GUO et al.: ON THE PERFORMANCE OF LMS COMMUNICATION 1493
where P
s
denotes the transmit power at S, f
u,k
∈ C
N
1
×1
the
uplink channel vector between the multi-antenna source and
the k-th satellite beam; P
j
the transmit power of the j-th inter-
ference; x
j
(t) the j-th interference with unit average power
E
|x
j
(t)|
2
=1, f
j,k
the interference link channel coefficient
from the j-th interferer to the k-th beam; η
1
(t) the distortion
noise term due to the satellite receive HIs, and n
1
(t) ∼
CN
0,δ
2
1
is the additive white Gaussian noise (AWGN)
at R, with variance δ
2
1
given by δ
2
1
= KTB; K, T , B
being the Boltzmann’s constant, receive noise temperature,
and noise bandwidth, respectively. According to [33], η
1
(t)
is often modeled as η
1
(t) ∼CN(0, Υ
1
), with Υ
1
given by
Υ
1
= k
r
P
s
f
H
u,k
w
1
2
+
M
j=1
k
I
P
j
|f
j,k
|
2
, (5)
where the coefficients k
r
and k
I
are introduced to characterize
the aggregate level [51] for the received impairments at R from
the source signal power and interference power, respectively.
2
In the second time slot, the regenerative satellite first
exploits the DF relaying protocol to decode the signal s (t)
from y
r,k
(t) correctly, and then forwards it to the D through
the l-th on-board beam and the corresponding downlink chan-
nel. Similarly, by considering the effect of transmit impair-
ments, the output signal at D can be written as
y
d,l
(t)=
P
r
w
H
2
f
d,l
(s (t)+η
2
(t)) + w
H
2
n
2
(t), (6)
where P
r
is the transmit power of the l-th on-board beam,
w
2
∈ C
N
2
×1
obeying w
2
2
=1is the destination BF
weight vector used to enhance the received signal power,
f
d,l
∈ C
N
2
×1
is the downlink channel vector between the
l-th satellite beam and D, η
2
(t) is the additional distortion
noise term due to satellite transmit HIs which is modeled as
η
2
∼CN(0, Υ
2
) with Υ
2
= k
t
, where the coefficient k
t
characterizes the aggregate level of HIs, and finally n
2
(t) ∼
CN
0,δ
2
2
I
stands for the AWGN vector with δ
2
2
= KTB.
C. Problem Formulation
Here we employ the above described system and signal
models to derive the optimal transmit and receive BF
such that the output signal-to-interference-plus-noise-and-
distortion-ratio (SINDR) is maximized. By using (1) and (4),
the SINDR at the k-th satellite beam can be expressed as
γ
1,k
=
P
s
f
H
u.k
w
1
2
k
r
P
s
f
H
u,k
w
1
2
+(1+k
I
)
M
j=1
P
j
|f
j,k
|
2
+ δ
2
1
=
P
s
C
k
h
H
u,k
w
1
2
k
r
P
s
C
k
h
H
u,k
w
1
2
+(1+k
I
)
M
j=1
P
j
|f
j,k
|
2
+ δ
2
1
. (7)
Since γ
1,k
is monotonously increased with C
k
h
H
1
w
1
2
,
the max output SINDR in (7) is obtained when both
2
In this paper, we assume that each interference ES is equipped with a
parabolic antenna. However, our analysis can be directly extended to the case
of multi-antenna ES.
C
k
and
h
H
u,k
w
1
2
are max. According to (3), it can be
found that the satellite beam gain is determined by the source
position. By considering that the locations of all earth stations
are available at the gateway [49], the on-board beam with
θ
k
→ 0 is selected to receive the signal from the source
during the first time slot, resulting in G
k
≈ G
max
. Meanwhile,
in order to obtain the optimal transmit BF weight vector w
1
,
we formulate a constrained optimization problem as
arg max
w
1
h
H
u,k
w
1
2
, (8a)
s.t. w
1
2
=1. (8b)
From [52], we know that when w
1
=
h
u,k
h
u,k
,
h
H
u,k
w
1
2
gets the largest value. By substituting w
1
=
h
u,k
h
u,k
into (7)
and after some necessary computations, the max outage
SINDR at R can be expressed as
γ
max
1,k
=
γ
s
γ
s
k
r
+ γ
I
(1 + k
I
)+1
, (9)
where γ
s
= P
s
C
max
h
u,k
2
/δ
2
1
Δ
=
γ
s
h
1
2
and γ
I
=
M
j=1
P
j
|f
j.k
|
2
/δ
2
1
Δ
=
γ
I
M
j=1
|f
j,k
|
2
.
In a similar manner, when the l-th beam with θ
l
→ 0 is used
to transmit the signal to D, and when w
2
= h
d.l
/h
d.l
2
is
chosen as the receive BF weight vector at the D, the max
output signal-to-noise-and-distortion-rate (SNDR) at D can be
written as
γ
max
2,l
=
γ
r
γ
r
k
t
+1
, (10)
where γ
r
= P
r
C
max
h
d,l
2
/δ
2
2
Δ
=
γ
r
h
2
2
. In the sequel,
we will use γ
1
to replace γ
max
1,k
and γ
2
to replace γ
max
2,l
for brevity. Thus, the equivalent end-to-end SINDR of the
regenerative satellite-based relay system with DF protocol is
given by
γ
e
=min
γ
max
1,k
,γ
max
2,l
Δ
=
min (γ
1
,γ
2
). (11)
III. S
TATISTICAL PROPERTIES OF SATELLITE CHANNELS
According to [3], the probability density function (PDF) of
γ
i
= γ
i
|h
i
|
2
with |h
i
|
2
Δ
=
[h
k
]
i
being the i-th element of SR
channel vector h
k
is given by
f
γ
i
(x)=
α
i
γ
i
e
−
β
i
γ
i
x
1
F
1
m
i
;1;
δ
i
γ
i
x
,x>0, (12)
where
1
F
1
(a
1
; b
1
; z) denotes the confluent hyper-
geometric function defined in [53, eq. (9.210)].
α
i
=
2b
i
m
i
2b
i
m
i
+Ω
i
m
i
/2b
i
, β
i
=
1
2b
i
, δ
i
=
Ω
i
2b
i
(2b
i
m
i
+Ω
i
)
with Ω
i
, 2b
i
and m
i
≥ 0 being the average power of the LOS
component, the average power of the multi-path component,
and the fading severity parameter ranging from 0 to ∞,
respectively [54]. In what follows, by considering two cases
of m
i
, namely, integer and rational, we will discuss the PDF
expressions of γ =
N
i=1
γ
i
in detail.