Content uploaded by Stefano Cioni
Author content
All content in this area was uploaded by Stefano Cioni on Oct 31, 2017
Content may be subject to copyright.
From S-MIM to F-SIM: Making Satellite
Interactivity Affordable at Ku and Ka-band
A. Arcidiacono, D. Finocchiaro, F. Collard, Eutelsat, France.
S. Scalise, F. Lazaro Blasco, DLR, Germany
R. De Gaudenzi1, S. Cioni, N. Alagha ESA/ESTEC, The Netherlands
M. Andrenacci, MBI, Italy
Abstract
This paper describes the evolution of the ETSI S-band Mobile Interactive Multimedia (S-
MIM) protocol to support Fixed Interactive Multimedia Services (F-SIM) exploiting
existing Ku and Ka-band satellites. The key F-SIM protocol differences for both physical
and upper layers are described and justified. The F-SIM protocol has been adopted by
the recently deployed Eutelsat Broadcast Interactive System (EBIS) whose architecture,
key system parameters, link budget examples and key composing elements are also
described. Finally, a summary of laboratory and field trials results over the Eutelsat Ka-
Sat multi-beam satellite are illustrated.
1. Introduction and Problem Outline
The demand for low cost satellite interactivity has been constantly growing in recent years.
New applications are emerging in two major domains: firstly, television (TV) broadcasters
are attracted by the added value of connected TV services like audience measurement,
enhanced customer support, Push Video on Demand (VOD) or personal subscription
management; secondly, we are now in the era of Internet of Things (IoT) the intelligent
connected objects, areas in which very small data are collected from or exchanged with
remote devices. However, there is a strong competition from terrestrial networks (cabled
or wireless) that expanded their reach, offering lower prices than classical satellite
solutions. Consequently, satellite providers must offer the cost-effectiveness required by
most service operators, both in terms of equipment costs and optimisation of the satellite
capacity.
One part of the solution came from the adoption of multi-beam satellite architectures, also
known as High Throughput Satellite Systems (HTS), such as Eutelsat Ka-Sat [1]. Allowing
high levels of frequency re-use and increasing the satellite antenna gain thanks to small
1 Correspondence address: European Space Agency, European Space and Technology Centre, ESA-ESTEC,
Keplerlaan 1, Postbus 299, 2200 AG Noordwijk, The Netherlands, e-mail: rdegaude@xrsun0.estec.esa.nl.
1
spot beams, the overall bandwidth cost can be reduced by a factor of ~10 with respect to
traditional Ku-band wide-beam satellites.
Another important element was the design of low cost terminals, and protocols adapted
for low-bitrate applications. This work was started in the framework of the deployment of
Mobile Satellite Services (MSS) in S-band over Europe, with the specification and
standardization of the ETSI S-band Mobile Interactive Multimedia (S-MIM) protocol [2]-
[3], and the development of corresponding prototype terminals. However, this protocol
had to be extended to match the specific requirements of applications operating at higher
frequency bands (mainly Ku and Ka but also C-band). This work gave birth to the Fixed
Satellite Interactive Multimedia (F-SIM) protocol specifications that are illustrated in this
paper. The main S-MIM evolutions required in F-SIM can be summarized as follows:
a) Support of higher frequency bands (C, Ku and Ka);
b) Adoption of DVB-S2 as forward link standard;
c) Adoption of a more stringent phase noise mask for the terminal;
d) Exploitation of the DVB-S2 Network Clock Reference for the provision of an
accurate frequency reference at the terminal;
e) Major extension of the physical layer configurations supported in terms of bit rate
and packet size supported;
f) Modifications of the uplink power control algorithm to best suit the Ku/Ka-band
propagation channel;
g) Enhancement of the upper layer design to be more IP friendly by enabling a more
flexible management of the different Quality of Service (QoS) classes at data link
layer ;
The F-SIM protocol represents the core of the Eutelsat Broadcast Interactive System
(EBIS), as implemented by the “Smart LNB” terminals [4], as well as by the “Starfish”
head-end architecture that is also presented in this paper. The EBIS system is already
deployed in Ka-band on Eutelsat Ka-Sat satellite, as well as in Ku-band on several Eutelsat
fleet satellites. The EBIS advantages reside in the fact of proposing a global architecture
for satellite interactivity with reduced costs at the terminal level, while keeping flexibility
and high spectral efficiency at the system level.
The paper is organized as follows: Sect. 2 describes the high level architecture of the EBIS
system. Sect. 3 illustrates the key system elements (services support, frequency bands,
space and ground segment). Sect. 4 focuses on the S-MIM evolution towards F-SIM. Sect.
5 describes the system field trial results. Sect. 6 provides the conclusions.
2. System Architecture
The EBIS is typically based on a geostationary transparent satellite with Ku-band (global
beam coverage) in the forward link, whereas the return link may use Ku-band or Ka-band
(or C-band where regulations allow it). The system is designed to operate with existing or
planned transparent bent-pipe telecom satellites with global or multi spot beam coverage
(typical for Ka-band systems). See Sect. 3.2 for details.
2
A high level representation of the EBIS concept is shown in Figure 1. A hub is connected
in star topology with many user terminals. The forward and return links are separated,
and could be in different bands or even on different satellites. The forward link uses DVB-
S2 protocol, while the return link protocol is specified within F-SIM as an evolution of the
ETSI S-MIM protocol as summarized in Table 1.
ETSI S-MIM
F-SIM
Remarks
Frequency band
S-Band
C/Ku/Ka bands
Potential coexistence
with other return link
services enforces a
more stringent control
on interference and
generally on terminal
behaviour by the hub
Forward Link
DVB-SH
DVB-S2
F-SIM System
signalling shall be
carried over IP in
order to ensure
compatibility with
legacy MPE/MPEG
multiplexers
Footprint
Typically symmetric
between forward and
return links
Might be asymmetric
in both footprint and
band
Need for an enhanced
logon procedure to
identify the proper
return link beam
Forward Link
Signalling
Optimised to reduce
bandwidth
consumption
Optimised to reduce
terminal storage and
processing capabilities
Multicast and unicast
signalling added on
top of broadcast
signalling to better
control individual
terminals or group of
terminals
Physical Layer
Configurations
Typically, only one
configuration used in
each return link
carrier with bit rate
limited to 5 kbps and
bandwidth from 300
kHz up to 5 MHz
Larger bit rate (up to
160 kbps) supported
with carriers from 2.5
MHz up to 10 MHz
bandwidth
Increased number of
configurations
supported and
different
configurations might
be used by different
terminals sharing the
same return link
carrier
Main source of
Fading
Mobility
Atmosphere (e.g. rain)
Power control
algorithm was adapted
to account for the
different fading
dynamic. Changes
also required to take
into account
coexistence of
different physical
configurations in the
same return link
carrier
Return Link Traffic
Mainly UDP based,
with very small
payload size and low
duty cycle
Still sporadic activity,
but depending on
terminal classes TCP
can also be used,
together with larger
payloads
Need for enhanced
QoS support at data
link layer
3
Table 1: Summary of F-SIM features compared to S-MIM.
The hub provides all modulation and demodulation functions, and is connected via an
Internet Protocol (IP) network to the Network Operating Center (NOC) and to the service
providers.
Figure 1 - High level EBIS architecture
On the user side (see Figure 2), a typical terminal implementing F-SIM is the Eutelsat
Smart LNB that provides Internet Protocol (IP) connectivity to the user’s network. The
user is able to get access to interactive multi-media systems exploiting the F-SIM protocols
and a small dish. All RF and baseband processing functions are integrated in the Outdoor
Unit (ODU), which communicates with an Indoor Unit (IDU) via a normal coaxial cable.
The IDU could also be integrated in a set-top box.
4
Figure 2: Terminal side of the EBIS system.
Other scenarios are possible as well: in a typical M2M installation, for example, the local
network could be attached directly to the ODU.
3. Key System Elements
3.1 Services supported
F-SIM was designed primarily to offer IP connectivity for a variety of services, especially
those with medium/low bit rate and low duty-cycle requirements.
The native support for transporting IP packets allows a very easy integration of F-SIM
based terminals into a generic IP network, without requiring costly customizations. In this
way, a wide range of applications is covered.
The main services targeted by the F-SIM protocol and Smart LNB terminals are:
• “Connected TV” services: providing interactivity related to legacy TV content,
including multiscreen video distribution, DRM services, voting, audience
measurement, web browsing, datacast.
• Machine-to-Machine (M2M) connectivity: providing low-bitrate connectivity to
objects or networks. This includes Internet of Things (IoT) services, Supervisory
Control and Data Acquisition (SCADA), backhauling of M2M wireless networks.
5
3.2 Frequency bands and planning
The F-SIM protocol can support all the bands currently adopted by commercial
geostationary satellites, as well as a mixed architecture where forward and return channels
use different bands. The first generation of terminals under development are supporting
the following configurations:
Ku/Ku: the forward and return links use Ku-band (typically single beam);
Ku/Ka: the forward link is in Ku-band whereas the return link is in Ka-band. The
data link layer has been designed in order to support multiple spot beams system.
The use of two different satellites can also be considered as long as the satellites are
located close to each other;
Ka/Ka: both forward and return links use Ka-band (typically multi-spot satellite
architecture).
In the forward link, the frequency bands of interest are:
Ku-band: 10.70-12.75 GHz;
Ka-band: 17.7-20.2 GHz.
In the return link, the target frequency bands are:
Ku-band: 13.75-14.5 GHz;
Ka-band: 29.5-30 GHz.
In all cases, the return link supports three possible channel bandwidths of 2.5 MHz, 5 MHz
or 10 MHz. Each channel is shared among a very large number of terminals. The F-SIM
Enhanced Spread-Spectrum (E-SSA) Random Access (RA) allows reusing the same
spreading sequence and bandwidth among many terminals in a truly asynchronous and
efficient manner [10]. The Return Link System is composed of one or more channels (also
called “RL carriers”) distributed over several beams. The number of required channels is
related to the number of terminals supported and their traffic statistics (e.g. bit rate,
packet size, activity factor).
The Forward Link (FL) is composed of one FL carrier for each satellite beam. Each FL
carrier contains:
Appropriate DVB Program Specific Information/Service Information (PSI/SI)
[7] tables (legacy DVB tables with specific content): Network Information Table
(NIT), Program Map Table (PMT), Program Association Table (PAT), IP
Notification Table (INT), etc.;
DVB Network Clock Reference (NCR) information in a Program Clock
Reference (PCR) Packet Identifier (PID) (27 MHz counter);
System-wide signalling (transported via multicast IP packets in either bitwise
or Protobuf format, according to F-SIM Signalling specifications [9]);
Multicast signalling (according to F-SIM Signalling specs [9]) for a group of
terminals logged on this FL carrier (transported via multicast IP packets);
6
Unicast signalling (according to F-SIM Signalling specs [9]) for individual
terminals logged on this FL carrier (transported via unicast IP packets);
Unicast data for terminals logged on this FL carrier (transported via unicast IP
packets);
Additional IP Multicast data (e.g. encoded video, push files) to be routed by the
terminal, for broadcast-type services.
Further details about the system signalling are provided in Sect. 4.2.
3.3 Space segment requirements
In this section we present some link budgets examples. Since the forward link is based on
the well-known DVB-S2 standard, we will focus the discussion on the F-SIM return link
budget.
For Ku-band a satellite antenna G/T of 4 dB/K has been assumed. The Ka-band is divided
in two sub-cases corresponding to two different satellite antenna patterns: a) small spot
beams (Ka-Sat like) with a G/T of 18 dB/K; b) larger spot beams with a G/T of 12 dB/K.
Table 2 presents an example of F-SIM deployment using channels of 5 MHz and an 80
kbps physical layer configuration (see Sect. 4.1 for physical layer assumptions and details).
The Ka-band terminal RF power is assumed to be limited to 100 mW while 500 mW is
baselined for Ku-band operations. These link budgets are computed for a terminal
transmitting with the maximum power, simultaneously with 50 other ones that use
nominal power spreading (see Sect. 4.2). The signal from this terminal will suffer from the
interference generated by the remaining terminals, but since it transmits at maximum
power, the gateway shall be able to demodulate and cancel it, thus reducing the multiple
access interference (MAI) for the remaining signals.
We assumed an optimal case where the received power distribution of terminal packets is
uniformly spread (in dB) over the maximum range, according to the available link margin
[12]. In order to consider the MAI effect in a general case, a simple link budget is not
sufficient because the presence of the iterative Successive Interference Cancellation (iSIC)
process at the demodulator side [10]. The impact of MAI cannot be modelled by simple
formulas. It is necessary to run an ad-hoc semi-analytical simulator to assess this impact.
The implementation of a semi-analytic process simulating the SIC process allows the
computation of the maximum number of simultaneously transmitting terminals and the
maximum aggregate bit rate as described in detail in [12]. The selected E-SSA RA scheme
performance depends on several crucial parameters:
• The FEC characteristic (Frame Error Rate vs. Eb/N0) relative to the packet size;
• The type of random spreading power distribution for incoming gateway packets
obtained considering the power control implemented in the terminals;
• The spreading factor adopted;
• The residual power after each packet cancellation, generating additional
interference during the iSIC process at the gateway demodulator;
• The number of SIC iterations supported by the gateway demodulator, depending
on hardware configuration.
7
According to the values of these parameters, the aggregate spectral efficiency can vary
from 1 bit/s/Hz to 3 bits/s/Hz (see Sect. 5). This corresponds to an aggregate bit rate
between 5 Mb/s and 15 Mb/s over a 5 MHz channel. As explained in [12], for the E-SSA
iSIC demodulator, the larger the maximum incoming packet power spread (thus link
margin), the larger the achievable aggregate throughput. It is noted that the E-SSA open
loop Random Access (RA) scheme throughput for packet loss rate below 10-3 as typically
required by satellite networks, is three orders of magnitude better than conventional
ALOHA or Slotted ALOHA (see Chapter 3 of [11]). Also in terms of latency E-SSA is by far
superior to other protocols as it allows packets to get through in pure RA mode without
the need for packet retransmission.
Table 2 shows three return link budget examples encompassing both Ku and Ka-band
typical satellites.
Return link budget
Unit
Ku-band
Ka-band multi-spots
large spot
beams
small spot
beams
System characteristics
Chip rate (Rc)
Mcps
3.84
3.84
Bit rate
kbps
80.00
80.00
Coding rate
-
0.33
0.33
Symbol rate
kbaud
240.00
240.00
Spreading Factor (SF)
-
16.00
16.00
Roll-off factor
-
0.22
0.22
Gross carrier bandwidth
MHz
4.68
4.68
Downlink frequency
GHz
11.66
19.00
Downlink wavelength
m
0.03
0.02
Uplink frequency
GHz
13.96
30.00
Uplink wavelength
m
0.02
0.01
Terminal Tx characteristics
Max Tx power
dBm
27.00
20.00
Antenna diameter
m
0.80
0.80
Antenna gain
dB
36.00
44.00
EIRP
dBW
33.00
34.00
Satellite propagation
Link attenuation
dB
205.50
213.30
213.30
G/T
dB/K
4.00
12.00
18.00
Result for terminal at max power
Overall link C/N (normalized to Rc)
dB
-11.40
-7.20
-1.20
Eb/N0 without MAI (at max Tx power)
dB
5.41
9.61
15.61
Aggregate return link capacity (estimate)
Max number of simultaneous transmissions
-
60
100
130
Aggregate capacity
Mbps
5.0
8.0
10.5
Table 2: Example of F-SIM link budget.
8
3.4 Terminal design
The terminal unit is typically composed of:
- One satellite consumer grade dish capable of receiving Ku/Ka-band Direct-to-
Home (DTH) satellite signals and sending F-SIM packets in either in Ku or Ka-
band.
- One ODU, acting as electronic interactive feed that can be mounted into a
consumer grade satellite dish. The ODU contains all RF, baseband, link layer and
IP routing functions.
- One IDU capable of interacting with the ODU via a coaxial cable, to provide
electrical power to the ODU and interfacing with existing multimedia devices or
home appliances via Ethernet.
For some applications, such as M2M, the ODU could be directly connected to a sensor or
a local IP gateway. The terminals consist at this stage of a standard satellite dish of 70-80
cm diameter, like those used for satellite television DTH reception, equipped with a more
complex feed that embeds all the functionalities related to the reception and transmission
of IP packets. The terminal’s low power/energy requirements allow its deployment in a
stand-alone fashion, powered by solar panels, and therefore totally independent of the
terrestrial infrastructure.
Summarizing, the typical F-SIM terminals, such as Eutelsat Smart LNB, have the
following features:
• Low cost: bill of materials of tens of dollars, with target price under $100;
• Low power consumption;
• Easy installation with 70-80 cm dish;
• Support for different Rx/Tx bands, currently Ku/Ku, Ku/Ka and Ka/Ka;
• Downlink bitrates up to 50 Mbps for receiving broadcast and unicast IP;
• Support of legacy DVB-S/S2 receivers (pass-through of legacy IF signal), reusing
existing coaxial cable in old installations;
• Uplink bit rates from 10 to 160 kbps (at physical level) in F-SIM IP routing
capabilities.
The terminal is intended to be installed at customer premises and left unattended, with all
the configuration and management being done via satellite from the satellite service
operator. Figure 3 shows the overall functional block diagram of the Smart LNB terminal.
The complete terminal is composed by an ODU and an IDU. The ODU is connected to the
IDU by means of a coaxial cable carrying the (legacy) Intermediate Frequency (IF)
downlink signal, the IP packets encoded with Ethernet-over-coax technology, and the
power for the ODU. The power supply is contained in the IDU. The IDU interfaces the end
user terminal and the IP home network. The IDU could also be integrated into a Set-Top
Box (STB), or connected to it as a USB key, so as to provide connectivity to the STB itself.
9
Figure 3: Overall functional block diagram of a SMART LNB terminal
The ODU functional block diagram is presented in Figure 4. On the receive side it contains
a DTH type Low Noise Block converter (LNB) which is used to provide the IF signal for
the embedded DVB-S2 demodulator, as well as for the legacy STB (via the IDU). The ODU
contains a DVB-S2 demodulator providing the forward link F-SIM signalling information
to the terminal, as well as the stable frequency reference for the modulator thanks to the
NCR extraction. The F-SIM modulator gets the data from the IDU through the coaxial
cable and drives the up-converter and power amplifier.
10
Figure 5 - Smart LNB prototypes in Ku-band (left) and Ka/Ku-band (right) used for tests and
validation.
Figure 6 - Ku-band Smart LNB terminal.
3.5 Gateway design
MBI, in collaboration with Eutelsat and the European Space Agency, has developed the
first complete F-SIM gateway platform, named “Starfish”, which is fully compliant with
the F-SIM protocol. The Starfish platform illustrated in Figure 7 is a turn-key, cost-
effective, easily scalable and compact solution for the F-SIM system. It is based on a
Software Defined Radio (SDR) architecture that allows easy algorithm upgrades and
leveraging of advances in hardware performance.
The main platform capabilities are:
• Easy integration with existing architectures (Radio Frequency Systems (RFS),
DVB-S2 head-end, etc.) on both forward and return links;
12
• Support for both single and multiple-beam satellite configurations;
• High scalability, compact and adaptive characteristics, integration with existing
management facilities;
• A number of geographically distributed gateways can be included (in the case of
multiple gateway systems such as Ka-Sat);
• Large number of users supported.
The overall functional architecture of the F-SIM Starfish platform for Ku (forward)/Ka
(return, multi spots) configuration is shown in Figure 7.
Figure 7: Starfish gateway architecture for Ku/Ka configuration.
In the following paragraphs the main design drivers are illustrated.
Geographically Distributed Architecture
In the case of multi-beam satellites (e.g. HTS such as Eutelsat Ka-Sat), a number of
Starfish gateways have to be geographically distributed in order to accommodate the
multi-beam multi-hub satellite architecture. The Starfish gateways are connected to the
central starfish Hub using an IP-based backbone.
13
SDR based RAN
The Starfish gateway return link demodulator shown in Figure 8 provides the Radio
Access Network (RAN) functionalities for the return link. The return link gateway is based
on a Level 2 Software Defined Radio (SDR) architecture2. A COTS radio front-end acquires
the analogue signal and converts it to digitized samples, while the Digital Signal
Processing (DSP), which is from the physical to the IP layers, is implemented by software
running on general-purpose reprogrammable units.
Figure 8: SDR based RAN hub.
The processing units are multi-core Intel servers equipped with a number of Graphical
Processor Units (GPUs) supporting the Compute Unified Device Architecture (CUDA)
computing platform. This setup allows the software to reach a high performance level in
terms of the number of elementary operations per unit of time, with an excellent trade-off
between cost and power consumption. The software has been entirely developed in a high
level language (C++), including the physical layer components, which greatly simplifies
reuse, testing and validation, while supporting a seamless portability to different hardware
platforms.
On one hand, the return link capacity can be easily scaled-up by increasing the
computational power available at the gateway by adding new servers. In this case, a single
RF front-end is required, and the baseband digital signal is distributed between the servers
through high-capacity InfiniBand links. On the other hand, an entry-level solution, for low
traffic scenarios, is very light and compact, in addition to cost effective.
Finally, this architecture ensures a simplified scalability in terms of signal bandwidth
especially when Ka-band is used on the return link. Current state of the art SDR devices
used by the Starfish implementation are capable of processing signals with bandwidths
higher than 10 MHz.
The random-access protocol used by F-SIM is based on the innovative concept of sliding
window detection, demodulation and iterative cancellation of the received traffic to
maximize the RA throughput even in the presence of packet power unbalance [10]. This
2 According to the SDR Forum (now Wireless Innovation Forum) definition –
http://www.wirelessinnovation.org/.
14
type of iSIC requires significant computational burden at the gateway. The following tasks
must be independently performed for every received packet:
• Preamble detection;
• Channel estimation;
• Demodulation and decoding of the control and data channels;
• Regeneration of a baseband replica of the received packet;
• Cancellation from the sliding window memory of the decoded packet using the
regenerated replica.
Depending on the channelization and traffic load, there may be thousands of physical layer
packets received at the same time. In the Starfish architecture this challenge is solved by
using a parallel implementation in each of two tiers. Firstly, the overall DSP operations
are split between the Central Processing Units (CPUs) and GPUs depending on which
architecture is best suited for any given operation. The basic idea behind this functional
split is that GPUs are efficient in executing operations composed of independent and
simple tasks (e.g. Fast Fourier Transform operations), whereas the CPU is designed for
monolithic complex or iterative tasks (e.g., turbo decoding). Secondly, both CPU- and
GPU-assigned operations are executed with true parallelism: in multiple cores for the
CPU, and in multiple devices for the GPU cards, each one equipped with multiple cores.
We now illustrate some example results obtained with the F-SIM Starfish demodulator on
the preamble search function (see Sect. 4.1), which is one of the most complex and
computationally-intensive processing functions performed by the gateway. In general, the
execution time required depends on the channelization, spreading factor, and several hub
configuration parameters.
In Figure 9 we show with boxes the normalized computational effort of the preamble
search, defined as the ratio between the transmission time of a window of baseband
samples and the time required to complete preamble search on that window. If this
number exceeds 1 then it is not possible to carry out real time processing. In the chart we
compare the computational effort of CPU (single core Intel Xeon E5-2690@2.90 GHz
using SIMD instructions) vs. GPU (single NVIDIA K10 device) implementations. As it can
be seen, the CPU is very far from being able to cope with the stringent real-time
requirement of the preamble search. As a reference, we also report a rough estimate of the
number of elementary operations, in Floating Point Operations per Second (FLOPS), for
the three possible F-SIM channelization. This estimate is largely optimistic since it only
includes part of the actual computations required (specifically, only FFT operations), and
it does not take into account memory copies and other types of overhead.
15
Figure 9: Preamble search function performances on CPU and GPU.
Finally, the demodulator accepts as a configuration parameter the maximum expected
range of frequency drift. Most preamble search operations must be performed for every
possible candidate frequency in that range. Under the reasonable assumption that the
frequency drift span remains the same during the packet transmission, then the preamble
search cost increases linearly with the frequency drift span size. Under significant
frequency uncertainty, this can easily jeopardize the overall gateway performance, since
less computational resources will be available for non-preamble search operations such as
demodulation and decoding.
Starfish Hub
The hub (see Figure 10) is the central engine of the Starfish platform, with the functions
of coordinating the forward and the return links. To this aim, the hub processes the logon
requests and IP packets received from the terminals through the return link, and performs
a number of operations to guarantee network access, including network address
translation (NAT), terminal filtering, quality of service (QoS), and traffic shaping. These
operations are executed by a number of ancillary blocks, including a terminal control,
which is connected to the customer servers to deliver services and contents associated with
each terminal.
F-SIM communication performances are enhanced thanks to the software modules
implementing the power control and the congestion control algorithm. These modules
constantly monitor the channel conditions, and generate and transmit appropriate
signalling to terminals, so as to optimize F-SIM return link performance.
Finally, full compatibility with existing DVB-S2 family uplink facilities, which could be
available at the operator side (DVB-S2 Head-end), allows an easy upgrade of existing pure
broadcasting platforms to bidirectional IP-based platforms.
16
Figure 10 – Starfish gateway front panel appearance.
4. F-SIM Protocol Architecture and Its Evolution from S-MIM
The transmission in the forward-link (see Figure 11) follows the DVB-S2 standard [5].
Different Physical Channels (PCH) correspond to the different DVB-S2 carriers used
within the system. Only one transport channel is provided in the forward link, namely a
Broadcast Channel (BCH) where data and control information directed to different
terminals are multiplexed together. Demultiplexing and demodulation take place in two
sub-layers, one at Multi-Protocol Encapsulation (MPE) level and another one at MPEG-
TS level.
User data can be transported using TCP/IP or UDP/IP. Link layer signalling is transported
using UDP/IP. Additional control and management plane information (e.g. the antenna
pointing and firmware update applications) can be transported using TCP/IP or UDP/IP.
17
Demodulation &
Decoding
Demultiplexing
Decapsulation
PHY Link Layer
MPEG-TS
PCH PCH
MPE
Transport &
Network Application
Layers
Signalling Tables and
Messages
BCH
DATA PLANE CONTROL PLANE
Antenna Pointing
Link Layer ACKs
. . .
Firmware Update
(SDDP)
IP
UDP
TCP
Also in the return link case (see Figure 12), user data can be transported using TCP/IP or
UDP/IP and optionally IPsec to provide an Encrypted Channel (EC). Link layer signalling
(logon requests) are directly encapsulated into Layer 2 PDUs. Additional control and
management plane information can be transported using TCP/IP or UDP/IP.
The Header Compression sub-layer provides two access services to the IP layer:
• Non-Compressed Channel (nCC)
• Unidirectional stateless Compressed Channel (UsCC). This is applicable only to
IPv6.
The Transmission Mode sub-layer offers two access services to the Header Compression
sub-layer:
• Transparent Mode (Tr Mode): PDUs mapped into this transmission mode will be
transmitted without expecting any link layer acknowledgements in the forward
link.
• Acknowledged Mode (ACK Mode): PDUs mapped into this transmission mode will
require a link layer acknowledgement in the forward link.
Figure 11: Forward Link Protocol Stack (Terminal Side).
18
The encapsulation sub-layer offers one access service to the Transmission Mode sub-layer,
based on the Return Link Encapsulation (RLE) protocol fully specified in [6]. RLE is used
to encapsulate and if necessary fragment L3 PDUs and to insert the terminal MAC layer
address. The latter shall be unique for each terminal which is part of the EBIS network.
The physical layer in the return-link supports asynchronous access based on Spread
Spectrum ALOHA (SSA) random access (RA) technique as specified in [8] and enhanced
at the demodulator side by E-SSA iSIC techniques [10].
Modulation &
Transmission
Header
Compression
Encapsulation &
L2 Addressing
Transmission
Mode
PHY Link Layer
RLE
nCC
Tr Mode
UsCC
PDCH PCCH
ACK Mode
IP/UDP
Transport &
Network Application
Layers
Logon Requests
RACH
DATA PLANE CONTROL PLANE
Antenna Pointing
IPSec Negotiation
NECEC
HC Control
ARQ Management
Congestion & Rate
Control Format
&
Power
Control
TFI
4.1. Physical Layer Aspects
Figure 12: Return Link Protocol Stack (Terminal Side).
19
The F-SIM air interface represents an evolution of the S-MIM protocol [2]-[3]. The key
drivers for the air interface evolution have been the support of fixed (non-mobile)
terminals operating at higher frequency bands, with higher data rates and more flexible
support of different packet sizes and varying channel conditions. Similar to S-MIM the F-
SIM air interface has been designed to minimize terminal cost and maximize exploitation
of satellite spectrum given the very bursty traffic nature.
One of the aspects specifically designed in S-MIM to cope with the Land Mobile Satellite
(LMS) channel is the uplink power control. In the F-SIM system, we will not face the
fading/shadowing due to the user mobility, but still we will face time and location
dependent attenuation due to atmospheric fading, variability of satellite receive antenna
gain and geometrical path loss. Differing from fading due to mobility, Ka-band
atmospheric fading, although not infrequent, is a relatively rare event compared to mobile
shadowing/fading and characterized by a slower dynamic. Fading events are even less
frequent when operating at Ku or C-band. As the user link has to be sized for the worst
case link attenuation (geometry dependent path loss, satellite antenna gain and
atmospheric loss for the required link availability), it is of interest to exploit the intrinsic
link margin to increase the system performance. The way to optimize the gateway
demodulator incoming packets power distribution and how to approximate in practice this
distribution in a multi-beam Ka-band type of systems has been extensively discussed in
[12]. The idea beyond the F-SIM modified open loop power control algorithm is to exploit
the existing individual terminal link margins to generate an incoming packet power
distribution at the gateway demodulator which maximizes throughput. Its
implementation in F-SIM is described in Sect. 4.2.
Another aspect specific to F-SIM is operation at much higher carrier frequencies (up to
Ka-band, e.g., 30 GHz uplink) compared to S-MIM which was devised to operate at S-
band (e.g., 2 GHz uplink). It is known that a packet direct-sequence spread-spectrum
system is quite sensitive to possible carrier frequency errors for packet initial acquisition
[10]. The complexity of the gateway demodulator is largely dependent on the packet
acquisition unit and to the incoming packets’ carrier frequency uncertainty. It is therefore
important to limit the transmitted packet’s carrier frequency deviation from the nominal
value. Based on practical implementation of the gateway demodulator, a typical F-SIM
terminal is expected to satisfy the following requirements:
• Each burst is transmitted with a frequency error less than 1.5 kHz with respect to
the nominal frequency in Ku/Ka-band. Moreover, within a burst, the peak-to-peak
frequency variation shall be less than 20 Hz.
• The average chip rate of each burst shall have an error less than 0.25 ppm with
respect to the nominal chip rate. Moreover, within a burst, the peak-to-peak chip
rate variation shall be less than 0.1 ppm.
An accurate reference clock is needed at the terminal to control both the chip rate and the
uplink frequency. A straightforward (but rather expensive) solution is to use a highly
stable reference oscillator at the user terminal. However, this requirement contrasts with
the need to keep the terminal cost low thus avoiding expensive thermally controlled
oscillators. Alternative cost-effective solutions to obtain the required frequency accuracy
20
exist at the network level. As discussed in Section 3.2, F-SIM protocol envisions
distribution of Network Clock Reference (NCR) as part of the forward-link signalling.
The NCR is carrying an accurate time stamp (a 27 MHz counter) as part of the forward-
link signalling. By extracting the counter value from the forward-link and comparing its
value to a locally generated counter, the user terminal can synchronise its local oscillator
to the master clock at the gateway. NCR broadcasting consumes only a fraction of the
forward link bandwidth. However, it should be noted that for a network with fixed user
terminals, the repetition rate of NCR packets can be as low as 10 packets per second to
reduce the overhead. From the implementation aspect, the NCR extraction is performed
based on the content of NCR packet (i.e. the 42-bit counter within the Packet Clock
Reference). As such, the NCR extraction is independent of the forward link physical layer
characteristics (symbol rate, modulation and coding). There are existing and emerging
DVB-S2 demodulator chipsets that integrate a “start of field” (SOF) signal which allows a
very accurate extraction of the NCR reference clock [13]. Without the SOF signal, it is still
possible to achieve good accuracy by averaging over a time window so as to reduce the
jitter inside the demodulator.
In addition to the terminal reference clock stability, the overall carrier frequency error
budget is also characterized by the Doppler frequency shift induced by the satellite motion
and the transponder frequency translation error. Both parameters are somewhat
proportional to the carrier operating frequency thus more relevant for Ka-band than S-
band. Due to the slowly varying and periodic nature of frequency offset variation
(following almost a daily pattern), in principle, a frequency variation profile for such
change could be established and used at the gateway to pre-compensate for the offset.
Satellite distance measurements and/or beacon signals may also be used to create such
profiles. However, the frequency offset predictions may not follow precisely the actual
frequency offset or such data may not be readily available to satellite network operators.
Alternatively, some monitoring techniques could be envisaged to track and compensate
for such slowly varying frequency deviations as a common mode, affecting signals received
from all user terminals. The latter technique is adopted in the Starfish gateway.
Finally, when moving from S-band to higher frequency values, the impact of the carrier
phase noise needed to be considered. In general, it must be recalled that the lower the
transmission symbol rate, the higher the expected degradation due to the phase noise. For
this reason, it was confirmed by numerical results (see Figure 13) that the exploitation of
known pilot symbols in the transmitted F-SIM signal control channel is necessary to limit
the degradation of the Packet Error Rate (PER) for PER < 5∙10-3. This requires a proper
optimization of the power splitting (e.g., the β parameter in [8]) between the data and
control channels and the carrier phase estimator window length.
21
Figure 13: Simulated F-SIM Packet Error Rate (PER) in Additive White Gaussian Channel
with carrier phase noise mask in line with the DVB-RCS versus the Carrier-to-Noise ratio
(C/N) for both data-aided (Pilot) and not data-aided (Viterbi & Viterbi) phase estimation
techniques.
The F-SIM physical layer follows the same S-MIM 3GPP Wideband Code Division
Multiple Access design philosophy [3] with a number of extensions and adaptations.
Compared to the original S-MIM specifications, the number of physical layer
configurations supported has been largely extended to be able to cope with very diverse
link budgets in the different satellites. The key F-SIM physical layer parameters have been
summarized in Table 3.
F-SIM Physical Layer Parameter
Value(s)
Channelization bandwidth [MHz]
2.5, 5, 10
Chip rate [Mcps]
1.92, 3.84, 7.68
Spreading factor
16, 32, 64, 128, 256
Bit rates [kbps]
4.9 - 161
Symbol rates [ksps]
15 - 480
Frame payload size [bytes]
38, 78, 150, 300, 625, 1513
Frame duration [ms]
0.5, 10
Preamble duration [chips]
1536, 3072, 6144, 12288, 24756
Table 3: F-SIM key physical layer parameters.
Within a single channel, different modes can be used by different terminals according to
their current channel conditions. The list of allowed modes is determined by the hub and
broadcast in the control signalling.
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
-29.00 -28.00 -27.00 -26.00 -25.00 -24.00 -23.00 -22.00
PER
C/N [dB]
AWGN - K=1200 (b=0.1)
PhNoise@15ksymb/s - Pilot Window=135
(b=0.1)
PhNoise@15ksymb/s - Pilot Window=105
(b=0.15)
PhNoise@15ksymb/s - V&V window=135
22
A major F-SIM deviation from S-MIM is that the preamble and the Physical Control
Channel (PCCH) adopt the same spreading factor as the Physical Data Channel (PDCH)
and not a constant value (i.e., 256) as before. This means that the gateway demodulator
has to have a distinct acquisition unit for each spreading factor adopted in the network.
4.2. Data Link Layer Aspects
Signalling
The F-SIM forward link signalling is carried by means of IP packets, following the format
typical of DVB signalling tables [7]. This additional signalling is advertised as a series of
programs in the Program Map Table (PMT). The return link signalling only foresees the
logon request, which is native layer 2 signalling. Table 4 provides an overview of the
different F-SIM signalling tables.
Name
Link
Nature
Type
Content
Configuration
Table (SCT).
FWD
Broadcast
M+p
Static information about all the
return link carriers (beam,
polarization, center frequency,
chip rate, etc…) and all the
allowed transmission formats
Access Table
(SAT)
FWD
Broadcast
M+p
Association between RL carriers
and transmission
formats to
service classes, static parameters
for congestion control
Dynamic Table
(SDYT).
FWD
Broadcast
M+P
Dynamic load and power control
parameters for the different
profiles
Link Layer Service
signalling table
(LLST)
FWD
Broadcast
M+p
List of allowed IP addresses per
service class, together with header
compression parameters
Link Layer
Acknowledgement
(LLA)
FWD
Unicast
O+C
Logon Response
(LRS)
FWD
Unicast
M+C
Beam and carrier assigned to the
terminal, allowed service,
terminal IP addresses
Individual
Message (SIM)
FWD
Unicast
O+C
Force a terminal to stop
transmitting, log off, or assign the
terminal to different multicast
signalling group(s)
Group Message
(SGM)
FWD
Multicast
O+C
Same as above, but for all
terminals belonging to a given
group
Logon Request
(LRQ)
RTN
NA
M+C
Table 4: Overview of the F-SIM signalling tables: Legend: M = mandatory, O = optional, P =
periodically sent with high periodicity, p = periodically sent with low periodicity, C =
occasionally sent.
23
Logon procedure
The logon procedure is the procedure through which the terminal registers into the EBIS
network. The logon procedure is initiated by the terminal through a logon request and it
follows an algorithm which, in case of multi-beam coverage in the return link, allows the
terminal to select the correct beam according to its geographical position. This is necessary
because F-SIM supports an asymmetric satellite coverage for forward and return links.
Three different types of logon procedures can be distinguished depending on the network
information knowledge:
• Beam search logon: Is used the first time the terminal registers into the EBIS
network or when the previously known system configuration is outdated and the
terminal needs to once again select a return link beam. The procedure uses power
ramping, starting with a low transmit power and monotonically increasing the
transmit power in the event no response is received from the hub.
• Cold logon: This procedure is used when the previous logon is no longer valid but
the configuration available at the terminal is compatible with the signalling so that
it is not necessary to select the beam to be used. The procedure also uses transmit
power ramping but it is only aimed at the beam and carrier that the terminal was
using before performing the cold logon.
• Warm logon: This procedure is executed periodically upon expiration of the
validity of the previous logon (e.g. once a day). This procedure does not rely on
transmit power ramping but rather uses one transmit format selection strategy like
any other service in the system.
The logon response from the hub contains always the Es/N0 value with which the logon
request was received at the hub. This allows the terminal to compute the maximum Es/N0
that it can achieve at the hub by using its maximum transmit power, so as to implement
the required F-SIM power control algorithm.
Power Control and Transmission Format Selection
Power Control
F-SIM makes use of enhanced power control algorithms in order to maximize the
throughput and minimize packet loss in the return link channels. As previously explained,
the return link of F-SIM uses SSA with iSIC at the receiver (demodulator). When such a
receiver is used the power distribution of the received packet has a strong impact on the
packet error rate and the maximum traffic load which can be supported. In fact, the
optimal Es/N0 distribution is known to be a uniform distribution in dB ranging from
[Es/N0]min to [Es/N0]max , where [Es/N0]min is the minimum Es/N0 required for a burst
to be decoded error free, and [Es/N0]max is the maximum achievable Es/N0 from any
terminal [12].
In a scenario in which all terminals have the same link budget (same [Es/N0]max ) the
optimal Es/N0 distribution could be achieved simply by letting terminals use a uniform
distribution of their transmit power in dB [12]. However in a real setting a smarter
24
approach is needed to achieve the optimal distribution, since terminals have different
instantaneous link budgets. The differences could be caused, for example, by:
• different transmitted EIRP (e.g. different antenna size or amplifier power);
• different positions of the terminals in the beam, resulting in different G/T at the
satellite antenna;
• dynamic channel attenuations (e.g. atmospheric attenuation due to rain).
The power control signalling in F-SIM consists of an array of n increasing Es/N0 values
(in dB) and “n-1” probability values p[i] (see Figure 14). Each terminal has obtained at
logon an estimate of the maximum Es/N0 with which its packets arrive at the hub,
[Es/N0]max. According to its value of [Es/N0]max the terminal is assigned to one of the “n-
1” implicitly defined segments as follows:
• If [Es/N0]max ≤ [Es/N0][0], the terminal is not assigned to any segment
• If [Es/N0][i] ≤ [Es/N0]max < [Es/N0][i+1], the terminal belongs to segment i, for
i=0,..., n-2
• If [Es/N0][n-1]≤ [Es/N0]max, the terminal belongs to segment n-1
If a terminal is not assigned to any segment, it is not allowed to transmit. Otherwise, the
terminal is allowed to transmit, but according to specific rules. Let us assume a terminal
belongs to segment “i”. The terminal proceeds as follows, for each packet to transmit:
• With probability p[i] it randomizes its transmit power so that the Es/N0 at the
receiver gateway lies between [Es/N0][i] and the minimum of [Es/N0][i+1] and
[Es/N0]max ;
• With probability 1-p[i] it randomizes its transmit power so that the Es/N0 at the
receiver gateway lies between [Es/N0][0] and the minimum of [Es/N0][i+1] and
[Es/N0]max .
Figure 14 provides an example for the case n=4 and a terminal belonging to segment 2.
E
s
/N
0
[0] E
s
/N
0
[1] E
s
/N
0
[2] E
s
/N
0
[3]
p[0]p[1] p[2]
segment 0 segment 1segment 2
[E
s
/N
0
]max
Prob = p[2]
Prob = 1- p[2]
Figure 14: Power control signalling example.
In the following we will provide an example to illustrate how the power control algorithm
in F-SIM works. Let us assume the terminals are divided into two groups according to
their [Es/N0]max as follows:
25
• 50% of the users can achieve [Es/N0]max = 20 dB.
• 50% of the users can achieve [Es/N0]max = 10 dB due to, for example, a smaller
maximum transmit power, a marginal position in the beam, or due to rain
attenuation.
The minimum required Es/N0 to decode the burst error free shall be 0 dB in this case.
Table 5 shows two different sets of power control parameters for this setting. In the first
case we let each terminal randomize its Es/N0 at the receiver uniformly between the
minimum required 0 dB and the maximum it can achieve. The resulting distribution of
the Es/N0 can be observed in the left part of Figure 15. This distribution is far from being
uniform. However, by using the optimal set of power control parameters, shown in Table
5 one can obtain a perfectly uniform Es/N0 distribution at the hub, as it can be seen in the
right part of Figure 15.
i 0 1 2
Uniform
Es/N0 [i] 0 20
p[i] 0
Optimal
Es/N0 [i] 0 10 20
p[i]
0
1
Table 5: Example power control parameters.
E
s
/N
0
probability
E
s
/N
0
probability
20 dB
10 dB0 dB 20 dB
10 dB
0 dB
Figure 15: Es/N0 histogram. On the left the case is shown in which all terminals
randomize their power uniformly. On the right we show the case with optimized
parameters.
Transmission Format Selection
In general, when a terminal wants to transmit a packet of a given service through the
return link it will have to choose among a set of possible transmission formats which
correspond to different physical layer configurations. In F-SIM, it was assumed that the
different transmission formats employ the same modulation and channel coding but
different spreading. Given the fact that the chip rate (and therefore the bandwidth) is fixed
for a given channel, transmission formats with larger spreading factors will be more robust
but they will also provide lower information bit rates.
26
In F-SIM there are two alternatives for the transmission format selection. The first aims
at maximizing the bit rate of every terminal and lets a terminal use a transmission format
whenever its achievable Es/N0 at the hub is high enough to be decoded error free. When
this strategy is used terminals usually employ the transmission format with the highest
bitrate for the service class. The second alternative aims at maximizing the overall
throughput of the system and usually leads to terminals using transmission formats with
lower bit rates.
Scheduling and QoS
The scheduling of F-SIM has also been largely modified compared to S-MIM. This
scheduling algorithm provides a QoS based on the service classes in which retransmissions
are treated as new packets.
The scheduling algorithm in F-SIM in simplified form operates as follows:
1. The packets arriving from layer 3 (L3) are sent to different ARQ buffer according
to their service class. At the insertion in the ARQ buffer, each L3 PDU shall be
marked with a timestamp (TS).
2. The terminal always selects the L3 packet to be transmitted from the lowest
available service class (highest priority). If there is more than one packet in this
class, the packet with the smallest timestamp (TS) is selected (i.e. the oldest packet
in the queue). This packet is removed from the ARQ buffer.
3. The terminal then executes the congestion control algorithm using the parameters
distributed by the gateway for the corresponding class of service and actual
congestion level. Due to the outcome of this algorithm, the transmission
opportunity might be used or dismissed with a certain probability. In the latter
case, the L3 packet is left in the ARQ buffer and its TS increased by a back-off time.
4. Otherwise, the terminal encapsulates and if necessary fragments the packet into
several frames and it selects a transmission format. The fragments are then placed
in a Frame PDU (FPDU) buffer which is a FIFO queue.
5. The terminal extracts frames from the FPDU buffer periodically after waiting for
an inter fragment time interval and the selected frame is transmitted over the
return link.
6. If an acknowledgement for a given packet is not received within the ACK timeout
and the packet has not been retransmitted the maximum number of allowed
retransmissions, the packet is reinserted in the ARQ buffer and its timestamp is
increased by the ACK timeout.
IP Header Compression
F-SIM can use IPv4 and IPv6. When IPv6 is used there is an optional header compression
mechanism which can decrease the protocol overhead. This header compression
mechanism is derived from RFC 6282 and allows compressing the IPv6 and UDP headers
from 48 bytes down to 5 bytes (7 bytes if Cyclic Redundancy Check (CRC) is used). The
compression works by assuming typical values for the different fields of the IP packet.
When a packet uses a typical value for an IPv6 packet field this is indicated (for example)
by setting a bit to 0. In case another value is used, the bit is set to 1 and the actual value of
27
the field needs to be provided. Hence this compression only works effectively in case the
terminal uses these typical values for the different fields.
This algorithm compresses the IPv6 destination addresses (128 bits) and the destination
port (2 bytes) by defining 256 different destination contexts (8 bits) which are short names
for a set of 256 different IPv6 addresses and ports. Furthermore, when IP header
compression is used the source IPv6 address is not transmitted since it can be obtained
from the MAC layer address. Finally 256 (8 bit) source contexts are defined which allow
using 256 different UDP source ports.
The source contexts are defined a priori and hard coded in the system. The destination
contexts are defined in the LLST table and can be updated on demand.
5. Field Trials
Validation tests
Validation tests on F-SIM compliant terminals have been performed using the laboratory
and satellite platform illustrated in Figure 16. In both platforms, the different F-SIM
physical layer formats detailed in Sect. 4.1 have been individually validated through the
CRC checking of long series of bursts generated by the different terminals. Moreover, we
verified the correct demodulation of several terminals using simultaneously different F-
SIM physical formats through the satellite platform (return link on Ka-Sat and forward
link on Eutelsat 10A in Ku-band).
F-SIM demodulator performance using a single terminal
The aim of this test was assessing the implemented terminal and gateway performance in
presence of a single user affected by Additive White Gaussian Noise (AWGN) and phase
noise. Terminals output bursts are recorded and demodulated by a Starfish demodulator
for different Eb/N0 values. Through the terminal’s management interface, a long series of
bursts is transmitted and the transmission is repeated for a sufficiently long time to
estimate the demodulator Packet Loss Ratio (PLR) and Bit Error Rate (BER) at the current
signal-to-noise ratio working point. The PLR and BER statistics have been collected and
then compared with theoretical results. The physical configuration used for the example
processed in the Figure below is the following:
• Chip rate : 3.84 Mc/s
• Packet size : 150 bytes
• Spreading factor : 256
In Figure 17, the results are compared to the theoretical performance of the rate 1/3 GPP
turbo FEC implemented in F-SIM demodulator. The implementation losses between
theory and practical tests are mainly caused by the noisy gateway channel estimation. In
particular, as shown also by the simulation reported in Figure 13, the main reason for
demodulator performance loss is the Ka-band terminal carrier phase noise. Further
gateway channel estimation algorithms optimization is 0n-going to reduce the observed
implementation losses.
28
Figure 16: F-SIM validation platforms.
29
Figure 17: Measured PLR vs Eb/N0 versus the theoretical results.
Another important value to measure is the residual power after cancellation of a packet
due to channel estimation errors at the gateway demodulator side. This residual power
value impacts the SIC process efficiency, and depends on the channel estimation quality.
Figure 18 provides the measured residual packet power after cancellation, as a function of
the received Eb/N0, during the demodulation process on the Starfish demodulator.
Figure 18: Residual packet power after gateway demodulator cancellation vs Eb/N0.
30
F-SIM spectral efficiency using traffic emulation
This series of tests has been realized using a traffic emulator instead of one single terminal,
in order to validate the total capacity of the F-SIM system when operating in RA mode
with Poisson type of bursty traffic. The traffic emulator is able to generate traffic
equivalent to several hundreds or thousands of terminals simultaneously transmitting.
Different F-SIM configurations have been used with:
• one unique channel of 5 MHz
• the highest (256) and the lowest (16) spreading factors
• three different packet sizes (38, 78, 150 bytes)
The AWGN added during these tests is equivalent to the link budget using a satellite G/T
of 18 dB/K in line with the link budget assumptions of Sect. 3.3. The incoming packets
power spreading assumed corresponds to the theoretical one obtained thanks to the
implementation of the power control techniques described in Sect. 4.2. An additional
margin is introduced compared to the reference link budgets in order to account for the
implementation losses reported in Figure 17.
As shown in Table 6, the measured throughput performance results span from 1.5 to 3
b/s/Hz depending on the selected physical layer configuration. This is a truly remarkable
performance range considering that the terminals are operating in a totally uncoordinated
RA fashion and in time asynchronous mode.
Looking at the experimental results two main conclusions can be drawn. The first is that
a larger spreading factor generates a greater link margin for all terminals at the expense
of a lower individual terminal bit rate. Consequently, as expected [12], the power control
can spread the packets’ power over a wider range, and thus achieve a larger aggregated
throughput. The highest aggregate throughput is obtained with the configuration
corresponding to a larger population of active terminals each transmitting at a lower bit
rate. The second consideration is related to the packet duration: the larger the size, the
more efficient the power control. In fact, with a larger spreading factor, the number of
simultaneous users is increased and the power distribution gets closer to the ideal uniform
(in dB) distribution. Over a long transmission time, this increases the average total
capacity of the F-SIM system.
31
5 MHz Bandwidth
Spreading Factor 16
Spreading Factor 256
CR3840SF16DS38
CR3840SF16DS78
CR3840SF16DS150
CR3840SF256DS38
CR3840SF256DS78
CR3840SF256DS150
MAC-Load
bits/s/Hz
1.43
1.6
1.66
2.33
2.79
2.83
PLR
0.0004
0.0002
0.0002
0.00268
0.00029
0.00021
SIC iterations
14
16
14
30
29
26
MAC-Load
bits/s/Hz
1.55
1.72
1.79
2.46
2.93
2.93
PLR
0.0087
0.006
0.002
0.00724
0.05
0.00045
SIC iterations
30
32
25
32
32
31
MAC-Load
bits/s/Hz
1.69
1.86
1.92
2.6
3.06
3.07
PLR
0.13
0.19
0.12
0.117
0.41
0.2
SIC iterations
32
32
32
32
32
32
Table 6: F-SIM spectral efficiency emulated results.
6. Conclusions
This paper described the evolution of the S-MIM ETSI protocol into the F-SIM protocol
developed by Eutelsat in cooperation with ESA, DLR and MBI. The F-SIM protocol has
been adopted by the recently deployed Eutelsat Broadcast Interactive System (EBIS). The
new EBIS concept allows TV platform operators to deploy their own ecosystem of linear
television and connected television services directly via existing Ku or Ka-band satellites.
At the heart of the system are the ‘smart LNB’ and the Starfish hub technologies.
The Smart LNB is a new-generation electronic feed connected to an antenna with an
embedded transmitter that makes it easy for service providers to deploy interactive
applications. End customers will be attracted by value added connected television services
such as: Push Video On Demand, pay-per-view, social TV, Hybrid Broadcast Broadband
TV, multiscreen viewing, personal subscription management and live show participation.
Broadcasters benefit from a wide range of applications in direct contact with their installed
base for customer intelligence, promotion of new personalised services, enhanced
customer support and remote diagnostics, audience measurement and more.
The Smart LNB addresses the wide consumer market, with the intention to reach millions
of users. Transforming unidirectional experience into a full interactive one, at costs
compatible with consumer grade equipment, is a key for large adoption of this technology.
32
The F-SIM random access protocol combined with Starfish gateway advanced signal
processing achieves an unprecedented return link spectral and power/energy efficiency
i.e. three order of magnitudes higher throughput than conventional ALOHA or Slotted
ALOHA schemes. This element combined with the low terminal cost, high protocol
reliability and very reduced packet transmission latency represent essential characteristics
to reach the overall service provision cost target.
The Starfish hub is characterized by a very scalable design which allows easy adaptation
to the evolving network needs by exploiting state-of-the-art SDR technologies and signal
processing techniques. To our knowledge, this is the first realization of a satellite two-way
gateway fully based on SDR techniques.
The F-SIM protocol and associated technology have been designed with flexibility in mind.
So they can find application in other important markets, particularly M2M applications
that are targeting a very large worldwide market. We believe that F-SIM technology,
thanks to its flexibility and high efficiency, is an appropriate tool that will allow satellites
to capture a large share of these important markets.
References
[1] H. Fenech et al, “Future High Throughput Satellite Systems”, First IEEE AESS
European Conference on Satellite Telecommunications (ESTEL), October 2-7 2012,
Rome, Italy, pp. 1-7 .
[2] S. Scalise, C. Párraga Niebla, R. De Gaudenzi, O. Del Rio Herrero, D. Finocchiaro, A.
Arcidiacono, ”S-MIM: a Novel Radio Interface for Efficient Messaging Services over
Satellite”, IEEE Communications Magazine, Vol. 51 , No. 3, March 2013, pp. 119-125.
[3] ETSI TS 102 721, “Satellite Earth Stations and Systems (SES); Air Interface for S-band
Mobile Interactive Multimedia (S-MIM),” Ver. 1.2.1 (2013-08).
[4] Eutelsat Smart LNB initiative web page:
http://www.eutelsat.com/en/services/broadcast/direct-to-home/smart-LNB.html
[5] ETSI EN 302 207, "Digital Video Broadcasting (DVB); Second generation framing
structure, channel coding and modulation systems for Broadcasting, Interactive
Services, News Gathering and other broadband satellite applications (DVB-S2)", Ver.
1.3.1, March 2013.
[6] ETSI TS 103 179, "Satellite Earth Stations and Systems (SES); Return Link
Encapsulation (RLE) protocol", Ver. 1.1.1, August 2013
[7] EN 300 468, “Digital Video Broadcasting (DVB); Specification for Service Information
(SI) in DVB systems”, Ver. 1.14.1 (2014-05).
[8] Air Interface for Fixed Satellite Interactive Multimedia (F-SIM); Physical Layer
Specification, Return Link Asynchronous Access, Eutelsat proprietary specification.
[9] Air Interface for Fixed Satellite Interactive Multimedia (F-SIM): Link Layer and
System Signalling Specification, Eutelsat proprietary specification.
[10] O. del Rio Herrero, R. De Gaudenzi, ”High Efficiency Satellite Multiple Access
Scheme for Machine-to-Machine Communications”, IEEE Trans. on Aerospace
Systems, Vol. 48, No. 6, October 2012, pp. 2961-2989.
33
[11] S. Chatzinotas, B. Ottersten, R. De Gaudenzi, “Cooperative and Cognitive Satellite
Systems”, 1st Edition, Elsevier Academic publishing, May 2015, ISBN No.
9780127999487.
[12] F. Collard, R. De Gaudenzi, "On the Optimum Packet Power Distribution for
Spread Aloha Packet Detectors with Iterative Successive Interference Cancellation",
IEEE Trans. Wireless Comm. Vol. 13, No. 12, December 2014, pp. 6783-6794.
[13] ST Microelectronics STiD135 data sheet: http://www.st.com/st-web-
ui/static/active/en/resource/technical/document/data_brief/DM00136025.pdf
34
