Professional Documents
Culture Documents
Taneli Riihonen, Stefan Werner, Risto Wichman, and Eduardo Zacarias B
Taneli Riihonen, Stefan Werner, Risto Wichman, and Eduardo Zacarias B
SR
LI
RD
Fig. 1. Two-hop relay link with potential loop interference.
The normalized
1
transmit powers in the source and in the
relay are denoted by p
S
and p
R
, respectively. The normal-
ized
1
channel gain factors
SR
,
LI
, and
RD
represent the
respective frequency-at sourcerelay, residual loop interfer-
ence, and relaydestination channels. The signal model is
omitted here for brevity as the end-to-end capacities can be
determined from straightforward modications to the capac-
ity expressions available in literature for the HD mode.
The signal-to-interference and noise ratio of the rst hop
is
R
=
p
S
SR
p
R
LI
+1
and the signal-to-noise ratio of the second
hop is
D
= p
R
RD
. Thus, by exploiting [1, Eqs. 12, 13, and
15] and modifying the signal models thereof, the end-to-end
capacities in the full-duplex mode are expressed as
C
AF
FD
= log
2
_
1 +
p
S
SR
p
R
LI
+1
p
R
RD
p
S
SR
p
R
LI
+1
+ p
R
RD
+ 1
_
, (1)
C
DF
FD
= log
2
_
1 + min
_
p
S
SR
p
R
LI
+ 1
, p
R
RD
__
(2)
for amplify-and-forward (AF) and decode-and-forward (DF)
protocols, respectively.
Relaying in the HD mode is also described by the above
model, except that the loop interference is eliminated and
transmission consumes twice the channel resources of the full-
duplex mode. Thus, the end-to-end capacities in the half-
duplex mode can be determined from (1), (2) by setting
LI
=
0 and dividing the expressions by two:
C
AF
HD
=
1
2
log
2
_
1 +
p
S
SR
p
R
RD
p
S
SR
+ p
R
RD
+ 1
_
, (3)
C
DF
HD
=
1
2
log
2
(1 + min {p
S
SR
, p
R
RD
}) . (4)
1
Normalization means here that the channel gains are dened relatively
to receiver noise power, and that all signal amplication and attenuation is
embedded in the channel gains such that p
S
+ p
R
2.
3. FEASIBILITY OF THE FULL-DUPLEX MODE
In this section, we evaluate the range of loop interference
levels for which the full-duplex mode outperforms the half-
duplex mode in terms of capacity.
Trivially, C
FD
= 2C
HD
with protocol {AF, DF} if
LI
= 0. Furthermore, C
FD
/C
HD
is continuous and monoton-
ically decreasing in terms of
LI
, and lim
LI
FD
/C
HD
=
0. Thus, there exists a break-even loop interference level
LI
=
LI
for which C
FD
= C
HD
. For all loop interference
levels below the break-even level
LI
, the FD mode results in
better capacity than the HD mode.
Analytically, we dene the break-even loop interference
level
LI
with protocol {AF, DF} as the loop interference
channel gain that satises the statement
C
FD
C
HD
if and only if
LI
LI
. (5)
In the following, we propose three power allocation (PA)
methods and derive closed-form solutions for the correspond-
ing power optimization problems. Previously, similar power
allocation methods have been studied only for the HD mode,
for example, in [8,9]. With each PA method and both relaying
protocols, we determine closed-form expressions for
LI
.
3.1. Uniform power allocation
Let us rst consider transmission without transmit power opti-
mization. The system uses uniform power allocation given by
p
S
= p
R
= 1 with both modes and protocols in this section.
The advantage of this approach is that channel state informa-
tion is not needed.
The break-even loop interference level
LI
for protocol
{AF, DF} is determined by solving
LI
in terms of
SR
and
RD
from the inequality C
FD
C
HD
with (1)(4). Thus,
(5) is true when
AF
LI
=
_
SR
+ 1
RD
+ 1
(
SR
+
RD
+ 1) (6)
with the amplify-and-forward protocol and
DF
LI
=
SR
_
_
min{
SR
,
RD
} + 1 + 1
_
min{
SR
,
RD
}
1 (7)
with the decode-and-forward protocol. We note that
LI
1
for {AF, DF}. Therefore, C
FD
C
HD
for all
SR
and
RD
if
LI
1.
Furthermore, we notice with the decode-and-forward pro-
tocol that C
DF
FD
= 2C
DF
HD
if and only if
LI
SR
RD
1 or
LI
= 0. This means that performance is the same as without
loop interference when the loop interference power is reason-
ably low or when
SR
RD
. Consequently, if
SR
<
RD
then loop interference degrades performance with all
LI
> 0.
3.2. Power allocation with individual constraints
Transmit power optimization is motivated by the fact that the
end-to-end performance is basically limited by the weakest of
the two hops as seen from (1)(4). Thus, if the limiting factor
in the FD mode is the rst hop due to excessive loop interfer-
ence power, the end-to-end performance can be improved by
decreasing the relay transmit power. This reasoning leads to
optimization of p
S
and p
R
under individual constraints, which
is formulated as
(p
S
, p
R
) = arg max
(p
S
,p
R
)
C
subject to 0 p
S
1 and 0 p
R
1
(8)
for protocol {AF, DF} and mode {FD, HD}. In the
rest of the paper, we assume that necessary ideal channel state
information is available for determining p
S
and p
R
.
In the HD mode the solution to (8) with both protocols is
trivially given by p
S
= p
R
= 1, because C
AF
HD
and C
DF
HD
are
monotonically increasing in terms of both p
S
and p
R
. Simi-
larly p
S
= 1 in the FD mode since C
AF
FD
and C
DF
FD
are mono-
tonically increasing in terms of p
S
with any p
R
. Then, by
nding the maxima of the capacity expressions (1), (2) for
0 p
R
1, the optimal relay transmit power in the FD
mode becomes
p
R
=
min
_
1,
_
SR
+1
RD
LI
_
with AF,
min
_
1,
1
2
LI
__
4
SR
LI
RD
+ 1 1
__
with DF.
Especially, we note that the transmit power constraint is in-
voked (p
R
= 1) if
LI
SR
+1
RD
with the AF protocol or
LI
SR
RD
1 with the DF protocol.
To satisfy condition (5) for the AF protocol,
LI
is solved
from the inequality C
AF
FD
C
AF
HD
with the substitution of the
optimal transmit powers into (1), (3). Thereby, the break-even
loop interference level for the AF protocol is given by
AF
LI
=
SR
RD
_
2 +
1
A
+
1
SR
2
_
(1 +
1
A
)(1 +
1
SR
)
_
(9)
where A =
_
1 +
SR
RD
/(
SR
+
RD
+ 1) 1 if
(
SR
+ 1)(
SR
+
RD
+ 1)
RD
+ 1
SR
+ 1
RD
(10)
and otherwise
AF
LI
is given by (6) due to the transmit power
limitation (p
R
= 1). Now
AF
LI
1 for all
SR
and
RD
.
Likewise, the break-even loop interference level for the
DF protocol is determined by solving
LI
in terms of
SR
and
RD
from the inequality C
DF
FD
C
DF
HD
with (2), (4), yielding
DF
LI
=
_
SR
_
min{
SR
,
RD
} + 1 + 1
_
RD
_
_
min{
SR
,
RD
} + 1 1
_
2
. (11)
In particular, we note that
DF
LI
2 for all
SR
and
RD
.
3.3. Power allocation with a sum constraint
From the system perspective, it may be reasonable to impose
a constraint on the total transmit power of the system as some-
times assumed in related literature. By optimizing p
S
and p
R
with a sum constraint, the transmit power that is saved on one
hop can be used for improving the other hop. The optimiza-
tion problem is formulated as
(p
S
, p
R
) = arg max
(p
S
,p
R
)
C
subject to 0 p
S
, 0 p
R
, and p
S
+ p
R
2
(12)
for protocol {AF, DF} and mode {FD, HD}.
It is easy to verify that the optimal solution to (12) lies at
the boundary of the feasible region, i.e., p
S
+ p
R
= 2 with
all protocols and modes. Then, by nding the maxima of the
capacity expressions (1), (2) in terms of p
R
with substitution
p
S
= 2p
R
, the optimal relay transmit power in the FD mode
becomes
p
R
=
2
SR
+1
2
SR
+1+
(2
RD
+1)(2
LI
+1)
with AF,
4
SR
SR
+
RD
+
(
SR
+
RD
)
2
+8
SR
RD
LI
with DF.
The optimal transmit powers in the HD mode are obtained by
substituting
LI
= 0 into above expressions. Finally, with
both protocols and both modes, the optimal transmit power in
the source is given by p
S
= 2 p
R
.
The break-even loop interference level for the AF protocol
is determined by solving
LI
in terms of
SR
and
RD
from
the inequality C
AF
FD
C
AF
HD
with (1), (3) using the optimal
transmit powers, which yields
AF
LI
=
SR
+
RD
+ 2
SR
RD
(2 +
1
A
)
2
SR
RD
_
(1 +
1
A
)(2 +
1
SR
)(2 +
1
RD
),
(13)
where
A =
1 +
2
SR
RD
1 +
SR
+
RD
+
_
(2
SR
+ 1)(2
RD
+ 1)
1.
For the decode-and-forward protocol, similar calculations re-
sult in the break-even loop interference level given by
DF
LI
=
SR
+
RD
1
_
SR
+
RD
SR
+
RD
+2
SR
RD
. (14)
Especially, we see that
AF
LI
2(2
2) 0.69dB and
DF
LI
4 for all
SR
and
RD
.
3.4. Discussion
The calculated break-even loop interference levels, given by
(6), (7), (9), (11), (13), and (14), are illustrated Fig. 2. Fur-
thermore, the relative capacity improvement available by us-
ing the FD mode instead of the HD mode is exemplied in
Fig. 3 using (1)(4) in a scenario where
LI
= 6dB.
1
1
3
3
5
5
7
7
9
9
1
1
1
1
1
3
1
3
1
5
1
5
1
7
1
7
1
9
1
9
2
1
2
1
2
3
2
5
2
7
2
9
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(a) AF and uniform PA.
1
3
3
5
5
7
7
9
9
9
1
1
1
1
11
1
3
1
3
13
1
5
1
5
15
1
7
1
7
17
1
9
1
9
19
19
2
1
2
1
21
21
23
2
3
2
3
2
5
2
5
2
5
2
7
2
9
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(b) AF and PA with individual constraints. Maxi-
mum power is used (p
R
= 1) in the shaded region.
2
3
5
5
7
7
9
9
1
1
1
1
11
1
3
1
3
13
1
5
1
5
15
1
7
1
7
17
1
9
19
19
2
1
2
1
21
21
2
3
2
3
23
23
25
2
5
2
7
2
9
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(c) AF and PA with a sum constraint.
1
1
3
3
3
5
5
5
7
7
7
9
9
9
1
1
1
1
1
1
1
3
1
3
1
3
1
5
1
5
1
5
1
7
1
7
1
7
1
9
1
9
2
1
2
1
2
3
2
3
2
5
2
5
2
7
2
7
2
9
3
1
3
3
3
5
3
7
3
9
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(d) DF and uniform PA.
5
5
7
7
9
9
9
1
1
11
1
1
1
3
1
3
1
3
1
5
15
1
5
1
7
1
7
17
1
9
1
9
19
1
9
2
1
2
1
21
2
1
2
3
2
3
23
2
3
2
5
2
5
25
2
5
2
7
2
7
27
27
2
7
2
9
2
9
29
29
2
9
31
3
1
3
1
3
1
3
3
3
3
3
5
3
5
3
7
3
7
3
9
3
9
4
1
4
1
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(e) DF and PA with individual constraints.
7
9
9
1
1
11
1
3
13
1
5
1
5
1
5
1
7
17
1
7
1
9
1
9
1
9
2
1
21
2
1
2
3
2
3
23
2
3
2
5
2
5
25
2
5
2
7
2
7
27
2
7
2
9
2
9
29
2
9
31
3
1
3
1
31
33
3
3
3
5
3
5
3
7
3
7
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(f) DF and PA with a sum constraint.
Fig. 2. Contour plots for the break-even loop interference level
LI
[dB]: if
LI
LI
then C
FD
C
HD
.
The numerical results in Fig. 2 show that the FD mode
is feasible in the presence of reasonably strong loop interfer-
ence. The FD mode can also offer signicant capacity im-
provement below the break-even level as shown in Fig. 3. The
HD mode is better than the FD mode only when both hops are
weak or the loop interference power is very high.
The DF protocol is more robust to loop interference than
the AF protocol. This is because the DF protocol avoids ac-
cumulation of the loop signal due to signal regeneration. For
the same reason, with the DF protocol, the loop interference
does not degrade the performance at all when comparing to
the ideal case without loop interference if
LI
SR
RD
1.
The value of
SR
is critical with uniform PA. Especially,
when
RD
SR
, the break-even loop interference levels
are low due to the excessive transmit power usage. Hence,
the end-to-end capacity can be improved by applying the PA
methods. Especially, we see that (10) is satised for all
SR
if
RD
> (1+
SR
[dB]
R
D
[
d
B
]
(a) AF and uniform PA.
0
.
4
5
0
.
5 0
.
6
0
.
7
0.7
0
.
8
0.8
0
.
9
0.9
1
1
1
.
1
1.1
1
.
2
1
.2
1
.
3
1
.
3
1.3
1
.
4
1.4
1.4
1
.
5
1.5 1.5
1
.
6
1
.
6
1.6
1.6
1
.
7
1
.
7
1.7
1.7
1
.
8
1
.
8
1
.
9
1
.
9
1
.
9
5
1
.
9
8
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(b) AF and PA with individual constraints. Maxi-
mum power is used (p
R
= 1) in the shaded region.
0
.6
0
.7
0
.8
0
.
9
0.9
1
1
1
.
1
1.1
1
.
2
1.2
1
.
3
1.3
1
.
4
1
.4
1.4
1
.
5
1.5
1.5
1
.
6
1
.6
1.6
1
.
7
1
.
7
1.7
1.7
1.8
1.8
1
.
8
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(c) AF and PA with a sum constraint.
0
.
4
5
0
.
4
5
0
.
5
0
.
5
0
.6
0
.
6
0
.
7
0
.
7
0
.
8
0
.
8
0
.9
0
.
9
1
1
1
1
.1
1
.
1
1
.
1
1
.2
1
.
2
1
.
2
1
.3
1
.
3
1
.
3
1
.4
1
.
4
1
.
4
1
.5
1
.
5
1
.
5
1
.7
5
1
.7
5
1
.7
5
2
2
2
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(d) DF and uniform PA.
1
.9
1
.9
1
.9
5
1
.9
5
1
.9
8 1
.9
9 1
.9
9
5
0
.91 1
1
.
1
1
.1
1
.2
1.2
1
.3
1
.
3
1
.3
1
.4
1
.
4
1
.4
1
.5
1
.
5
1
.5
1
.5
1
.7
5
1
.
7
5
1
.
7
5
1
.7
5
1
.7
5
1
.7
5
2
2
2
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(e) DF and PA with individual constraints. Maxi-
mum power is used (p
R
= 1) in the shaded region.
1
.1
1
.
2
1
.2
1
.
3
1
.3
1
.
4
1.4
1
.
5
1.5
1
.5
1
.
6
1
.
6
1
.6
1
.
7
1
.
7
1.7
1
.7
1
.
8
1
.
8
1.8
1
.8
1
.8
1
.
9
1
.
9
1
.9
1
.9
1
.9
5
1
.9
5
1
.9
5
1
.
9
5
1
.9
8
1
.
9
8
10 5 0 5 10 15 20 25 30
10
5
0
5
10
15
20
25
30
SR
[dB]
R
D
[
d
B
]
(f) DF and PA with a sum constraint.
Fig. 3. Contour plots for the capacity ratio
C
FD
C
HD
when
LI
= 6dB in the full-duplex mode.
C
FD
C
HD
2 for all
SR
and
RD
.
5. REFERENCES
[1] J. N. Laneman, D. N. C. Tse, and G. W. Wornell, Cooperative
diversity in wireless networks: Efcient protocols and outage
behavior, IEEE Transactions on Information Theory, vol. 50,
no. 12, pp. 30623080, Dec. 2004.
[2] W. T. Slingsby and J. P. McGeehan, A high-gain cell en-
hancer, in Proc. IEEE 42nd Vehicular Technology Conference,
May 1992, pp. 756758.
[3] W. T. Slingsby and J. P. McGeehan, Antenna isolation mea-
surements for on-frequency radio repeaters, in Proc. 9th Inter-
national Conference on Antennas and Propagation, Apr. 1995,
pp. 239243.
[4] C. R. Anderson, S. Krishnamoorthy, C. G. Ranson, T. J. Lemon,
W. G. Newhall, T. Kummetz, and J. H. Reed, Antenna isola-
tion, wideband multipath propagation measurements, and inter-
ference mitigation for on-frequency repeaters, in Proc. IEEE
SoutheastCon, Mar. 2004, pp. 110114.
[5] H. Suzuki, K. Itoh, Y. Ebine, and M. Sato, A booster cong-
uration with adaptive reduction of transmitterreceiver antenna
coupling for pager systems, in Proc. IEEE 50th Vehicular Tech-
nology Conference, Sept. 1999, pp. 15161520.
[6] H. Hamazumi, K. Imamura, N. Iai, K. Shibuya, and M. Sasaki,
A study of a loop interference canceller for the relay stations
in an SFN for digital terrestrial broadcasting, in Proc. IEEE
Global Telecommunications Conference, Nov. 2000, pp. 167
171.
[7] K. M. Nasr, J. P. Cosmas, M. Bard, and J. Gledhill, Per-
formance of an echo canceller and channel estimator for on-
channel repeaters in DVB-T/H networks, IEEE Transactions
on Broadcasting, vol. 53, no. 3, pp. 609618, Sept. 2007.
[8] Q. Zhang, J. Zhang, C. Shao, Y. Wang, P. Zhang, and R. Hu,
Power allocation for regenerative relay channel with Rayleigh
fading, in Proc. IEEE 59th Vehicular Technology Conference,
May 2004, pp. 11671171.
[9] J. Zhang, Q. Zhang, C. Shao, Y. Wang, P. Zhang, and Z. Zhang,
Adaptive optimal transmit power allocation for two-hop non-
regenerative wireless relaying system, in Proc. IEEE 59th Ve-
hicular Technology Conference, May 2004, pp. 12131217.