Content uploaded by Giuseppe Cocco

Author content

All content in this area was uploaded by Giuseppe Cocco on Mar 20, 2018

Content may be subject to copyright.

Co-Channel Interference Cancellation at the User

Terminal in Multibeam Satellite Systems

G. Cocco†, M. Angelone¶and A.I. Perez-Neira1,2

†German Aerospace Center - DLR

Oberpfaffenhofen, D-82234, Wessling, Germany

¶European Space Agency - ESTEC, Noordwijk – The Netherlands

1Centre Tecnol`ogic de Telecomunicacions de Catalunya – CTTC

Parc Mediterrani de la Tecnologia, Av. Carl Friedrich Gauss 7 08860, Castelldefels – Spain

2Department of Signal Theory and Communications

Universitat Politecnica de Catalunya, Barcelona, Spain

giuseppe.cocco@dlr.es, martina.angelone@esa.int, ana.perez@cttc.es

Abstract—We study the applicability of soft interference can-

cellation in the forward link of commercial multibeam satellite

systems with focus on mobile terminals. We adopt a standard

currently used in commercial satellite systems as a reference.

The multibeam satellite antenna radiation diagram has been

generated using a physical optics reﬂector model, while state-of-

the art channel models have been used for the land mobile satel-

lite (LMS) channel. The interference pattern has been derived

through a system simulator developed by the European Space

Agency (ESA). Starting from the analysis of the interference

pattern we study the application of a low complexity soft

interference cancellation scheme. Our results show that, under

realistic interference and propagation conditions and for existing

standards, a two-colors frequency reuse scheme can be employed

while guaranteeing service availability across the coverage and

keeping the complexity at the user terminals relatively low.

I. INTRODUCTION

Bandwidth limitation is one of the main capacity-limiting

factors in wireless communications systems. A common prac-

tice to overcome bandwidth shortage in both satellite and

terrestrial networks using multiple beams/cells consists in

dividing the available spectrum into sub-bands (colors) and

reuse them over non-adjacent geographical regions. However,

if on the one hand the reuse of frequency allows for a more

efﬁcient utilization of the spectral resources, on the other hand

it determines an increase of the co-channel interference (CCI)

due to the non-ideal antennas radiation patterns. Despite the

improvements in antennas technology, undesired side lobes

still constitute challenging problem in geostationary (GEO)

satellite communications since the interference coming from

co-channel beams can heavily affect the reception of the

desired signal at the user terminal such that either the link

throughput or the availability are penalized. This problem is

exacerbated by the use of aggressive frequency reuse patterns.

Interference cancellation techniques at the user terminal (UT)

represent a possible solution to this problem. Many different

interference cancellation techniques have been proposed up to

date. A comprehensive overview is presented in [1]. A low

complexity option for mitigating interference is to adopt a

maximum a posteriori (MAP) symbol detector. Such detector

has the drawback of having a complexity that grows expo-

nentially with the number of signals to detect. In order to

keep complexity low, trading part of the performance, several

simpliﬁed schemes have been proposed in literature such as

[2], [3] and [4]. Iterative decoding has been shown to achieve

the multiple access channel (MAC) capacity in [5], by inte-

grating error control coding with multiple access interference

suppression. In [6] two iterative low complexity algorithms

for adjacent channel interference (ACI) cancellation in satellite

systems are presented. In [7] the authors proposed a parallel

multi-user detector for adjacent channel interference cancella-

tion in the return link of Inmarsat’s Broadband Global Area

Network (BGAN) system.

In the present paper we study the applicability of soft

co-channel interference cancellation in the forward link of

a satellite system with high frequency reuse based on a

realistic scenario with focus on mobile terminals. The results

presented here have been developed within the Advanced

Research in Telecommunications Systems (ARTES) [8] project

Next Generation Waveform for Increased Spectral Efﬁciency

(NGWISE) founded by the European Space Agency (ESA) [9].

The standard adopted in the European Telecommunications

Standard Institute (ETSI) Satellite Component of UMTS (S-

UMTS) [10] has been used as a baseline. Such standard is

currently used in commercial satellite systems [11]. The multi-

beam satellite antenna radiation diagram has been generated

through a commercial software used for satellite antenna de-

sign and analysis, while the interference pattern has been cal-

culated using a system simulator developed by ESA. A state-

of-the art channel model has been adopted for the land mobile

satellite (LMS) channel. Unlike most of previous works, we

start from the analysis of the interference distribution across

the coverage area. Based on the interference distribution we

propose a low complexity interference management solution.

It is worth noting that our work differs from [6] and [7]

in that co-channel rather than adjacent channel interference

is considered. In fact, whenever standard channel spacing is

considered 1and an aggressive frequency reuse scheme is

1this may not be the case if techniques such as time-frequency packing are

applied [12]. However, this falls out of the scope of the present work.

applied, indeed CCI becomes the most relevant source of

interference in the system as its level is much higher with

respect to the ACI. Unlike in [7] we consider the forward

link rather than the return link. Interference cancellation in

the forward link is constrained by the complexity at the UT,

especially within the LMS context. We show that, assuming a

realistic interference spatial distribution, the optimal detector

can be applied at the receiver with affordable complexity if

an appropriate synchronization of the signals at the feeder

stations and at the satellite is provided. Our results show that

frame error rates as low as 10−3can be achieved in the whole

covered area while using a two-color frequency reuse scheme

with a dramatic increase in spectral efﬁciency with respect

to coloring schemes usually adopted in commercial satellite

systems.

The rest of the paper is organized as follows. In Section II

the system model is presented while in Section III we describe

the proposed solution specifying the required modiﬁcations to

ETSI standard [13]. The numerical results are presented in

Section IV while Section V summarizes the main contributions

of the paper.

II. SYSTEM MODEL

Let us consider the forward link of an interactive geosta-

tionary (GEO) multibeam satellite system with 210 user-link

beams operating in L/S band. Each beam occupies half of the

available user-link bandwidth and a two-color frequency reuse

pattern is adopted, with a single polarization per beam.

Due to the satellite antenna radiation pattern each beam

suffers from the interference generated by the closest co-

channel beams. For a frequency-reuse of two, ten co-channel

beams can be identiﬁed as the most signiﬁcant contributors to

the total interference. The geographical position of reference

and interfering beams is shown in Figure 1, where the red

color identiﬁes the reference beam while orange is used for

the co-channel interfering beams. In order to be representative

of the best case and the worst case scenarios we considered

two beams for our analysis, namely beam 105 and beam 110,

one at the center and one at the edge of the global coverage,

respectively.

Given a UT in a given beam we refer to the desired signal

as reference signal. Single antenna terminals are considered.

The received signal at time twhen Nint interferers are

present is:

y(t) = h(t)gC(t)xC(t)

+

Nint

X

nint=1

gI

nint (tnint )xI

nint (tnint )#+n(t),(1)

where tnint =t−τnint ,τnint being the time offset of interferer

number nint, while

gI

nint (t) = GI

nint ej(2π∆νnint t+ϕnint ),(2)

GI

nint being the antenna gain of the co-channel interfering

beam nint in the direction of the UT, normalized to the gain of

the reference signal, while ∆νnint and ϕnint are the frequency

Fig. 1. Considered reference and interfering beams and conventional

numbering. Reference and interfering beams are shown in red and yellow,

respectively.

and phase offsets with respect to the local oscillator at the UT,

respectively. Similarly we deﬁned

gC(t) = GCej(2π∆νCt+ϕC),(3)

with GC= 1.

Signals xC(t)and xI

nint (t),nint ∈ {1,...,Nint}, are

the reference and the interfering signals, respectively. The

interfering signals (and similarly the desired one) can be

expressed as

xI

nint (t) =

NCW

Nint

X

l=1

snint (l)g(t−lT nint

s),(4)

where g(t)is a root-raised cosine pulse with roll-off α,snint (l)

represents the l−th received symbol from interferer nint ,

Tnint

sis the symbol duration while NC W

nint is the number of

modulated symbols in a codeword for interferer nint .h(t)

takes into account the channel effect (phase rotation and prop-

agation loss). Note that h(t)is a common multiplying factor

for all signals, since all waveforms originate from the same

spacecraft and in a forward link system all signals cover the

same path to the UT. We assume that the maximum frequency

offset is such that ∆νnint TS≪1/100,∀nint ∈ {1,...,Nint}.

The sample taken at time tkafter matched ﬁltering and

sampling of signal y(t)is:

yk=h(tk)gC(tk)s(k)

+

Nint

X

nint=1

gI

nint (tnint

k)

NCW

nint

X

l=1

snint (l)g(tnint

k−lT nint

s)

+wk,

(5)

where tnint

k=tk−τnint while wk’s are independently and

identically distributed (i.i.d.) zero mean complex Gaussian

random variables with variance σ2in each component. The

interfering signals gains GI

nint are determined by the satellite

antenna radiation pattern. The use of a realistic antenna pattern

is of fundamental importance for the selection and the per-

formance assessment of an adequate interference cancellation

technique at the UT. In the following we give details about

the antenna pattern and system model adopted in the present

paper.

A. System Simulations and Antenna Pattern Models

This section describes the system simulator developed by

ESA and used to compute the interference pattern as well as

the models used to create the considered antenna pattern.

The ESA satellite communication systems analysis tool,

developed in MATLAB, performs a multi-dimensional space-

time link budget over a uniform latitude-longitude grid of

users, averaging over a user-deﬁned set of time availabilities

with the related channel attenuations and availability proba-

bilities. The reference propagation models are based on ITU

recommendation [14] and it is assumed that the trafﬁc request

across different beams is uniform. For the sake of this study we

focus on clear sky conditions, since atmospheric attenuation

does not represent a serious impairment in L/S band. Each

user of the grid is assigned to a speciﬁc beam if the gain of

such beam in its location is the highest across the coverage.

Then, based on the frequency plan and on the consequent beam

coloring, the resulting interference pattern and distribution are

calculated. The simulated system foresees the use of Adaptive

Coding and Modulation (ACM) that enables each user to select

the most efﬁcient modulation and coding (ModCod) scheme

allowed by the link condition. In general the ACM in LMS

systems is more challenging with respect to the case of ﬁxed

terminals due to the rapid changes in the communication

channel induced by the terminal motion. In [13] a return

channel is used to feed-back the measured SNR (or SINR)

to the Bearer Control Layer. The information is used at the

control unit to select the bearer according to a target QoS.

Such system is used to adapt the communication rate to the

long-term channel variations only, since short-term fading is

covered by the link margin [13, Section 7]. Further analysis

in the implementation of the ACM mechanism is out of the

scope of this paper.

The down-link signal to interference ratio in the point x

belonging to beam iis given by:

C

IDL

co

(x) = PT X SAT (i)Gsat

T Xco−po (i, x)

PNco−ch

j=1 PT X SAT (j)Gsat

T Xco−po (j, x),(6)

where:

•PT X SAT (i)is the saturated power per carrier of beam i

•Gsat

T Xco−po (i, x)is the co-polar satellite TX antenna gain

of beam i in the location x

•PT X SAT (j)is the saturated power per carrier of beam

j; note that in the analysis it has been assumed that all

the carriers have equal power and therefore this term can

be assumed to be a constant

•Gsat

T Xco−po (j, x)is the co-polar satellite TX antenna gain

of co-channel beam beam jin the location x.

We assume that solid state power ampliﬁers (SSPAs) are used

on-board the satellite payload. As in this analysis we focus

on the ﬁrst Nco = 10 strongest interferers received at the user

terminal we further deﬁne for each of them:

C

Ijco

(x) = Gsat

T Xco−po (i, x)

Gsat

T Xco−po (j, x),(7)

as the signal to co-channel interference related to the j-th co-

channel interferer, assuming that Ij≥Ij+1 ∀j∈ {1,...,Nco}

and INco+1 = 0. As for the considered antenna pattern, a

commercial software for antenna design analysis and coverage

planning has been used to reproduce a beam pattern similar

to the one of the commercial system BGAN [11], which

adopts the ETSI standard [13]. The software is based on

physical optics reﬂector modeling and allows for accurate

characterization of the directivity of both the co-polar and the

cross-polar ﬁelds, as well as scan-aberrations and losses [15].

A geostationary satellite in the 25 deg East orbital position

has been considered.

Finally, the reﬂector has been modeled with the parameters

listed in Table I.

III. PROPOSED SOLUTION

We consider the forward bearers family of the standard [13].

We aim to ﬁnd an interference cancellation solution at the UT

that is at the same time efﬁcient and that has low complexity.

We propose to split the complexity between system and UT

levels. In the following the modiﬁcations required at system

level with respect to the standard are detailed.

TABLE I

SATELLI TE REFLECTOR PARAMETERS.

Parameter Value

Aperture size [m] 9

F/D 1.34

Beam spacing/θ3dB [deg] 1.363

Crossover Level [dB] -3

Aperture Efﬁciency 59.1%

Directivity [dBi] 40.85

A. System Level

The modiﬁcations that would be required to [13] are here-

after speciﬁed and the related implications and feasibility

discussed.

1) It is assumed that the symbol rate RI

s= 1/T I

sof the

strongest interferer is the same as that of the reference

signal RC

s= 1/Ts,Tsbeing the symbol period of the

reference signal. Although in principle different channel

code rates, modulations and FEC block sizes may be

used in the two signals, the simulation results we present

in Section IV show that there are some restrictions to the

modulations and code rates that can be adopted. Note

that assuming the same symbol rate for the reference

and the interfering signals implies Tnint

s=Ts,∀nint ∈

{1,...,Nint}, in expression (5).

2) The symbols of reference and interfering signals are

aligned such that the intersymbol interference (ISI)-free

sample instants of the reference signal correspond to the

ISI-free sample instants of the interferer, which implies

τnint = 0 ∀nint ∈ {1,...,Nint}in expression (5).

However, in Section IV we show that this constraint can

be relaxed up to a certain extent.

3) The receiver knows the modulation used by the inter-

ferer. This information can be made available to the UT

through the global beam and using knowledge of the

user position, which is currently foreseen in [13] through

GPS signal. Knowing the position with respect to the ref-

erence beam, a user could derive which is the strongest

interfearing beam. The information about the modulation

used in each beam (and thus also in the interfearing

beam) during a given time slot is transmitted over the

global beam. We use this assumption as it simpliﬁes the

description of the proposed scheme, although in Section

IV we will show that it can be actually removed.

B. MUD at the User Terminal

We assume that the channel of both the reference signal and

the strongest interferers as well as the ISI-free sample instants

of the reference signal can be estimated. This assumption is

usually taken in most multi-user detection (MUD) systems.

Channel estimation can be performed using the pilot symbols

inserted at regular intervals in the frame as foreseen in [13].

In case the pilot symbols of reference and interfering signal

overlap, joint estimation methods may be adopted (e.g., E-M

algorithm [16])2. An extensive literature is available on the

subject and further analysis is out of the scope of this paper.

We further assume that conditions 1→3of Section III-A

hold.

In case no interference is present, in a typical receiver

the turbo decoder is fed with the log-likelihood ratio (LLR)

vector of the sampled received signal. The j-th component,

j∈ {1,...,RNcw

C}of the LLR vector for QPSK signalling

2A similar problem has been addressed in [17] where the feasibility of the

joint estimation of phase, amplitude and frequency offsets of ﬁve colliding

signals is studied.

and using the Grey mapping scheme of [10] can be expressed

as:

LLRj= log P r{bj= 1 |yk}

P r{bj= 0 |yk}= log Pk,s2+Pk,s3

Pk,s0+Pk,s1,(8)

for j= 2k−1, while

LLRj= log Pk,s1+Pk,s3

Pk,s0+Pk,s2,(9)

for j= 2k, where Pk,snis the probability to observe the

sample ykconditioned to the transmission of the symbol sn,

n∈ {0,1,2,3}, while bjindicates the j-th coded bit in the

transmitted codeword. Equation (8), and similarly Equation

(9), is derived taking into account that, according to the

considered mapping, symbols s2and s3correspond to a bit

pair with the ﬁrst bit equal to 1, while the ﬁrst bit of the pair

mapping to s0and s1is equal to 0. Equation 8 can be easily

extended to the case of 16 QAM modulation. The probability

Pksnis proportional to:

Pk,sn∝exp |yk−h(tk)GCsn|2

2σ2.(10)

In the case of a single interferer with constellation size M,

the probability that the k-th symbol of the reference signal

s(k)is equal to sncan be expressed as:

Pk,sn=

M−1

X

m=0

PsI

mPk,sn,sI

m,(11)

where Pk,sn,sI

mrepresents the probability to receive ykcon-

ditioned to symbols snand sI

min the reference and in

the interfering signals, respectively, while PsI

mrepresents the

probability of transmitting symbol sI

m, which is assumed to

be equal to 1/M. The probability Pk,sn,sI

mis proportional to:

Pk,sn,sI

m∝exp |yk−h(tk)gC(tk)sn−h(tk)gI

nint (tk)sI

m|2

2σ2.(12)

This can be easily extended to the case of a generic number of

interferers Nint each with its phase and frequency offsets and

amplitude, leading to the following expression for the optimal

symbol detector,

Pk,sn=

M1−1

X

m1=0

···

MNint −1

X

mNint =0

Nint

Y

j=1

PsI

jPk,sn,sI

m1,...,sI

mNint

,(13)

where Mnint is the constellation size of interferer number

nint. The complexity of expression (13) grows exponentially

with the number of interferers.

Once the a-priori probabilities for the desired signal have

been derived they can be used to calculate the L-values to be

fed to the turbo decoder. In this case the only modiﬁcation

at the receiver side with respect to the standard terminal is

limited to the signal detector, while no modiﬁcation would

be needed at the decoder. The performance of the receiver

in terms of FER (and potentially in terms of throughput, as

higher ModCods could be adopted) can be further improved

through an iterative detection-decoding scheme, which would,

on the other hand, increase the receiver complexity and require

further modiﬁcations to the existing structure of the receiver

described in [10]. Such possibility is not further discussed here

for a matter of space.

IV. NUMERICAL RESULTS

In the following we evaluate the performance of the pro-

posed algorithm for the scenario described in Section II.

We start by describing in detail the reference scenario and

the interference distribution generated through the system level

simulator presented in SectionII. The beam numbering and

geographical location are those shown in Fig. 1. In Table II,

at the bottom of the page, we show the C/I related to each

of the 10 strongest interferers for both beam 105 and 110 in

two points, namely at the center of the beam (CoB) and at the

edge of the beam (EoB).

Fig. 2. Interference pattern and conventional numbering of the co-channel

beams. The central red rectangle represents the reference beam, while the

yellow rectangles represent the ten strongest interfering beams.

The relative numbering of the interferers is given according

to Figure 2, where the central rectangle represents the refer-

ence beam while the yellow rectangles represent the strongest

co-channel interferers.

With reference to Table II at the bottom of next page, it

can be seen that in the EoB cases the power of the interferer

number 5is comparable to that of the reference signal while

the second strongest interferer is attenuated more than 11

dB. On the other hand, in the CoB the strongest interferer

is at least 12 dB lower than the reference signal. Let us

consider the worst case scenario, i.e., the EoB. In this case

there is only one strong interferer plus nine interferers with

a relatively weak power, that, by the Central Limit Theorem,

can be modeled as Gaussian noise. Trying to apply MUD to

these low-power interferers is not likely to have a relevant

impact on the system performance while it would increase

signiﬁcantly the complexity of the receiver. A better choice

is to apply the MUD to the desired signal and the strongest

interferer while treating the rest of the interferers as noise. In

order to understand whether the assumption of having at most

one signiﬁcant interferer is realistic in each point of the beam

footprint, we analyzed the distribution of the total C/I across

the whole beam. The distribution is shown in Fig. 3, where

three cases have been considered for each point in the two

beams: i) all the interferers are present (top-left), ii) only the

ﬁrst strongest interferer has been removed (bottom-left) iii)

the ﬁrst two strongest interferers have been removed (bottom-

right). From the ﬁgure it can be seen that the total C/I reaches

negative values, in logarithmic scale, in some areas of the

beam (e.g., in the EoB points considered in the table shown

Fig. 3. Probability density function (PDF) of interference across the

covered area (horizontal axis represents C/I , expressed in dB) in case: 1) all

interferers are present (top-left), 2) the strongest interferer has been removed

(bottom-left), 3) the two strongest interferers have been removed (bottom-

right). The cumulative distribution function (CDF) of the difference between

the two strongest interferers across the beam is also shown (top-right).

in table II) when all the interferers are present. Removing the

strongest interferer determines a minimum C/I larger than

or equal to 6dB in any point of the two considered beams.

We further notice that the cancellation of the second strongest

interferer further increases the minimum C/I of only about

1-1.5dB. From the analysis of Fig. 3 we conclude that the

total C/I is mainly limited by the ﬁrst strongest interferer while

the second one has only limited impact on performance. We

propose therefore to deal with only one interfering signal while

treating the others as noise in order to keep the complexity

low. The detector described in Section III. We recall that this

implies that the only modiﬁcation needed at the decoder side

is in the detector, for which the a-priori probability in case of

one interferer reduces to:

Pk,sn=

M−1

X

m=0

exp |yk−hgC(tk)sn−hgI(tk)sI

m|2

2σ2

eq ,(14)

Mbeing the cardinality of the interferer’s constellation. The

correspondent block scheme is shown in Fig. 4.

In order to take into account the inﬂuence of the other

interferers (which reduces the reliability of the detection) in

the received signal’s statistics we increase σ2

eq by several

dBs (6in the following simulations) with respect to the

actual variance of the thermal noise σ2. The optimal choice

would be to choose the value of σeq by estimating the noise-

plus-interference power. However, in practice keeping a ﬁxed

value of the variance can be a good compromise since i) the

thermal noise component can be either given by the terminal

manufacturer or easily estimated, while the power due to

residual interference may not be easy to measure, as the

Fig. 4. Proposed SIC scheme. Only the strongest interferer is taken into

account and no iterative detection is applied (i.e., reference signal is detected

as described in Section III and decoded using the BGAN turbo decoder).

Optionally also the strongest interfering signal can be decoded.

received signal is made up by the sum of the (strong) reference

signal, a (possibly strong) dominant interferer and the residual

interferer (estimation of the residual interference power in

such conditions would increase the complexity of the receiver)

and ii) we observed that the FER shows little sensibility to

the exact value of σeq . In the simulations presented in the

following the signal model described in equations 1-5 has

been adopted: all 10 interferers have been simulated including

channel code, modulation and channel effect, and scaling the

powers according to Table II. We ﬁrst present the results

obtained in AWGN channel and then those for the LMS

scenario. The simulation setup for the two cases is depicted

in Fig. 5. The simpliﬁed scheme shown in Fig. 4 (i.e., the

received signal passes through the detector and through the

turbo decoder just once) has been used.

A. AWGN Channel

In ﬁgures 6, 7 and 8 we show the FER curves for the

considered MUD using the interference pattern detailed in

Table II. Different combinations of MODECODs available in

the standard [13] have been used, namely QPSK rate 1/3for

all signals in Fig. 6, QPSK with rate 2/5 for all signals in Fig.

7 and QPSK with rate 1/3for the reference signal and 16

QAM rate 1/3for interferers in Fig. 8.

From the plots it emerges that the target FER of 10−3can

be achieved using QPSK modulation in all signals up to rate

2/5while if 16 QAM is used in one of (or both) the signals the

target FER cannot be achieved for values of C/N of practical

interest.

In the following subsection we present the results for an

LMS scenario.

Fig. 5. Simulation setup in AWGN and LMS channels.

4 5 6 7 8 9 10 11 12 13 14 15

10−3

10−2

10−1

100

C/N (dB)

FER

Reference signal: bearer F80T1Q4B−L8, rate=0.3375, QPSK

Interferer: bearer F80T1Q4B−L8, rate=0.3375, QPSK

Fig. 6. FER in AWGN with MUD. Bearer F80T1Q4B-L8 (QPSK rate

1/3, symbol rate 33600 sym/sec, roll-off 0.25) is used for all signals. The

interference pattern for beam 110 EoB detailed in Table II (worst case

scenario) has been used.

B. LMS Channel

The channel model used in the simulations presented in the

following is a land-mobile satellite (LMS) channel for vehicles

moving at a speed of 50 kmph in a suburban environment. A

channel realization of 30 minutes (25 km path at 50 kmph)

has been used, corresponding to about 2.7·104FEC blocks for

bearer F80T025Q1B-L8 (QPSK, rate 1/3, symbol rate 8400

symbols per second). The time series has been generated using

an LMS channel generator implementing the Perez-Fontan

model [19].

TABLE II

TABLE W ITH FOU R SAMPLES OF THE INTERF EREN CE PATTERN. EACH ROW CONTAINS THE C/I RELATED TO TH E TEN STRONGEST INTERFERING

SIGNALS FOR EITHER A CENTER-O F-BE AM (BEST LO CATIO N)POINT OR AN EDGE-OF-BE AM (WORST LOCATION)POI NT IN BEAMS 105 (BEST BEAM)AND

110 (WORST BEAM). THE TOTAL C/I IS ALSO REPORTED F OR EACH CASE [18].

C/I [dB] C/I Total [dB]

Beam 1 2 3 4 5 6 7 8 9 10

105 CoB 37.1107 21.5885 32.1618 37.2214 17.9294 14.5272 27.3961 31.9697 20.7406 29.2564 11.46467

EoB 15.6046 15.5337 29.8048 15.8007 0.3881 15.1211 21.4581 44.3936 38.1297 21.875 -0.17835

110 CoB 30.3903 19.541 32.9636 35.0503 13.7636 12.1374 21.4154 29.4771 18.9879 30.8531 8.607253

EoB 27.2207 29.9124 22.4402 17.9726 0.1185 11.5821 18.8873 14.2254 15.2343 27.9627 -0.6047

4 6 8 10 12 14 16

10−3

10−2

10−1

100

C/N (dB)

FER

Reference signal: bearer F80T1Q4B−L7, rate=0.4, QPSK

Interferer: bearer F80T1Q4B−L7, rate=0.4, QPSK

Fig. 7. FER in AWGN with MUD. Bearer F80T1Q4B-L7 (QPSK rate

2/5, symbol rate 33600 sym/sec, roll-off 0.25) is used for all signals. The

interference pattern for beam 110 EoB detailed in Table II (worst case

scenario) has been used.

2 4 6 8 10 12 14 16

10−3

10−2

10−1

100

C/N (dB)

FER

Reference signal: bearer F80T1Q4B−L8, rate=0.3375, QPSK

Interferer: bearer F80T1X4B−L3, rate=0.33437, 16 QAM

Fig. 8. FER in AWGN with MUD. Bearer F80T1Q4B-L8 (QPSK rate 1/3,

symbol rate 33600 sym/sec, roll-off 0.25) is used for the reference signal

while bearer F80T1X4B-L3 (16 QAM rate 1/3, symbol rate 33600 symbols

per second , roll-off 0.25) is used for the interferers. Note that rate 1/3is

the lowest code rate available in BGAN. The interference pattern for beam

110 EoB detailed in Table II (worst case scenario) has been used.

In Fig. 9 we show the frame error rate for the reference

signal using the proposed simpliﬁed SIC scheme. 10 interferers

have been considered using the C/I values in Table II. Bearer

F80T025Q1B-L8 (QPSK, rate 1/3) of standard [13] has been

adopted for all signals.

Fig. 9 shows that the SIC scheme reaches the target FER of

10−3in all the considered cases, showing a neat enhancement

with respect to the case in which no interference cancellation

is applied. Thus it can be seen that decoding is possible in

all considered points, while it is not feasible without the

MUD algorithm. A relatively high C/N is required in order

to fulﬁll FER requirements in EoB which is due partly to

the challenging propagation scenario. As a matter of facts

it can be seen in Fig. 9 that, even in case no interference

is present in the system, a C/N of about 14 dB is needed

to reach a target FER of 10−3. We also note that the FER

obtained in the LMS channel in case of no interference is

almost the same as that in CoB. This is because the total

0 5 10 15 20 25 30

10−3

10−2

10−1

100

C/N (dB)

FER

Beam 110 EoC no IC

Beam 110 EoC

Beam 105 EoC

Beam 110 CoC no IC

Beam 110 CoC

Beam 105 CoC

No interference

Fig. 9. Frame error rate for the reference signal using the simpliﬁed SIC

scheme with one iteration (one detection and one decoding iteration). A 30

minutes LMS channel series in suburban environment generated according

to [20] has been used. 10 interferers have been considered using the C/I

values in Table II. Bearer F80T025Q1B-L8 (QPSK, channel code rate 1/3,

symbol rate 8400 symbols per second) of standard [10] has been adopted for

all signals.

interference level in CoB is low enough to allow for correct

decoding even without SIC, which justiﬁes the fact that the

same performance is achieved by the SIC and the no IC (no

interference cancellation) schemes.

An important outcome of the simulations is that the system

results to be interference-limited mainly in the EoB area,

while interference has little effect in the CoB area. We also

showed that dealing with a single interferer is enough to make

decoding possible. We emphasize that these results have been

achieved with a limited increase in the receiver complexity,

as only the demapper has been modiﬁed with respect to

the receiver described in [10]. The fact that a C/N larger

than 20 dB is needed in EoB could be addressed by using

a code with longer codewords (the turbo code of DVB-SH,

for instance, has codewords which are an order of magnitude

larger than those used in the simulations just presented), an

interleaver with an adequate depth or a combination if the

two, compatibly with memory and latency constraints in the

user terminals.

V. CONCLUSIONS

We studied the application of co-channel soft interference

cancellation in multibeam mobile satellite systems with a two-

colour frequency reuse scheme. We took the ETSI standard

[13], currently used in commercial satellite systems, as a ref-

erence and simulated the beam radiation and the interference

patterns using a realistic antenna model. The calculation of the

interference pattern has been carried using simulator developed

by ESA. Due to strong complexity limitations in mobile

terminals, we proposed to move part of the complexity to

the system level, by aligning signals transmitted over different

beams and adding speciﬁc signalling information in the global

beam. We started from the analysis of the interference levels

across the beams selecting two of them as best and worst case

scenarios. In order to keep the complexity at the receiver low,

we proposed a simpliﬁed scheme in which only the detector

is modiﬁed with respect to the standard [10]. Our results

showed that even under challenging propagation conditions

and with strong interference, the considered scheme leads to

interesting results, achieving a target FER of practical interest.

We showed that the proposed approach may constitute a

concrete possibility to live together high levels of interference

with a relatively limited increase in complexity at both system

and user level. This comes at the expense of a limitation in

the ModCods that can be used.

ACKNOWLEDGEMENTS

The present work has been carried out under the ARTES

1programme founded by the European Space Agency. The

view expressed herein can in no way be taken to reﬂect the

ofﬁcial opinion of the European Space Agency.

The research leading to these results has received funding

from the Spanish Ministry of Science and Innovation under

projects TEC2011-29006-C03-02 (GRE3N-LINK-MAC) and

the Catalan Government (2009SGR0891).

REFERENCES

[1] J. G. Andrews, “Interference cancellation for cellular systems: a con-

temporary overview,” IEEE Wireless Comm., vol. 12, no. 2, pp. 19–29,

2005.

[2] X. Wang and H. V. Poor, “Iterative (turbo) soft interference cancellation

and decoding for coded CDMA,” IEEE Transactions on Comm., vol. 47,

no. 7, pp. 1046–1061, 1999.

[3] M. Kopbayashi, J. Boutros, and G. Caire, “Successive interference

cancellation with SISO decoding and EM channel estimation,” IEEE

Journal on Selected Areas in Comm., vol. 19, no. 8, pp. 1450–1460,

2001.

[4] G. Colavolpe, D. Fertonani, and A. Piemontese, “SISO detection over

linear channels with linear complexity in the number of interferers,”

IEEE Journal of Selected Topics in Signal Processing, vol. 5, no. 8, pp.

1475–1485, 2011.

[5] C. Schlegel, “Achieving the multiple-access capacity of the AWGN

channel with iterative processing,” in AESS European Conference on

Satellite Telecommunications (ESTEL), Rome, Italy, Oct. 2012.

[6] B. F. Beidas, H. El-Gamal, and S. Kay, “Iterative interference can-

cellation for high spectral efﬁciency satellite communications,” IEEE

Transactions on Comm., vol. 50, no. 1, pp. 31–36, 2002.

[7] M. Moher, W. Zhang, P. Febvre, and J. Rivera-Castro, in Advanced

Satellite Multimedia Systems conf. (ASMS) and Signal Processing for

Space Communications workshop (SPSC).

[8] European Space Agency (ESA), “Advanced Research in Telecommuni-

cations Systems (ARTES),” http://telecom.esa.int/telecom/.

[9] European Space Agency, “Next generation waveform for improved

spectral efﬁciency,” http://telecom.esa.int/telecom/, 2013.

[10] European Telecommunications Standards Institute, “Satellite component

of UMTS (S-UMTS) family SL satellite radio interface; part 2: Physical

layer speciﬁcations; sub-part 1: Physical layer interface,” May 2012.

[11] Inmarsat, “Broadband Global Area Network (BGAN),”

http://www.inmarsat.com/services/bgan.

[12] A. Barbieri, D. Fertonani, and G. Colavolpe, “Time-frequency pack-

ing for linear modulations: spectral efﬁciency and practical detection

schemes,” IEEE Transactions on Communications, vol. 57, no. 10, pp.

2951–2959, 2009.

[13] European Telecommunications Standards Institute, “Satellite component

of UMTS (S-UMTS) family SL satellite radio interface; part 1: General

speciﬁcations; sub-part 3: Satellite radio interface overview,” May 2011.

[14] International Telecommunications Union - Radiocommunications Sector,

“Propagation data and prediction methods requred for the design of

Earth-space telecommunication systems: methods requred for the design

of Earth-space telecommunication systems,” Reccomendation ITU-R

P.618-10 (10/2009), Oct. 2009.

[15] Satsoft, http://www.satsoft.com/.

[16] M. Feder and E. Weinstein, “Parameter estimation of superimposed

signals using the EM algorithm,” IEEE Trans. on Acoustics, Speech

and Signal Processing, vol. 36, no. 4, pp. 477–489, Apr. 1988.

[17] G. Cocco, N. Alagha, C. Ibars, and S. Cioni, “Practical issues in multi-

user physical layer network coding,” in IEEE Advanced Satellite Mobile

Systems conf. (ASMS), Baiona, Spain, Sep. 2012.

[18] European Space Agency, “NGWISE project technical Annex B: Tech-

nologies, models and requirements for the work ofRG2 on MSS,” 2013.

[19] F. Perez-Fontan, M. A. Vazquez-Castro, S. Buonomo, J. P. Poiares-

Baptista, and B. Arbesser-Rastburg, “S-band LMS propagation channel

behaviour for different environments, degrees of shadowing and eleva-

tion angles,” IEEE Trans. on Broadcasting, vol. 44, no. 1, pp. 40–76,

Mar. 1998.

[20] F. Perez-Fontan, A. Mayo, D. Marote, R. Prieto-Cerdeira, P. Marino,

F. Machado, and N. Riera, “Review of generative models for the

narrowband land mobile satellite propagation channel,” Int’l Journal of

Satellite Comm. and Networking, vol. 26, pp. 291–316, 2008.