Page 1
CARRIER PAIRING, A TECHNIQUE FOR INCREASING
INTERACTIVE SATELLITE SYSTEMS CAPACITY. AN
ASSESSMENT OF ITS APPLICABILITY TO DIFFERENT
SYSTEM ARCHITECTURES
by G. Gallinaro(1), R. Rinaldo(2), A. Vernucci(1)
(1) Space Engineering S.p.A.  Rome, Italy
(2) European Space Agency  ESTEC  Noordwijk, Holland
ABSTRACT
The Carrier Pairing technique, i.e. the sharing of the same frequency band for
both Forward Link and Reverse Link carriers in a star or multistar satellite
network, is here discussed. In particular a possible mechanization and its
performance in some reference scenarios are discussed to understand merits
and limitations of such technique.
It appears that, in multibeam satellite systems, carrier pairing is a viable
approach for increasing the capacity of a few selected hot spots. However, the
use of carrier pairing in all the beams, in a systematic way, may lead to a
lower overall spectral efficiency, with respect to alternative system
approaches, when the number of beams in the satellite coverage is large.
1. INTRODUCTION
The Carrier Pairing technique was originally conceived with the aim to
increase the spectral efficiency of interactive satellite systems comprising a
Hub station and a great number of User Terminals (UTs), by allocating a
common band segment to signals transmitted by the Hub and the UTs.
Generally speaking, in the Forwardlink (FL, Hub UTs) it is feasible to
manage interference resulting from signals spectral overlap, thanks to the
much higher level of the signal transmitted by the Hub compared to those
transmitted by the UTs. In the Reverselink (RL, UTs Hub) the otherwise
intolerable interference caused by the Hubtransmitted signal can be
mitigated, at the Hub receive side, by locally adding to the received
composite signal a suitably modified replica of the signal transmitted by the
Hub itself into the FL. Carrier Pairing is a known technique that has already
been adopted for commercial equipment mainly intended for operation in
globalcoverage satellite systems.
The urgent need to improve satellite systems competitiveness is leading
research to conceive solutions permitting to increase their capacity, thus
increasing economy of scale and ultimately permitting to reduce service
tariffs. The road being followed is that of proposing and assessing new
Page 2
payload architectures on the one hand (e.g. multibeam), and, on the other
hand, investigating new highperformance access solutions which Carrier
Pairing is an example of. At this regard an important issue to be dealt with is
the consistency between advanced payload architectures and the enhanced
access solution. For instance, for the Carrier Pairing case, it would be
important to assess its advantages in different system scenarios (number of
beams, traffic distributions, frequency reuse factor), so as to understand to
which extent the advantages stemming from the adoption of advanced
architectures can be addedup to those deriving from the optimization of the
access technique.
The subject paper, which is based on some of the results obtained in the
course of an ongoing contract awarded by ESA to Space Engineering, begins
introducing briefly the Carrier Pairing concept and its possible
mechanizations, and discussing the issues to be kept under control for
maximizing the technique effectiveness. Then, after defining some reference
system scenarios, the performance of Carrier Pairing in those scenarios is
discussed, showing the applicable results of a comprehensive simulation
campaign carried out in the context of the cited ESA study.
It appears that, in multibeam systems, carrier pairing is a viable approach for
increasing the capacity of a few selected hot spots. However, the use of
carrier pairing in all the beams may not be advantageous when the number of
beams in the satellite coverage is large as the intrabeam interference
becomes the limiting factor.
The paper is organized as follows. Next section contains a brief introduction
to the carrier pairing techniques and the related interference cancellation
scheme used at the GW side for recovering the RL signals. Section 3 shows
the BER /FER performances achievable on a nonlinear satellite channel at
the RL GW demodulator.
Section 4 gives finally the overall system throughput which could be
achieved in a multibeam system scenario using this technique and compares
it with that achievable with a conventional approach in which separate
frequency bands are utilized for the FL and RL. The comparison is done
assuming that the same total bandwidth and onboard power are used in both
approaches to eventually assess if carrier pairing is a viable choice for
improving the spectral efficiency of next generation broadband multimedia
satellite systems.
2. CARRIER PAIRING TECHNIQUE
In a satellite system implementing a conventional star network architecture
(i.e. with a conventional frequency plan) we need for each carrier two
different frequency bands: one for uplink and one for downlink. In the end,
for a bidirectional circuit we need four frequency bands as shown in the
figure below:
Page 3
GWSatellite UTs
FreqFL
u
FreqFL
d
FreqRL
d
FreqRL
u
Figure 1 Forward/Reverse Link frequencies in conventional systems
For example, assuming Kaband system operation, each of the four links may
be accommodated within the bandwidth shown below:
FL uplink (from GW to satellite): 27.5 ÷ 28.0 GHz
FL downlink (from satellite to User Terminals): 19.7 ÷ 20.2 GHz
RL uplink (from User Terminals to satellite): 29.5 ÷ 30 GHz
RL downlink (from satellite to GW): 18.3 ÷ 18.8 GHz
To reduce the occupied system bandwidth it is possible to share the same
bandwidth for FL and RL uplink as well as for FL and RL downlink (see
figure below).
Satellite
GW
UT
Frequp
Freqdown
Figure 2 Carrier Pairing approach
With this approach only two frequency bands are required instead of four.
There is thus no distinction between user beams and GW beams. As a
consequence, a GW can only serve a single beam.
Page 4
FL Carrier
RL Carriers
Figure 3. Signal spectrum with carrier pairing
The onboard HPA is operated in multicarrier mode as it amplifies both the
FL and RL carriers. In practice the satellite HPA amplifies a single wideband,
highpower, FL carrier plus a multitude of lowpower, narrowband, RL
carriers.
The required power density of the FL carrier (Figure 3) is much higher than
that of the RL carriers due to the different G/T of GWs and UTs.
Approximately the required relative power density of the FL carriers with
respect to the RL carriers is equal to the ratio of the G/T between the GWs
and the UTs.
In a typical system scenario, for example, the clear sky G/T of the GWs and
UTs can be respectively 33.9 dB/k and 17.0 dB/k (see for example Table 1).
Hence, the power density difference between the FL carrier and the RL
carriers may be in the order of 17 dB. Actually, because there is some
difference in the Eb/No requirement of FL and RL, due to the larger
codeword length which is possible in the FL, such difference could be a little
lower (e.g. 12 or 13 dB).
Gateways
Saturated EIRP 44.5 dBW
Antenna Gain (Tx / Rx) 45.1 dBi / 41.4 dBi
HPA Saturated Power
PostHPA Loss
Minimum Operational OBO
PreLNA Losses
Receiver Noise Figure
Clear Sky G/T
User terminals
81.7 dBW
61.0 dBi / 57.5 dBi
120 W (for 4 carriers)
2. dB
2.5 dB
0.5 dB
2. dB
33.9 dB/k
1 W
1 dB
2 dB
0.5 dB
2.5 dB
17. dB/k
Table 1 GW and UT stations RF parameters assumed in this work
Such a power density ratio will allow the UTs to demodulate and decode the
FL carriers without any special processing. Even advanced Adaptive Coding
and Modulation (ACM) schemes, like the recently standardized DVBS2, can
Page 5
be used on the FL although operating modes with the highest spectral
efficiency would not be possible due the RL carrier interference floor.
On the other hand, the GW, in order to successfully demodulate and decode
the RL carriers, has to cancel its own transmitted signal. Being this signal
known at the GW, a conventional echo canceller can be used as shown in the
figure below.
delay
 io  L
Adaptive
Filter
coefficients
adaptation
x[i]
y[i]
GW Tx
Shaping
Filter
GW
Modulator
Rx Shaping
Filter
Rx Cover
Filter
Demod
To satellite
From satellite
Figure 4. FL Interference cancellation with an adaptive filter (echo canceller)
The adaptive filter may be implemented through the LMS (Least Mean
Square) algorithm trying to minimize the Mean Square Error (MSE) between
the recovered signal (received signal minus the echo signal estimated by the
adaptive filter) and the reference signal (i.e. the GW FL signal before
transmission to the satellite).
The required adaptive Filter length is depending on:
the degree of accuracy of bulk round trip delay estimation
the memory of the channel
The adaptive filter can be split into two independently adapted filters in case
of operation at two samples/symbol, obtaining a hardware complexity
reduction of a factor of two. Indicating with f(k) (k=0 or 1) the two filters, the
adaptation rule for computing the coefficients of the filter at iteration i+1 is:
2
1
i
x
Where y(k)[i] is the received signal at iteration i (the even or the odd sample
depending on the k value) and xi is the vector of reference signal samples
within the filter memory at iteration i.
The LMS adaptation step μ has to be optimized in order to guarantee good
performance and algorithm stability.
i
i
*)()()()(
i
H
kk
i
kk
yii
xfxff
Page 6
3. WAVEFORM SIMULATION RESULTS
A waveform simulation campaign was undertaken to understand the potential
performance of carrier pairing also taking into account actual satellite non
linearities. The simulation scenario was composed by a high data rate carrier
(FL) and several lowdata rate carriers (RL) within the same bandwidth (see
Figure 3).
The satellite has been modeled as the cascade of an IMUX and a nonlinear
amplifier.
The satellite OMUX was not simulated explicitly because we were interested
in the demodulation of the low rate RL carriers whose bandwidth is much
smaller than that of a typical satellite OMUX. Anyway, if desired, the
frequency response of the OMUX may be conglobated in the receiver cover
filter.
Thermal noise may be added on both Up and DownLink. However, only
results with the uplink noise are shown here because the downlink noise
contribution was shown to be irrelevant for the demodulation of the RL
carriers at the GW side as the G/T of the GWs is enough high to make the up
link the limiting factor here.
The general block diagram of the simulator is shown in Figure 5.
FEC
Encoder
PN
Generator
Modulator
Shaping
Filter
UpLink
Noise
SIT #1
OnBoard
HPA
Down
Link
Noise
DecoderDemod
BER
EValuation
GW
FEC
Encoder
PN
Generator
Modulator
Shaping
Filter
SIT #N
FEC
Encoder
PN
Generator
Modulator
Shaping
Filter
Rx Shaping
Filter
Interference
Canceller
UpLink
Atten.
UpLink
Atten.
UpLink
Atten.
IMUX
Figure 5 Carrier pairing simulator general structure
Page 7
FL signal structure in the simulator was assumed conforming with the DVB
S2 signal specifications [2]. Both QPSK and 16APSK modulations were
considered for the DVBS2 signal to test the performance of the echo
canceller with both constant and nonconstant FL signal envelope as
difference in envelope statistics may produce different losses in a nonlinear
channel.
For the RL signal structure a DVBRCS signallike structure was used [1].
Modulation was thus QPSK. However, the recently proposed turbo code
was used instead of the currently specified DVBRCS turbo codes due to its
higher performance and flexibility [4]. Simulations with the DVBRCS
standard convolutional code were also performed.
In linear channel, interference cancellation is practically perfect, even with a
short filter (few symbols), provided that the bulk delay compensation of the
reference signal is correct, i.e. the maximum delay error (plus the channel
memory) is less than the adaptive filter impulse response length.
The adaptive filter coefficients appears stable, even in very long simulation,
apart for the case where the reference signal delay error is at the limit of the
impulse response length. In Figure 6 the echo canceller performance are
shown in the linear and uncoded case, for different ratios of Forward/Return
link power. The capability to almost perfectly cancel the interference in such
linear conditions is apparent.
1.E04
1.E03
1.E02
1.E01
56789 1011
Es/No (dB)
BER
Ideal
FL/RL= 16 dB
FL/RL= 10 dB
Canceller Filter= 10 taps
QPSK, Uncoded
mu=0.0005
Figure 6. Echo canceller performance, uncoded case
In order to analyze the effects of satellite nonlinearity, the linearized TWTA
model specified in Annex H of the DVBS2 specifications [2] was used.
As expected, the performance degradation is strongly dependent on the
TWTA operating Input BackOff (IBO). In Figure 7 and Figure 8 simulation
results are shown for the case where a rate ½ convolutional FEC code is used
and the FL/RL power ratio is either 20 dB or 10 dB.
Page 8
BER degradations at IBO=10 dB is negligible. At IBO ≤ 7 dB, however,
degradations start to become significant. It is, in fact, not possible, in a non
linear channel, to completely cancel the FL carrier interference using a linear
canceller.
1.00E06
1.00E05
1.00E04
1.00E03
1.00E02
1.00E01
1.00E+00
4 4.55 5.566.57 7.5
Es/No (dB)
BER
IBO = 50 dB
IBO = 10 dB
IBO= 7 dB
IBO= 4 dB
FL Power / RL Power =20 dB
Canceller Filter= 10 taps
mu =0.0005
QPSK, Conv Code Rate 1/2
Without Predistortion
Figure 7. Echo canceller performance with TWTA nonlinearity
1.E07
1.E06
1.E05
1.E04
1.E03
1.E02
1.E01
44.55 5.5
Es/No (dB)
66.57 7.5
BER
IBO = 10 dB
IBO = 8 dB
IBO = 7 dB
IBO = 4 dB
Linear
FL Power / RL Power =10 dB
Canceller Filter= 10 taps
mu =0.0005
QPSK, Conv Code Rate 1/2
Without Predistortion
Figure 8. Echo canceller with TWTA and F/R = 10 dB
A predistorter replicating the HPA nonlinearity for the local reference signal
can be used before the canceller to try compensating for the nonlinearity (see
Figure 9). A waveform predistorter with 2 sample/symbol (Fractionally
Spaced Predistorter) was thus tested. Improvements were however limited
due to:
Page 9
Interference
Canceller
NonLinear
Device
GW Rx Signal
GW Reference Signal
Cleaned RL Signals
Figure 9. Predistorter structure
the nonlinearity is with memory (due to the IMUX filter presence) while
the assumed precompensator is without memory
the effects of the RL signal in the nonlinear channel cannot be neglected.
Some performance improvements were obtained only for the highest ratio
between FL and RL total signal power (greater than 10 dB). However, for a
FL/RL ratio equal to 10 dB, a performance improvement of only 0.2 dB
(@BER=1E5) at IBO= 7 dB was obtained with the assumed rate ½
convolutional FEC code (Figure 10).
1.E06
1.E05
1.E04
1.E03
1.E02
3 3.54 4.55 5.56
Es/No (dB)
BER
IBO= 50 dB
IBO = 10 dB
IBO = 8 dB
IBO = 7 dB
FL Power / RL Power =10 dB
Canceller Filters= 10 taps
mu =0.0005
QPSK, Conv Code Rate 1/2
With predistortion
Figure 10. Echo canceller performance with predistortion (F/R:10 dB)
Figure 11 and Figure 12 shows respectively the BER and FER in the same
conditions as in Figure 10 but with a rate ½ Turbo code. A DVBRCS
packet of 1504 symbols (1504 information bit) was assumed.
The effects of using the 16APSK modulation on the Forward Link carrier has
been also considered as the resulting envelope fluctuation of the FL carrier
may have an impact on the performance of the RL signals. In particular, the
BER performance obtained with the echo canceller when 16APSK is used on
the FL is shown in Figure 13. Further, Figure 14 compares the obtained RL
carrier BER/FER performances when either 16APSK or QPSK are used on
the FL carrier. The comparison is shown for the case of IBO=7 dB and FL/RL
Page 10
power ratio of 16 dB. It appears that 16APSK on the FL produces a loss
which is about 0.1 dB greater than the one obtained with a QPSK carrier on
the FL.
Figure 15 summarizes the achievable performances on RL carrier
demodulation for different FL/RL power ratio. Finally, in Figure 16 the effect
of the predistorter is shown for IBO = 6 dB and F/R = 16 dB.
1.00E08
1.00E07
1.00E06
1.00E05
1.00E04
1.00E03
1.00E02
1.00E01
1.00E+00
1.52 2.53 3.54 4.5
Es/No (dB)
BER
IBO = 4 dB
IBO = 6 dB
IBO = 7 dB
IBO = 8 dB
IBO = 10 dB
IBO = 50 dB
FL/RL Power =10 dB
Figure 11. BER with Turbo, QPSK and F/R=10 dB. Predistortion assumed.
Page 11
1.E04
1.E03
1.E02
1.E01
1.E+00
1.52 2.53 3.54 4.5
Es/No (dB)
FER
IBO = 4 dB
IBO = 6 dB
IBO = 7 dB
IBO = 8 dB
IBO = 10 dB
IBO = 50 dB
Figure 12. FER with Turbo, QPSK and F/R=10 dB. Predistortion assumed
1.E07
1.E06
1.E05
1.E04
1.E03
1.E02
1.E01
1.E+00
1.5 1.71.9 2.12.32.52.7 2.9 3.1
Es/No (dB)
BER
IBO = 7 dB
IBO = 8 dB
IBO = 1 0 dB
IBO = 50 dB
FL Modulation = 16APSK
RL Modulation QPSK,
RL Code rate=1/2
FL/RL Power Ratio=10 dB
Figure 13. BER on RL carriers with interference from a 16APSK Forward link
carrier with FL/RL power ratio equal to 10 dB. Rate ½ Turbo FEC Code.
Page 12
1.E06
1.E05
1.E04
1.E03
1.E02
1.E01
1.E+00
1.82 2.22.4
Es/No (dB)
2.62.833.2
BER  FER
BER with QPSK FL
BER with 16APSK FL
FER with QPSK FL
FER with 16APSK FL
IBO= 7 dB
FL/RL Power = 16 dB
Figure 14. Performance of RL carriers (IBO=7 dB, FL/RL power ratio = 16
dB) with either QPSK or 16APSK interfering carrier on the Forward link.
Rate ½ Turbo FEC Code
1.E05
1.E04
1.E03
1.E02
1.E01
1.E+00
5 10152025 30 35
FL/RL Power ratio (dB)
BER  FER
BER
FER
IBO = 6 dB
Es/No = 3 dB
Figure 15. RL performance for different FL/RL power ratio, Es/No=3 dB,
IBO=6 dB. QPSK modulation on the FL carrier. Rate ½ Turbo FEC Code
Page 13
1.E06
1.E05
1.E04
1.E03
1.E02
1.E01
1.E+00
1.522.53 3.54
Es/No (dB)
BER  FER
BER w/o predistoter
BER w predistorter
FER w/o predistorter
FER w predistorter
IBO = 6dB
FL/RL Power = 16 dB
Figure 16. Predistorter performance for QPSK forward link, IBO=6dB and
F/R=16 dB. Rate ½ Turbo FEC Code
On the basis of the obtained results the optimum OBO can be evaluated with
respect to the global losses introduced in the system. The optimum OBO was
found to be about 2 dB (corresponding to an IBO of about 7 dB for the
considered TWTA) and the corresponding total loss about 3 dB.
4. SYSTEM PERFORMANCES
The advantage provided by the Carrier Pairing technique as far as the increase
of overall spectral efficiency of a satellite system may be more or less
significant depending on the considered system configuration. Ideally, one
should compare the cost per transmitted bit of each possible alternative
system. However, assessing the system cost is not trivial. We took here a
pragmatic approach. In particular, we designed a Kaband reference satellite
system according to current best practice and evaluated its spectral efficiency
with and without Carrier Pairing.
Figure 17 shows the antenna coverage of the reference system assumed for
the analysis. In particular, a European coverage implemented by means of 88
spot beams was assumed. Each spot beam had a beamwidth of approximately
0.5° (corresponding to an antenna gain of about 47 dBi at beam edge). A
payload with such a large number of beams, although challenging, is actually
within the capability of current technology.
For the reference system we assumed that a conventional frequency reuse
based on a threecolor scheme is adopted.
Page 14
Figure 17. Assumed user link antenna coverage
Quantitative simulation results reported here assumed a ground segment
composed by GWs and UTs whose RF characteristics are shown in Table 1.
Such characteristics are in line with those typical found in GWs and UTs.
Similarly, characteristics of the onboard transponders are given in Table 2.
We assumed to allocate 13 W RF saturation power for each 25 MBaud FL
channel eventually used in each beam. Each beam carried a number of FL
carriers variable from 1 to 5 depending on the beam traffic load. The number
of RL carriers per beam was also proportional to the allocated FL beam
bandwidth where the proportionality factor was equal to the ratio of the FL
carrier bandwidth to the RL carrier bandwidth. A single HPA per beam was
assumed. Hence if the beam is sized for 5 FL carriers a saturation power of
13x5= 65 W is assumed. The assumed saturation power shall, in the case
carrier pairing is used, also be shared with the RL carriers which are hosted in
the same transponder. For the reference system not using carrier pairing the
same total RF onboard power was maintained but separate FL and RL
transponders were used. Each FL transponder then used a power slightly
lower than the one associated with carrier pairing operation as some power
allowance shall be also considered for the RL transponders. Further the FL
transponder OBO was dependent on the number of carrier per HPA. For
beams where a single FL carrier was allocated an OBO of 1 dB was
considered (also to allow for operation with 16APSK modulation). For
multicarrier HPAs the nominal OBO was 1.6 dB, 2.3 dB, 2.6 dB or 2.8 dB
depending whether the number of carriers were 2, 3, 4 or 5.
Page 15
HPA Saturated RF Power
Total Saturated RF Power
Nominal Operational OBO
Av. TWTA Eff. @3 dB OBO
HPA DC Power
Maximum HPA Bandwidth
Spot Beamwidth
Antenna Gain (EOC)
PostHPA Loss
Receiver Noise Figure
PreLNA Loss
Variable: 13 W to 65 W
4615 W
3 dB
27.7%
8.3 KW
325 MHz
0.5 deg
47.3 dBi
2.5 dB
2.5 dB
1.5 dB
Table 2 Satellite Repeater characteristics – Carrier Pairing case
For the carrier pairing case, taking into account the waveform simulation
results, an OBO in the order of 2 dB was found the best choice for operation
with a single FL carrier (plus the associated number of RL carriers). However,
taking into account that a single HPA may also be utilized by multiple FL
carriers, the optimal OBO may be larger. For system performance
computation we assumed a conservative value of 3 dB independently of the
number of FL carriers per beam.
Modulation Code Rate
QPSK 1/4
QPSK 1/3
QPSK 2/5
QPSK 1/2
QPSK 3/5
QPSK 2/3
QPSK 3/4
QPSK 5/6
8PSK 3/5
8PSK 2/3
8PSK 3/4
8PSK 5/6
16APSK 3/4
16APSK 5/6
Req. Es/No (dB)
1.31
0.3
1.24
2.54
3.77
4.64
5.57
6.72
7.92
9.04
10.33
11.77
13.01
14.41
Table 3 FL ACM modes and corresponding required Es/No for the non
precoded mode. For the precoded case the above figures have to be degraded
by 0.5 dB.
Page 16
Modulation
Order
4
4
4
4
4
4
8
8
8
16
16
16
Coding
Rate
1/3
1/2
3/5
2/3
3/4
6/7
2/3
3/4
4/5
3/4
5/6
9/10
Efficiency
(bps/Hz)
0.66
1.00
1.20
1.33
1.50
1.71
2.00
2.25
2.40
3.00
3.33
3.60
Theor.
Eb/No (dB)
1.3
1.15
1.66
2.07
2.68
3.78
3.86
4.85
5.38
5.70
6.77
7.85
Theor.
Es/No (dB)
0.46
1.15
2.45
3.32
4.44
6.12
6.87
8.37
9.18
10.47
12.00
13.41
Req. Es/No
(dB)
1.55
3.1
4.5
5.3
6.4
8.1
9.2
10.7
11.5
13.5
15.0
16.4
Table 4 RL ACM modes and corresponding required Es/No
Adaptive Coding and Modulation has been assumed. The operating modes
and required Es/No for the FL are shown in Table 3. The required Es/No also
includes an additional 0.5 dB margin for ACM operation (apart for the lowest
mode).
For the RL, an ACM enhanced DVBRCS is assumed with Turbo coding.
The table of physical layer modes and required Es/(No+Io) are shown in
Table 4.
Also in this case, an additional 1.5 dB margin has been included in all modes
but the most protected one as safeguard against errors in ACM adaptation.
Finally, for the carrier pairing case, performance of the RL ACM modes have
been degraded by 0.5 dB with respect to those shown in Table 4 (used for the
conventional system) to account for the uncompensated residual FL carrier
interference.
A threecolours frequency reuse scheme was firstly used also for the carrier
pairing system.
A synthesis of the throughput and availability obtainable on the FL and RL
for different repartition between FL and RL of the onboard power is given in
Table 5.
Power repartition refers to the division of the HPA output RF power between
FL and RL carriers. It shall be observed that, being the transponder operated
in an almost linear region, the same power repartition is assumed at the
transponder input, i.e. any small signal suppression effect is assumed
negligible.
Page 17
Power Repartition FL to RL
7 to 1
99.78%
14.8 Gbit/s
99.24%
15.4 Gbit/s
4 to 1
99.77%
10 to 1
99.76 %
20.7 Gbit/s
8.9 %

FL Availability
FL Throughput
RL Availability
RL Throughput
18.9 Gbit/s
38.97%
5.03 Gbit/s
Table 5 Performance of carrier pairing for three different power repartition
hypothesis
Unfortunately, for FL to RL power allocation ratios of about 10 dB or higher,
the RL carriers suffer of too much interference from the other cofrequency
beams. It shall, in fact, be stressed that, whilst it is possible to cancelout
almost all FL interference from the same beam, canceling out the interference
caused by FL carriers from other beams is not possible (at least with a simple
echo canceller).
This poor performance of the RL with a 3 colour frequency reuse scheme
would have been probably predictable looking at the C/I distribution with a
three color frequency reuse scheme (see, at this regard, Figure 18 and Figure
19). It appears that the worst case uplink C/I is in the order of 11 dB. For a
power ratio between FL and RL of 10 to 1, this implies that C/I on RL is 10
dB worse than it would appears from Figure 18 and Figure 19 below. This
would lead to a significant penalization of the RL performance.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
10 1214 16 18202224
C/I (dB)
Cumulative Distribution
0
0.05
0.1
0.15
0.2
0.25
Probability Density
Cumulative Distribution
Probability Density
106 beams
0.5° beamwidth
Frequency Reuse 1 /3
Central beam
Figure 18 Cumulative distribution and probability density of uplink C/I in the
selected coverage with 0.5° beamwidth. Three color pattern frequency reuse.
Statistics computed for the central beam. Perfect power control
Page 18
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1012141618202224
C/I (dB)
Cumulative Distribution
0
0.02
0.04
0.06
0.08
0.1
0.12
Probability Density
Cumulative Distribution
Probability Density
106 beams
0.5° beamwidth
Frequency Reuse 1 /3
Beam #105
Figure 19 Cumulative distribution and probability density of uplink C/I in the
selected coverage with 0.5° beamwidth. Three colour pattern frequency reuse.
Statistics computed for the most peripheral beam. Perfect power control
A power ratio of 10 dB is thus only useful if an asymmetric traffic is expected
between FL and RL.
For symmetric traffic load a FL/RL power repartition of 4 to 1 would be
required to get an acceptable RL throughput. A total throughput of 14.8 Gbit/s
and 15.4 Gbit/s in fact results for that repartition case respectively for the FL
and RL. These numbers have to be compared against 14.4 Gbit/s and 18.0
Gbit/s which would be achievable respectively for the FL and RL of a
conventional system using separate FL and RL transponders (and frequency
bands).
There is a small improvement of the FL capacity paid by a reduction in the
RL throughput.
Availability is instead slightly reduced (particularly for the RL) with respect
to the performance figures achieved in the reference scenario (99.9% and
99.77%).
In conclusions, carrier pairing does not appear particularly attractive in the
present context at least for symmetric (FL to RL) traffic. The following
considerations are however to be done.
In our evaluation we have assumed that the RL carriers are fully used with a
filling factor for the TDMA frame of 100%. This in practice never happens.
So we could design the system with a nominal power ratio between FL and
RL, e.g. 4 to 1, but the actual interference experienced by the FL carrier
would be somewhat smaller due to the fact that, statistically, not all RL
carriers are simultaneously active. At this regard, to control latency in packet
Page 19
application on the RL, enough free capacity should be available to exploit the
free capacity assignment mechanism in DVBRCS. Whilst unused free
capacity assignments are a complete waste in traditional systems, in this case
there is a partial compensation with an increase of the FL efficiency which
can be exploited to further reduce the FL to RL nominal power allocation
ratio. However, with this approach interference on the FL carrier could be
quite unpredictable especially when a small number of beams is considered.
In conclusion, it is our feeling that, apart for possible cases with very
unbalanced FL to RL traffic, Carrier Pairing may find useful usage when only
a few beams needs to operate in carrier pairing mode (either because the
coverage is based on few beams or because there are a few hot spots in the
coverage).
For example Table 6 shows the results when only 8 out of 88 Beams at the
center of the coverage are used in carrier pairing mode, whilst the other
beams are used for RL only.
FL Availability
FL Throughput
RL Availability
RL Throughput
99.793%
1.2 Gbit/s
99.648%
18.1 Gbit/s
Table 6 Performance using carrier pairing in 8 hot spotbeams. A FL/RL
Power Allocation Ratio of 7 to 1 was used.
The FL throughput is the one achieved in the 8 beams used in carrier pairing
(to be added to the throughput provided in the normal FL transponders. The
RL throughput is the sum of throughput of all 88 RL transponders including
the 8 used in carrier pairing. For the transponders not used in carrier pairing
the characteristics of the conventional RL transponders are used.
5. CONCLUSIONS
An analysis of the performance of Carrier Pairing was performed to evaluate
if it can provide an improvement of system spectral efficiency in next
generation broadband satellite systems. It appears that in systems with large
number of beams the capability of carrier pairing to improve spectral
efficiency is quite limited at least when symmetric FL and RL traffic is
expected. However, it is expected that carrier pairing may be a possible
solution to improve the FL capacity if asymmetric traffic is expected
particularly when only a few hot spots are expected.
References
[1] ETSI EN 301 790, “Digital Video Broadcasting (DVB); Interaction
channel for satellite distribution systems, V1.3.1, 2002
Page 20
[2] ETSI EN 302 307 v.1.1.1, “Digital Video Broadcasting; Second
generation framing structure, channel coding & modulation systems for
Broadcasting, Interactive Services, News Gathering and other broadband
satellite applications, V1.1.1., 2005
[3] M Debbah, G. Gallinaro, R. Muller, R. Rinaldo, A. Vernucci,
“Interference Mitigation for the ReverseLink of Interactive Satellite
Networks,” submitted to Globecom 2006.
[4] S. Benedetto, R. Garello, G. Montorsi; C Berrou, C. Douillard et al,
“MHOMS: HighSpeed ACM Modem for Satellite Applications”, IEEE
Wireless Comm., April 2005.