communications from high-altitude platforms’. PhD thesis,
12 Khalighi, M.A., Schwartz, N., Aitamer, N., et al.: ‘Fading reduction by aperture Institut fur Nachrichtentechnik und Hochfrequenztechnik, Vienna University of
averaging and spatial diversity in optical wireless systems’, J. Opt. Commun. Technology, 2007
Netw., 2009, 1, (6), pp. 580–593 23 Gopal, P., Jain, V.K., Kar, S.: ‘Modulation techniques used in earth to satellite and
13 Vetelino, F.S., Young, C., Andrews, L., et al.: ‘Aperture averaging effects on the inter-satellite free space optical links’, Proc. SPIE, 2014, 9248, pp. 1–10
probability density irradiance fluctuations in moderate-to-strong turbulence’, 24 Kiasaleh, K.: ‘Performance of APD-based, PPM free-space optical communication
Appl. Op., 2007, 46, (11), pp. 2099–2108 systems in atmospheric turbulence’, IEEE Trans. Commun., 2005, 53, (9),
14 Wang, Z., Zhong, W., Fu, S., et al.: ‘Performance comparison of different pp. 1455–1461
modulation formats over free space optical (FSO) turbulence links with space 25 ‘The Wolfram Function Site’, available at http://functions.wolfram.com, accessed
diversity reception technique’, IEEE Photonics J., 2009, 1, (6), pp. 277–285 September 2015
15 Kaur, P., Jain, V.K., Kar, S.: ‘Effect of atmospheric conditions and aperture 26 Goldsmith, A.: ‘Wireless communications’ (Cambridge University Press, New
averaging on capacity of free space optical links’, Opt. Quantum Electron., Delhi, 2010)
2013, 46, (9), pp. 1139–1148

IET Optoelectron., 2016, Vol. 10, Iss. 4, pp. 119–127

& The Institution of Engineering and Technology 2016 127

Authorized licensed use limited to: AMITY University. Downloaded on July 11,2020 at 20:08:20 UTC from IEEE Xplore. Restrictions apply.