Analysis and optimization of the satellite-to-plane link of an aeronautical global system

Projecte Fi de Carrera Enginyeria de Telecomunicació __________________________________________________________________________________ Analysis and Optimization of the Satellite-to-Plane Link of an Aeronautical Global System Ricard Alegre Godoy __________________________________________________________________________________ Director: María Ángeles Vázquez Castro Departament de Telecomunicació i Enginyeria de Sistemes Escola Tècnica Superior d’Enginyeria (ETSE) Universitat Autònoma de Barcelona (UAB) Setembre 2010 MTQ CDT El sotasignant, María Ángeles Vázquez Castro, Professor de l’Escola Tècnica Superior d’Enginyeria (ETSE) de la Universitat Autònoma de Barcelona (UAB), CERTIFICA: Que el projecte presentat en aquesta memòria de Projecte Fi de Carrera ha estat realitzat sota la seva direcció per l’alumne Ricard Alegre Godoy. I, perquè consti a tots els efectes, signa el present certificat. Bellaterra, 2/09/2010. Signatura: María Ángeles Vázquez Castro Contents List of Acronyms vi List of Figures ix List of Tables xiii Introduction xvi 1 ANTARES Global System 1 1.1 Introduction to aviation communications . . . . . . . . . . . . . . 1 1.2 ANTARES System and Requirements . . . . . . . . . . . . . . . 2 1.2.1 Architecture ......................... 2 1.2.2 TypesofLinks ........................ 3 1.2.3 Protocol Stack . . . . . . . . . . . . . . . . . . . . . . . . 3 1.2.4 Global System Parameters . . . . . . . . . . . . . . . . . . 4 1.2.5 ANTARES System Requirements . . . . . . . . . . . . . . 5 1.3 Problem statement . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2 Design of the Radio Resource Management for the ANTARES system 11 2.1 System Level Architecture Design . . . . . . . . . . . . . . . . . . 11 2.1.1 Satellite Based Wide Area Network (SATBASED-WAN) Architecture ......................... 12 2.1.1.1 Distributed scheduling and resource allocation process ....................... 12 i 2.1.1.2 Centralized scheduling and resource allocation process ....................... 14 2.1.2 Terrestrial Based Wide Area Network (TERRBASEDWAN) Architecture . . . . . . . . . . . . . . . . . . . . . 15 2.1.2.1 Distributed scheduling and resource allocation process ....................... 15 2.1.2.2 Centralized scheduling and resource allocation process ....................... 16 2.2 MAC-level Architecture . . . . . . . . . . . . . . . . . . . . . . . 17 2.2.1 MAC-level architecture for distributed system level architectures............................ 17 2.2.2 MAC-level architecture for centralized system level architectures............................ 19 2.3 MF-TDMA SF Design . . . . . . . . . . . . . . . . . . . . . . . . 21 2.3.1 General MF-TDMA SF structure . . . . . . . . . . . . . . 21 2.3.2 Fixed T S MF-TDMA .................... 23 2.3.3 Dynamic T S MF-TDMA .................. 24 2.3.4 Figures of Merit for the MF-TDMA SF performance . . . 24 3 Analysis of the Erasure Channel and Proposed Solution 26 3.1 Introduction to LL-FEC . . . . . . . . . . . . . . . . . . . . . . . 26 3.1.1 LL-FEC concept . . . . . . . . . . . . . . . . . . . . . . . 26 3.1.2 From Information Theory . . . . . . . . . . . . . . . . . . 26 3.2 Justication of the LL-FEC use in ANTARES . . . . . . . . . . . 28 3.2.1 Link Layer channel vs Physical Layer Channel . . . . . . 28 3.2.2 Aeronautical Erasure Channel Models . . . . . . . . . . . 30 3.2.2.1 Erasure Model for Commercial Aviation . . . . . 30 3.2.2.2 Erasure Model for Helicopters . . . . . . . . . . 34 3.3 DesignOptions............................ 43 3.3.1 Reed Solomon (RS) Codes . . . . . . . . . . . . . . . . . . 43 3.3.2 RS+Channel Interleaver . . . . . . . . . . . . . . . . . . . 43 3.3.3 Low Density Parity Check (LDPC) codes . . . . . . . . . 44 3.3.4 RaptorCodes......................... 44 ii 3.3.5 OtherCodes ......................... 45 3.4 Comparative Performance Analysis . . . . . . . . . . . . . . . . . 45 3.4.1 Information from literature review . . . . . . . . . . . . . 45 3.4.1.1 Eciency...................... 45 3.4.1.2 Complexity . . . . . . . . . . . . . . . . . . . . . 46 3.4.2 Simulations on comparative analysis . . . . . . . . . . . . 47 3.4.2.1 Eciency simulations . . . . . . . . . . . . . . . 47 3.4.2.2 Complexity simulations . . . . . . . . . . . . . . 48 3.4.3 Conclusions on code comparative . . . . . . . . . . . . . . 51 3.5 Code Choice and Performance results . . . . . . . . . . . . . . . 52 3.5.1 CodeChoice ......................... 52 3.5.2 Discussion on the use of GSE . . . . . . . . . . . . . . . . 52 3.5.3 Simulator Construction . . . . . . . . . . . . . . . . . . . 53 4 Payload Dependent System Model 56 4.1 Conventional Payload Model . . . . . . . . . . . . . . . . . . . . 57 4.1.1 Payload Elements . . . . . . . . . . . . . . . . . . . . . . 57 4.1.2 Antenna Design . . . . . . . . . . . . . . . . . . . . . . . . 57 4.1.3 Payload Architecture . . . . . . . . . . . . . . . . . . . . . 58 4.1.3.1 Forward Link Architecture . . . . . . . . . . . . 58 4.1.3.2 Return Link Architecture . . . . . . . . . . . . . 59 4.2 Flexible Payload Model . . . . . . . . . . . . . . . . . . . . . . . 60 4.2.1 Payload Elements . . . . . . . . . . . . . . . . . . . . . . 60 4.2.2 Antenna Design . . . . . . . . . . . . . . . . . . . . . . . . 60 4.2.3 Payload Architecture . . . . . . . . . . . . . . . . . . . . . 60 4.2.3.1 Forward Link Architecture . . . . . . . . . . . . 60 4.2.3.2 Return Link Architecture . . . . . . . . . . . . . 62 4.3 Beam Hopping Payload Model . . . . . . . . . . . . . . . . . . . 62 4.3.1 Payload Elements . . . . . . . . . . . . . . . . . . . . . . 62 4.3.2 Antenna Design . . . . . . . . . . . . . . . . . . . . . . . . 64 4.3.3 Payload Architecture . . . . . . . . . . . . . . . . . . . . . 64 4.3.3.1 Forward Link Architecture . . . . . . . . . . . . 64 4.3.3.2 Return Link Architecture . . . . . . . . . . . . . 66 iii 5 Analysis of co-Channel Interference 67 5.1 Twobeammodel........................... 67 5.2 Generalmodel ............................ 70 6 Overall Performance 73 6.1 Simulations and Performance for the RRM design . . . . . . . . . 73 6.1.1 Conclusions and delay estimation for the architecture design 73 6.1.2 Simulations and Performance for MF-TDMA design . . . 75 6.1.2.1 Simulations for a system with 3 GES . . . . . . 76 6.1.2.2 Simulations for a system with 5 GES . . . . . . 77 6.1.2.3 Conclusions on simulation results . . . . . . . . 78 6.2 Simulations and Performance for LL-FEC design . . . . . . . . . 80 6.2.1 Simulations for the Erasures produced by airplanes in the same line of sight with the satellite scenario . . . . . . . 80 6.2.2 Simulations for the Erasures produced by the airplane itself when maneuvering scenario . . . . . . . . . . . . . 81 6.2.3 Simulations for the Erasures produced by the blades scenario.............................. 81 6.2.4 Simulations for the Erasures produced by blades and buildings scenario . . . . . . . . . . . . . . . . . . . . . . 82 6.3 Simulations and Performance on the Interference Analysis study 90 6.3.1 70 Beam system . . . . . . . . . . . . . . . . . . . . . . . 91 6.3.1.1 Dierences between payloads . . . . . . . . . . . 91 6.3.1.2 Interference generated by adjacent beams . . . . 92 6.3.1.3 Interference generated by non-adjacent beams . 93 6.3.1.4 Interference generated by far away beams . . . . 94 6.3.1.5 Conclusions for the 70 beam system . . . . . . . 94 6.3.2 ANTARES system . . . . . . . . . . . . . . . . . . . . . . 96 6.3.2.1 Analysis of the SINR . . . . . . . . . . . . . . . 96 6.3.2.2 Conclusions on the ANTARES system . . . . . . 100 7 Conclusions and Further Work 101 iv A Link Budget 103 A.1 Geometrical Aspects . . . . . . . . . . . . . . . . . . . . . . . . . 103 A.2 Link Budget Calculation in function of θ ..............105 A.2.1 Forward Uplink . . . . . . . . . . . . . . . . . . . . . . . . 105 A.2.2 Forward Downlink . . . . . . . . . . . . . . . . . . . . . . 106 A.2.3 Return Uplink . . . . . . . . . . . . . . . . . . . . . . . . 106 A.2.4 Return Downlink . . . . . . . . . . . . . . . . . . . . . . . 107 A.3 Link Budget and Payload Values . . . . . . . . . . . . . . . . . . 107 B Guidelines for LL-FEC implementation in ANTARES 110 B.1 Undened parameters and general procedure for RS implementation110 B.2 Characteristics of the Erasures produced by airplanes in the same line of sight with the satellite scenario . . . . . . . . . . . 112 B.3 Characteristics of the Erasures produced by the airplane itself scenario ................................113 B.4 Characteristics of the Erasures produced by the blades scenario 113 B.5 Characteristics of the Erasures produced by blades and buildings scenario................................113 Bibliography 115 v 6.15 Received SINR within the edge of the beam of interest and within the edge of the interferers for conventional payload . . . . . . . . 99 6.16 SINR evolution in the ANTARES system for the conventional and beam-hopping payloads . . . . . . . . . . . . . . . . . . . . . 100 A.1 Considered scenario . . . . . . . . . . . . . . . . . . . . . . . . . 104 A.2 Antenna pattern G(θ) with θ−3dB = 0.245 º for the conventional/beamhoppingpayloads...........................104 B.1 Fluxdiagram.............................112 xii List of Tables 1.1 TypesofLinks ............................ 3 1.2 Global system parameters . . . . . . . . . . . . . . . . . . . . . . 5 1.3 Satellite segment parameters . . . . . . . . . . . . . . . . . . . . 5 1.4 User segment parameters . . . . . . . . . . . . . . . . . . . . . . 5 1.5 Ground segment parameters . . . . . . . . . . . . . . . . . . . . . 5 1.6 Bitrate demanded per GES . . . . . . . . . . . . . . . . . . . . . 7 1.7 Unicast QoS requirements per CoS . . . . . . . . . . . . . . . . . 8 1.8 Multicast QoS requirements . . . . . . . . . . . . . . . . . . . . . 8 1.9 Message sizes per CoS . . . . . . . . . . . . . . . . . . . . . . . . 8 1.10 Unicast Messages arrival rates per CoS (messages/s) . . . . . . . 9 1.11 Multicast Messages arrival rates per CoS (messages/s) . . . . . . 9 2.1 MF-TDMA SF parameters . . . . . . . . . . . . . . . . . . . . . 22 2.2 Transmission time, t a for each MODCOD . . . . . . . . . . . . . 22 3.1 Standard airplane lengths . . . . . . . . . . . . . . . . . . . . . . 32 3.2 Minimum, mean and maximum values of erasure duration . . . . 33 3.3 Number of packets lost at each bitrate 4096 bit frame length . . 33 3.4 Number of packets lost at each bitrate 8192 bit frame length . . 33 3.5 Time in seconds the signal is blocked by helicopter blades . . . . 35 3.6 Duty cycle of the erasures . . . . . . . . . . . . . . . . . . . . . . 37 3.7 Raptor over RS outperformance for a PER@1e-4 at v=100km/h 45 3.8 RS encoding algorithm complexity . . . . . . . . . . . . . . . . . 46 3.9 RS decoding algorithm complexity . . . . . . . . . . . . . . . . . 46 xiii 3.10 LDPC encoding algorithm complexity . . . . . . . . . . . . . . . 47 3.11 LDPC decoding algorithm complexity . . . . . . . . . . . . . . . 47 3.12 Raptor encoding algorithm complexity . . . . . . . . . . . . . . . 47 3.13 Raptor decoding algorithm complexity . . . . . . . . . . . . . . . 47 3.14Simulatedcodes............................ 48 4.1 Conventional payload elements . . . . . . . . . . . . . . . . . . . 57 4.2 Flexible Payload Elements . . . . . . . . . . . . . . . . . . . . . . 60 4.3 Beam-hopping payload elements . . . . . . . . . . . . . . . . . . 64 6.1 Design options comparison . . . . . . . . . . . . . . . . . . . . . . 73 6.2 Obtained transmission time for a packet t a ............ 74 6.3 Lower delay bounds . . . . . . . . . . . . . . . . . . . . . . . . . 74 6.4 Scheduling complexity . . . . . . . . . . . . . . . . . . . . . . . . 75 6.5 Table of MODCODs and users distributions used in the simulations (MODCOD indexes are extracted from Table 2.2) . . . . . 76 6.6 Simulation parameters for the current scenario . . . . . . . . . . 80 6.7 Simulation parameters for the current scenario . . . . . . . . . . 81 6.8 Simulation parameters for the current scenario . . . . . . . . . . 82 6.9 Simulation parameters for the current scenario . . . . . . . . . . 82 6.10 Summary of results for the carried out simulations for 256 bytes link layer frame size . . . . . . . . . . . . . . . . . . . . . . . . . 85 6.11 Summary of results for the carried out simulations for 512 bytes link layer frame size . . . . . . . . . . . . . . . . . . . . . . . . . 86 6.12 Simulation set 1, RS(15,5) and RS(15,7) codes . . . . . . . . . . 87 6.13 Simulation set 2, RS(255,85) and RS(255,127) codes . . . . . . . 88 A.1 System and receiver parameters . . . . . . . . . . . . . . . . . . . 107 A.2 Satellite parameters for conventional payload . . . . . . . . . . . 108 A.3 Satellite parameters for exible payload . . . . . . . . . . . . . . 108 A.4 Satellite parameters for beam-hopping payload . . . . . . . . . . 108 A.5 β constant for the both systems and the three dierent payloads 108 B.1 Undened parameters and values assumed . . . . . . . . . . . . . 110 xiv B.2 Channel parameters . . . . . . . . . . . . . . . . . . . . . . . . . 112 B.3 Channel parameters . . . . . . . . . . . . . . . . . . . . . . . . . 113 B.4 Channel parameters . . . . . . . . . . . . . . . . . . . . . . . . . 113 B.5 Channel parameters . . . . . . . . . . . . . . . . . . . . . . . . . 113 xv Introduction Background of the project This project is part of a European project solicited by EUROCONTROL in the Single European Sky ATM Research (SESAR) programme called ANTARES. The project's prime contractor is Thales Alenia Space Italy and it is driven by several institutions such as Aedel Aerospace, Airtel ATN (IRL), Capgemini Norge (N), Commsonic (GB), Evolving Systems Consulting (CZ), Frequentis (A), Honeywell International (CZ), Indra Espacio (E), Iguassu Software Systems (CZ), IZT (D), Next (I), OHB-System (D), SINTEF ICT (N), Skysoft (P), DLR(D), Space Engineering (I), Syderal (CH), Thales Alenia Space Espana (E), Thales Alenia Space France (F), Thales Avionics (GB), Thales UK (GB), University of Salzburg Scientic Computing (A). The objectives of the project are to develop the communications protocols and satellite system for aviation communication. This satellite system is intended as a backup system for the currently operative beacon based system, but in the end because of advantages in terms of reliability, capacity etc. of this kind of system will, after some years, become the main system. Motivation and Objectives The purpose of this work is the analysis at system-level and a joint optimization at physical/link layer level of a satellite-to-plane link, forward link, of an aeronautical global system. The general design framework is driven by the highly increasing air-trac, which is expected to follow an exponential increase in the near future. Hence, currently distributed ground beacons-based approach may not scale enough and a turn to a rather centralized general approach will be assumed throughout the work. In particular, the modular design covered in the present work has consisted of: system-level architecture at link layer, functional denition, algorithms and protocols at link layer for radio resource management and forward error correction need and solutions at link layer considering physical layer design. xvi It is also presented a study of the co-channel interference in satellite mulitbeam system since ANTARES is a particular multibeam system. The objective has been to provide neat thorough payload-dependant models to aid the overall nal system design. This study is complementary to parallel work within our group on system-modelling and optimization of general multibeam systems with trac constraints. The objectives are to meet all the requirements, mandatory for aviation communication as it is established by the European Civil Aviation Conference (ECAC), whether they are relative to the link budget accomplishment, error rate ratios or qualities of service requirements needed at the analyzed layers in the work, physical layer and link layer. For this reason the methodology used is beginning with an exhaustive analysis of the system, the parts to optimize and its requirements to be able to take the right decisions to improve the overall performance of the system. Then we will simulate the proposed designs to observe if the requirements are met. As it has been mentioned before the ANTARES project is driven by several institutions, regarding to this it must be mentioned that during the development of the work not all the information needed was available as it was being developed by other contractors. This can be seen either as a disadvantage or an advantage since when all the information is available the obtention and analysis of results, and the optimizations tasks are easier. However as this has not been possible we have been pushed to do also and engineering task, suppose suitable values or congurations , analyze the obtained results and iterate with new values if necessary. Structure This work is structured as follows, after this brew introduction, is presented in Chapter 1 the reason for designing the ANTARES system, its operation at system level and its mandatory requirements imposed by the ECAC. Later on, in Chapter 2 are explained the design options for the Radio Resource Management in ANTARES, dening the architectures at system and medium access control level. It is also detailed in the named chapter the physical access to the medium in order to meet the specied requirements. In Chapter 3 is discussed and analyzed the need for using LL-FEC codes in the system and which are the most suitable codes. Chapter 4 explains, from a top view, the dierent payloads used in satellites, dierent payloads will lead to dierent link budget computations that will aect on our interference analysis on Chapter 5. This analysis is carried out in a generalized multibeam scenario and in a baseline option for ANTARES. Finally Chapter 6 presents the simulation results for the Chapter 2, 3 and 5 and it takes conclusion into the results. Chapter 7 concludes the project by xvii resuming and analyzing the work done, as well as indicating future research lines. An appendix, A, is added to give detailed information of some of the computation carried out in the work. Appendix B provides brief guidelines for the LL-FEC implementation in ANTARES. xviii Chapter 1 ANTARES Global System 1.1 Introduction to aviation communications Currently data aviation communications are based on transmitting information to ground beacons using VHF. These ground beacons retransmit the information through the Aeronautical Telecommunications Network (ATN) to its correct destination. ATN Internet Figure 1.1: Aviation communications scheme Ground beacons are distributed along airplane ight routes in order to maintain data communications continuous. However it is expected an exponential increase of the air trac in the next years as well as the aperture of new aerial 1 CHAPTER 1. ANTARES GLOBAL SYSTEM routes. These changes will lead to the increase of ground radio beacons and data trac. Hence the need to implement a backup satellite system for aviation communication in order to manage trac in a more centralized way, increase the capacity of the system, being able to manage the incoming higher volumes of air trac data and also unify all European countries under the same standard. 1.2 ANTARES System and Requirements 1.2.1 Architecture The ANTARES system is basically a three Geosynchronous Earth Orbit (GEO) satellite multibeam system, designed for providing control communications between base stations and aircrafts, and vice versa through a satellite. The architecture of the system is shown in Figure 1.2. Figure 1.2: ANTARES system architecture The elements constituting the system are described here after:  GEO satellites: The three GEO satellites should be enough to provide coverage to ight areas, the use of additional Highly Elliptical Orbit (HEO) is envisaged in order to provide coverage to the poles and remote areas. Each of the GEO satellites is backuped by an additional satellite separated α= 50o to increase the robustness of the system, providing an 2 CHAPTER 1. ANTARES GLOBAL SYSTEM additional link in case of degradation of the main one or substituting the main satellite in case of failure.  Aeronautical Earth Stations (AES): The user segment system is compound of 10 thousand AESs, normally airplanes equipped with omnidirectional antennas on the top of the fuselage.  Network Control Center (NCC): Is in charge of controlling all the satellite networking functions, one NCC per satellite is previewed.  Ground Earth Stations (GES): Provide the interface with the ATN, the trac network for airplanes communications. As a rst approach one GES would be enough to manage all the system, however some Air National Service Providers (ANSP) have demanded specic GES for them, hence the values considered are 1, 3 or 5. The system performs as follows, in the forward link when a GES wants to send information the NCC transmit a forward and a return signaling carrier to the GES and AES to indicate the forward and return trac carriers. The NCC must receive back the signaling information in order to correct errors such as satellite Doppler. Then GES uses the assigned forward trac carrier to transmit information to the AES, and the AES at the same time use the return trac carrier to send information to the GES. While GES and AES are exchanging information a synchronization procedure is carried out between GES, AES and NCC to keep the continuity of the trac carriers. 1.2.2 Types of Links The system will work over two frequency bands, the Ku band and the L band as it is shown in Table 1.1. Fixed Link Mobile Link Uplink Ku band 11.7-12.7 GHz L band 1646.5-1656.5 MHz Downlink Ku band 14-14.5 GHz L band 1545-1555 MHz Table 1.1: Types of Links The xed link comprehends from the GES to the satellite (unique beam structure) while the mobile link comprehends from the satellite to the AES (multibeam structure). 1.2.3 Protocol Stack The protocol stack of the system is shown in Figure 1.3. 3 CHAPTER 1. ANTARES GLOBAL SYSTEM C carriers TSF TS Subset of NUT carriers destined to M1users Subset of NUT carriers destined to MNusers Subset of NUT carriers destined to M2users Carriers for unicast traffic Carriers for multicast traffic Signaling carriers Figure 1.7: UT allocation  Analysis of the architecture of the ANTARES system and its requirements to being able to design the required parts of the system (done in the current chapter).  Designing the Radio Resource Management (RRM) for the selected Medium Access Control (MAC) (Chapter 2).  Analyze whether or not is necessary to use Link Layer Forward Error Correction (LL-FEC) in the system and in case its necessary choose a suitable code and design a preliminary coding option (Chapter 3).  Analysis of the co-channel interference of the system (Chapter 5), to this aim are studied rst the elements that inuence on this eect, which are the satellite payloads (Chapter 4). 10 Chapter 2 Design of the Radio Resource Management for the ANTARES system In this chapter is carried out the RRM design of the ANTARES system. The tasks involving this design are listed here below:  Denition of the types of link between AESs, GESs and NCC in section 2.1.  Denition at MAC level of the entities forming the system and the operations that must perform. Related section is 2.2.  Design of the MF-TDMA frame and its allocation algorithm explained in section 2.3. 2.1 System Level Architecture Design Within this section is specied at system level, the dierent available options for the architecture design, taking into account its advantages and disadvantages. Note that the elements constituting the system were specied in Figure 1.2 but not the interfaces between them. Considerations made for the architecture design are listed here below:  The link between GESs and AESs is always a satellite link.  The link between the NCC and the GESs can either be a satellite or a terrestrial link. 11 CHAPTER 2. DESIGN OF THE RADIO RESOURCE MANAGEMENT FOR THE ANTARES SYSTEM These leads to two basic dierent architectures, an architecture with a satellite based Wide Area Network (WAN), and an architecture with a terrestrial based WAN. Besides for the delay estimation in each of the architectures the following notation is used:  t GES,NCC stands for the propagation time between the GESs and the NCC and can take values either for a satellite or terrestrial link.  t NCC,GES stands for the propagation time between the NCC and the GESs and can take values either for a terrestrial or a satellite link.  t GES,AES stands for the propagation time between the GES and the AES and takes the value for a satellite end-to-end travel (0.25 seconds).  t SCH stands for the total time a packet is being scheduled which includes the amount of time the packet is queued and elapsed in the buer waiting for being transmitted.  t a stands for the amount of time needed to transmit a packet using a certain modulation and codication (MODCOD). It is of course important to reduce the quantity of satellite hops between the NCC and GESs as well as the information sent over these hops. The amount of time introduced in these links is relatively high when compared with any other delay and can not be reduced or optimized, so it will considerably aect to accomplish the QoS delay requirements. 2.1.1 Satellite Based Wide Area Network (SATBASEDWAN) Architecture 2.1.1.1 Distributed scheduling and resource allocation process The basic characteristics of this option, which is depicted in Figure 2.1, are the following:  The communication between the GESs and the NCC is carried out over a satellite link.  Scheduling process is performed at each GES while the resource allocation process is performed in the NCC leading to a distributed architecture. 12 CHAPTER 2. DESIGN OF THE RADIO RESOURCE MANAGEMENT FOR THE ANTARES SYSTEM Figure 2.1: Satellite based distributed architecture Therefore the top level process to send information to the AESs is as explained next:  Based on the incoming messages, each GES requests resources to the NCC.  The NCC performs the radio resource allocation and sends the assigned resources to each GES in a Resource Allocation Table (RAT).  Each GES perform a scheduling policy in order to send the most priority packets to its assigned AESs by using the assigned resources. A delay estimation for sending a packet can be performed as follows: d(s) = max{tSCH, tGES,NCC +tNCC,GES}+ta+tGES,AES (2.1) The term max{tSCH, tGES,NCC +tNCC,GES} stands for the minimum time a packet will be scheduled, which is the time for a GES demanding resources and the NCC assign it to the GES. As under this option all links are satellite based air hops are delayed 0.25 seconds, the total delay estimation turns into: d(s) = max{tSCH,0.5s}+ta+ 0.25s Hence there is always a minimum 0.75 seconds delay introduced by the satellite links. 13 CHAPTER 2. DESIGN OF THE RADIO RESOURCE MANAGEMENT FOR THE ANTARES SYSTEM 2.1.1.2 Centralized scheduling and resource allocation process The basic characteristics of this option, which is depicted in Figure 2.2, are the following:  The communication between the GESs and the NCC is carried out over a satellite link.  Scheduling process and resource allocation process are both performed in the NCC leading to a centralized architecture. Figure 2.2: Satellite based centralized architecture Therefore the process for sending information to the AESs is as follows:  NCC receives messages from the ATN and performs the radio resource allocation process and the scheduling.  The NCC sends the scheduled messages and the RATs to the GESs that retransmit it to the corresponding AESs by using the assigned resources from the NCC. Again we can do delay estimation for sending a packet, which result as follows: d(s) = tSCH + (ta+tNCC,GES)+(ta+tGES,AES) (2.2) 14 CHAPTER 2. DESIGN OF THE RADIO RESOURCE MANAGEMENT FOR THE ANTARES SYSTEM Where the terms ta+tx,y stand for the transmission time of a packet plus the delay of the link from x to y . As we are assuming all links are satellite based, this options leads to a minimum end-to-end delay of: d(s) = 2(ta+ 0.25s) + tSCH Hence there is always 0.5 seconds delay because of the satellite links. 2.1.2 Terrestrial Based Wide Area Network (TERRBASEDWAN) Architecture 2.1.2.1 Distributed scheduling and resource allocation process Under this design we assume the same architecture as in subsection 2.1.1.1; however the GESs-NCC path considered is terrestrial. Proposed architecture can be seen in Figure 2.3. Figure 2.3: Terrestrial based distributed architecture The top-level process to send the information is the same than in the mentioned section, however delays can be highly reduced because of the use of terrestrial links. Now the total delay can be simplied to: d(s) = max{tSCH, tGES,NCC +tNCC,GES}+ta+tGES,AES = (2.3) 15 CHAPTER 2. DESIGN OF THE RADIO RESOURCE MANAGEMENT FOR THE ANTARES SYSTEM =ta+tSCH +tGES,AES Where the term t GES,NCC +t NCC,GES has been neglected since in a terrestrial link these delays are considerably small compared with the rest of the delays. Considering t GES,AES equal to 0.25 seconds we get d(s) = ta+tSCH + 0.25s . Hence the delay introduced because of satellite links is 0.25 seconds. 2.1.2.2 Centralized scheduling and resource allocation process Under this design we assume the same architecture as in section 2.1.1.2; however the GESs-NCC path considered is terrestrial. Proposed architecture is shown in Figure 2.4. Figure 2.4: Terrestrial based centralized architecture The top-level process is the same than in the mentioned section, however delays can be highly reduced because of the use of terrestrial links. Now the end-to-end delay can be simplied to: d(s) = tSCH + (ta+tNCC,GES)+(ta+tGES,AES) = (2.4) =tSCH +ta+tGES,AES 16 CHAPTER 2. DESIGN OF THE RADIO RESOURCE MANAGEMENT FOR THE ANTARES SYSTEM Where the term and t a +t NCC,GES have been neglected since in terrestrial links these delays are considerably small compared with the rest of delays. Considering t GES,AES equal to 0.25 seconds we get d(s) = tSCH +ta+tGES;AES . Hence the delay introduced by the satellite links is just 0.25 seconds. 2.2 MAC-level Architecture In this section we specify at MAC level the blocks that form each of the architectures, the centralized and distributed architectures, whether based on satellite or terrestrial links. 2.2.1 MAC-level architecture for distributed system level architectures The MAC-level scheme for both distributed architectures can be seen here below in Figure 2.5. Radio Resources Allocator Load and Link Budget Monitoring NCC to UTs RAT Terrestrial link or satellite link ATN Information Queuing Manager GES1 Network layer Queuing MAC Queuing PHY Radio Resources Request Manager ATN Information Queuing Manager GESN Network layer Queuing MAC Queuing PHY Radio Resources Request Manager GES2to GESN-1 RAT to UTs Queuing statistics Queuing statistics Requests Requests Figure 2.5: MAC level architecture for distributed option A detailed scheme of how queuing is performed in the three dierent levels, Network, MAC and PHY can be seen in Figure 2.6 here after: 17 CHAPTER 2. DESIGN OF THE RADIO RESOURCE MANAGEMENT FOR THE ANTARES SYSTEM Queuing Buffering Queuing Unicast Traffic Signaling Traffic Multicast Traffic Queuing according to CoS ATN Information Scheduling Policy Buffering according to CoS ANTARES Encoder ANTARES Modulator Queuing statistics to Queuing Manager Requests to NCC RAT from NCC [...] [...] Round Robin Classificator Section 3 Section 1 Section 2 Buffering Information from Queuing Manager Network layer Queuing MAC Queuing PHY Figure 2.6: Detailed scheme of queuing performance In the list below the entities belonging to the GES and its functions are specied:  Incoming trac is divided in three sections, one for unicast trac, one for multicast trac and another one for signaling trac. Regarding to unicast section, trac is divided into eight queues; one per the CoS specied in Table 1.7 and then buered also into eight dierent buers before the scheduler polls them. In multicast and signaling section just one queue and buer is used.  The Queuing Manager gathers information from the queues and communicates this information to the scheduler in order to adapt the scheduling policy.  The Radio Resource Request Manager extracts information from the buers within each section and based on that sends requests to the NCC Radio Resource Allocator.  The scheduler polls the messages from the dierent sections and from the dierent buers within each section by following a scheduling policy. The scheduling policy can be adaptive and vary in function of the information send by the Radio Resource Allocator in the NCC or the Queuing Manager in the GES.  The ANTARES Encoder/Modulator shapes the signal with the correct MODCOD indicated by the Radio Resource Allocator in order to send it to the AESs. 18 CHAPTER 2. DESIGN OF THE RADIO RESOURCE MANAGEMENT FOR THE ANTARES SYSTEM In the list here below the entities belonging to the NCC and its functions are specied:  The Load and Link Budget Monitoring block gathers information in order to indicate to the Radio Resource Allocator information regarding to Adaptive Coding and Modulation (ACM).  The Radio Resource Allocator receives the resource requests from the Radio Resource Requests Manager in each of the GESs of the system and receives information from the Load and Link Budget Monitoring block. Then performs an algorithm to assign the dierent carriers and timeslots to all GESs in a fair way and communicate it through the RAT to each GESs. This RAT is received by GESs at the PHY level. Note that communication between the NCC and the GES can be done either by a satellite or a terrestrial link leading to dierent delays for the same architecture. 2.2.2 MAC-level architecture for centralized system level architectures The MAC-level scheme for both centralized architectures can be seen here below in Figure 2.7. The queuing scheme is the same seen in Figure 2.6 with the only dierence that after scheduling information is sent to each GES where it is modulated and encoded to send to the UTs in the AESs. Radio Resources Allocator Load and Link Budget Monitoring NCC to UTs Terrestrial link or satellite link ATN Information Queuing Manager Network layer Queuing MAC Queuing PHY Queuing statistics Requests RAT ANTARES Encoder ANTARES Modulator GES1 ANTARES Encoder ANTARES Modulator GES2 ANTARES Encoder ANTARES Modulator GESN to UTs to UTs Radio Resources Request Manager Figure 2.7: MAC level architecture for centralized option 19 Chapter 3 Analysis of the Erasure Channel and Proposed Solution 3.1 Introduction to LL-FEC 3.1.1 LL-FEC concept Forward Error Correction (FEC) solutions above the physical layer are proposed any time it is necessary to overcome the losses resilient to physical layer FEC mechanisms (and link layer retransmission schemes, if any). The concept is as follows. At the receiver side, each packet is considered by the link layers either completely received or completely lost usually based on a CRC (Cyclic Redundancy Check). This way, it can be considered as if there is a virtual erasure channel. Dierent upper layer FEC solutions are available in several specications for dierent applications. Current standards include solutions that are partly integrated in the application layer above the Network level, then referred to as Application Layer FEC (AL-FEC), or in the link layer (LL) below the Network level, then referred to as Link Layer FEC (LL-FEC) which will be the case under study. 3.1.2 From Information Theory In digital communications the transmitter sends one out of a nite set of codewords and the receiver would like to know which codeword has been transmitted. A general scheme of an end to end communication system can be seen just in Figure 3.1. 26 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION EncoderEncoder P(y|x)P(y|x) DecoderDecoder iϵ{0,1,…,c-1} xi=(xi[1], xi[2],...,xi[N]) y=(y[1], y[2], ..., y[N]) î Figure 3.1: End to end communications system The input and output, x[m] and y[m] , respectively lie in a nite set which are called the input and output alphabets, x∈X and y∈Y . The Binary Erasure Channel (BEC) has a binary input x= 0,1 , and a ternary output, y= 0,1, e , where e denotes the erasure. The transition probabilities, p(j|i)|j∈Y,i∈X , are depicted in Figure 3.2. As we can see the bits cannot be ipped as in the Binary Symmetric Channel (BSC), but can be lost or erased. 0 1 0 1 e 1-ε 1-ε ε ε Figure 3.2: Probability transitions of the BEC We can calculate the maximum capacity of the BEC channel as follows: C=maxp(x)I(X;Y) = maxp(x)H(Y)−H(ε) Letting E be the event Y=e and using the expansion: H(Y) = H(Y, E) = H(E) + H(Y|E) And letting p(x=1)= π , we have: H(Y) = ((1 −π) · (1 −ε), ε, π(1 −ε) = H(ε) + (1 −ε)H(π) Hence: C=maxp(x)H(Y)−H(ε) = maxπ(1 −ε)H(π) + H(ε)−H(ε) = 27 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION =maxπ(1 −ε)H(π) = 1 −ε This maximum is achieved when π=1 /2 . Intuitively we can see that as an epsilon proportion of the bits are lost in the channel (the erased ones), we can only recover up to one minus epsilon bits, hence the capacity is one minus epsilon.[20] 3.2 Justication of the LL-FEC use in ANTARES 3.2.1 Link Layer channel vs Physical Layer Channel As it is mentioned in the introduction, FEC can be used at dierent levels, normally at the physical layer, link layer or application layer. The use of FEC at a certain level does not exclude to use it at another level since its purposes are dierent. Therefore in this section we identify what type of errors LL-FEC codes can deal with in front of physical layer-FEC (PHY-FEC). In further subsections within this section we analyze whether in any of the ANTARES scenarios is given a situation for the LL-FEC use. So PHY-FEC is intended for correcting errors at bit level produced within a physical layer frame: FEC/CRC Correct /Detect errors Erroneous bit Physical layer frame L2 packet L2 packet L2 packet [...]LINK LAYER PHY LAYER Figure 3.3: PHY-FEC correcting/detecting errors idea Those errors can be produced by fast/medium term variations of the channels such it can be Rayleigh or Rice eects. However it is possible that under certain conditions a channel produced the eect shown in Figure 3.4, where we obtain a burst of erroneous bits so as several LL packets are lost. 28 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION FEC/CRC Cannot Correct/Detect errors Burst of erased bits Physical layer frame L2 packet L2 packet L2 packet [...]LINK LAYER PHY LAYER Discarded L2 packet Discarded L2 packet Figure 3.4: Burst erasure scenario Such burst of erroneous bits makes the PHY-FEC/CRC correction or detection useless and those frames would be simply discarded by the decoder. To overcome this problem FEC is used at link layer level as follows: Physical layer frame L2 packet L2 packet LL-FEC [...]LINK LAYER PHY LAYER LL-FEC Correct L2 packets Bursts of erased bits Figure 3.5: LL-FEC idea Where the fact that each of the FEC packets are calculated from dierent L2 packets makes possible to correct the burst of erasures. The specic way this is done is explained in subsection 3.5.3. Hence if we demonstrate the ANTARES aeronautical channel behaves as an erasure channel, the use of LL-FEC to correct errors is almost straightforward . The only drawback for the LL-FEC is the delay it introduces which must be under the ANTARES QoS requirements specied in Table 1.7. The purpose of the following subsections is:  Analyze if the ANTARES aeronautical channel behaves as a burst erasure channel and under which cases. 29 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION  Characterize these cases in order to design properly the LL-FEC codes to use. 3.2.2 Aeronautical Erasure Channel Models In this section we analyze two potential link layer erasure aeronautical channel models, one for commercial aviation and another one for helicopters. The erasures for each scenario are generated as follows:  For commercial aviation erasures can be produced due to:  Other airplanes in the line of sight between the satellite and the airplane intending to receive the communication.  By the airplane itself when maneuvering. For the rst case no statistics have been found on the literature. For the second case it is demonstrated in the literature that it may so happen that maneuvers of the airplane make the elevation angle with the satellite negative. First order statistics are available on the literature, but not second order statistics.  For the helicopter scenario erasures can be produced basically by two eects:  The blades of the helicopter when rotating over the antenna  When the helicopter is ying along a building located between the line of sight with the satellite. 3.2.2.1 Erasure Model for Commercial Aviation In this section two cases of erasure channel models with airplanes are explained, in the rst case the erasures are produced by an airplane blocking the satellite signal to another airplane and in the second case the erasures are produced because of the airplane itself. Erasures produced by airplanes in the same line of sight with the satellite The rst example of erasure channel model with airplanes is depicted in this section. Understanding the aeronautical channel as a mobile multiuser environment we nd feasible that burst erasures could be produced by another airplane positioned in the same line of sight with the satellite during certain time as shown in Figure 3.6. Due to the fast increase of air trac, the congestion in some aerial routes and airports this eect is reliable to happen in airplanes in en route phase and in airplanes maneuvering or landing. In Figure 3.7 we can see 30 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION an image of the aerial routes in a zone of Spain, extracted from the national air navigation service provider, AENA, and see how routes cross each other often. θvx1 vx1 vy1 vy1 vn1 vy2 300m Figure 3.6: Burst erasure channel scenario There are an innite number of possibilities for the way an airplane could block the signal to another airplane, depending on the relative speed of both planes, the angle between them and the position of the satellite. The minimum regulated distance between the two airplanes traveling at dierent heights is 300m. Figure 3.7: Catalonia aerial routes As shown in Figure 3.6, we consider the distance the airplane travels with its line of sight with the satellite blocked by another airplane, x, as a function of θ . Airplanes can travel in the same or in dierent directions. In the same gure v n indicates the nominal speed of the airplane, and v x and v y the x and y speed components respectively. All speeds are expressed in m/s . Note that this is a very specic case since we are assuming both planes are 31 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION traveling without any change on its altitude ight, without maneuvering, at constant speed and no variations in the angle θ are produced. Airplane Length (m) Landing/Maneuvering En route speed (m/s) speed (m/s) Airbus 60 83.3 245 Boeing 50 83.3 245 Regional 30 83.3 240 Cargo plane 20 83.3 240 Table 3.1: Standard airplane lengths As it is shown in Table 3.1 we have considered some dierent speeds that lead to two dierent cases:  One landing or maneuvering airplane at 83.3m/s is having the satellite signal blocked by an en route airplane at 240m/s-345m/s at a higher altitude.  One en route airplane at 240m/s is having the satellite signal blocked by another en route airplane traveling higher and at a 245m/s. We can nd the relative speed, v relative , in m/s between the two airplanes by doing: vx1=vn1 · sin(θ), vy1=vn1 · cos(θ) vx2= 0, vy2=vn2 vrelative =q(vy1±vy2)2+ (vx1±vx2)2 Where the operator + is applied when the airplanes travel in the same direction and the operator - is applied when travel in opposite directions. Then we can nd the time in seconds one airplane blocks the other one by: t(s) = x(m) v(m/s)=x cos(θ) · vrelative (3.1) Where x is the length of the aircraft. Signal blockage produces erasures consisting of bursts of packets. Minimum, mean and maximum values of duration of the erasure are given in Table 3.2. As a rst approach, Table 3.3 and Table 3.4 below show packets lost in the worst, mean and best case erasure for dierent 32 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION Minimum duration Mean duration Maximum duration 0.04 0.3 1.15 Table 3.2: Minimum, mean and maximum values of erasure duration bitrates (16kbps, 32kbps and 64kbps) and for dierent packet sizes (4096 and 8192 bits) in the forward downlink. Packet Size Packet Loss Case R b @16kbps R b @32kbps R b @64kbps PL min 111 4096 bits PL mean 236 PL max 5 10 18 Table 3.3: Number of packets lost at each bitrate 4096 bit frame length Packet Size Packet Loss Case R b @16kbps R b @32kbps R b @64kbps PL min 111 8192 bits PL mean 123 PL max 359 Table 3.4: Number of packets lost at each bitrate 8192 bit frame length Erasures produced by the airplane itself when maneuvering View angle statistics to an Immarsat I-4 GEO satellite have been obtained in reference [12] from simulated ights using a database of 8 thousand scheduled ights. The Cumulative Density Function (CDF) of the elevation angle in the Terminal Maneuvering Area (TMA) showed the results in Figure 3.8. Figure 3.8: Elevation angle CDF and zoomed region respectively The negative elevation angles show the case where the satellite view falls under the aircraft horizontal plane, these would cause blockage of the satellite signal. 33 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION As we can see in the gure in the TMA zone this can happen with an approximate probability of 2%. The P E,min values mean the probability of the aircraft to perform a maneuver resulting into a minimum elevation angle E min . This E min is computed from some pitch and roll angles of the airplane. P E,min is taken as a known parameter. As said, the literature does not provide second order statistics and we would need realistic data on maneuvering erasures to obtain how many packets would be erased. 3.2.2.2 Erasure Model for Helicopters Another example of erasure channel model happens in an helicopter aeronautical scenario. Here, we can distinguish two cases:  Erasures produced by the helicopter blades when rotating over the antenna.  Erasures produced when the receiving signal could be blocked by a building in the Line of Sight (LOS) of the helicopter. Both scenarios are illustrated in Figure 3.9. Standard helicopter blade measures have been chosen, 0.5 meters width and 6 meters length, while the revolution speed of the blades can be comprehend between 300-500 Revolutions Per Minute (RPM) and the speed of the helicopter can be comprehend between 150km/h and 250km/h. 400-500 RPM Vx=150km/h-250km/h 400-500 RPM Vx=150km/h-250km/h Figure 3.9: Helicopter scenarios, erasures produced because of the blades an erasures produced because of a building respectively Erasures produced by the blades of the helicopters Under this scenario we assume the antenna is under the helicopter blades and the RPM of the blades can be between 300RPM and 500RPM. The amount of time in seconds the blades block the LOS signal is shown in Table 3.5. 34 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION RPM Time (s) 300 0.160 400 0.0751 500 0.061 Table 3.5: Time in seconds the signal is blocked by helicopter blades We can trace an imaginary route of a helicopter between two points without any buildings blocking the signal, then the erasures produced only depend on the RPM of the blades. In Figure 3.10 and Figure 3.11 we can observe how many packets per per blade blockage (each time a blade rotates over the antenna) would be erased for 4096 bit and 8192 bit packet length and for three dierent bitrates (16kbps, 32kbps and 64kbps). Figure 3.10: 4096 bit length packets lost in function of the RPM of the blades for dierent bitrates 35 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION although at higher RPM the erasures are produced more often, its duration is lower. Figure 3.20: First order statistic: Probability of erasure Figure 3.21: Second order statistics: CDF of the duration of the erasure for the 150km/h 300RPM, 190km/h 400RPM, 230km/h 400RPM and 250km/h 500RPM from top to bottom and right to left In Figure 3.21 we can observe the probability that the duration of the erasure, 42 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION x, is lower than a certain level, F(x), this is the empirical CDF of the duration of the erasure. The gure shows that short erasures are produced almost always because while long erasure are barely produced. This happens because short erasures are produced because of the rotation of the blades which are a periodical eect while long erasure are produced by buildings which is a low-probability eect. This is also the reason why the graphic is so abrupt. No big dierences are noticed between the dierent pairs of speeds/blade rotation. 3.3 Design Options The basic idea in LL-FEC codes is that K blocks of source data are encoded at the transmission to produce N blocks of encoded data, such code is called an (N,K) code. An (N,K) code is able to correct up to N-K losses in a group of N encoded blocks. Several codes exist for this purpose, here below we list some of them together with a brief description and we give performance results on real applications for each of them. 3.3.1 Reed Solomon (RS) Codes The basic characteristic of RS codes is that they are MDS codes. Despite of its simplicity and attractiveness RS codes have some disadvantages; the encoding/decoding complexity grows up exponentially with the codeword size. To keep this complexity manageable several restrictions to the codeword length must be imposed [3]. RS codes have been adopted as the standard codes for the rst DVB generation. Simulation in reference [6] for a typical Rayleigh channel and in [9] for a burst noise channel shows that RS standalone codes are outperformed by both Raptor and LDPC codes respectively. However this result is not directly extrapolable to the ANTARES case because the channel it is not demonstrated to be an erasure+noise channel. 3.3.2 RS+Channel Interleaver This solution is proposed to mitigate the fact that the encoding/decoding complexity of RS codes with long codewords is unmanageable. It is shown in [4] that for a pure long burst erasure channel without added noise the optimal solution is using block interleaved RS codes, moreover this optimality does only depend on the length of the block interleaver. The basic idea scheme is shown in Figure 3.22. Several comparisons on the long burst erasure channel are carried out in this reference with the following summarized results:  For a xed N and K, the eciency can be made as close to one as desired by simply increasing the interleaving length suciently. 43 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION (N,K) RS encoder Depth I Channel Interleaver Burst Erasure Channel Depth I Channel de-Interleaver (N,K) RS decoder data data Figure 3.22: Basic idea of channel interleaved RS codes  Very short RS codes are less than 1% inecient at many lengths of practical interest.  In channels modeled as erasure+noise (not our assumed case), RS interleaved codes are outperformed by LDPC codes. 3.3.3 Low Density Parity Check (LDPC) codes LDPC are linear error correcting codes, although they are not MDS codes permit the use of longer codewords lengths than in RS with linear complexity increasing, providing much better performance. If well designed LDPC codes can be asymptotically (in N) MDS and can provide protection to long burst erasures if interleaved [3]. In references [8] and [9] it is proved that LDPC codes outperform RS and interleaved RS codes for burst erasures channels with added noise, however in pure burst erasure channels interleaved RS codes are the best option. As said before is still not clear if the ANTARES channel behaves as an erasure plus noise channel. 3.3.4 Raptor Codes Raptor codes are a class of Fountain codes based on the Luby Transform. Compared with RS codes provide more exibility, larger code dimensions, and lower decoding exibility. Another advantage is that show linear time encoding/decoding with N, however Raptor codes are not MDS. Raptor codes have been lately introduced in the DVB standards [6]. Also in [6] a comparison between RS codes and Raptor codes is carried out. The study was made for a typical correlated Rayleigh channel using both Multi Protocol EncapsulationForward Error Correction (MPE-FEC) and MPE-Interburst FEC (MPE-IFEC) and for dierent modulations and codications (MODCODs). The obtained results are summarized in Table 3.7. 44 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION Raptor outperformance over RS MODCOD MPE-FEC MPE-IFEC QPSK, 1/2 0.7dB 1.5dB 8PSK, 3/4 0.1dB 2.7dB Table 3.7: Raptor over RS outperformance for a PER@1e-4 at v=100km/h The results also showed that the best framework is MPE-IFEC which outperforms MPE-FEC for every MODCOD in RS and Raptor codes. The best overall result is the MPE-IFEC framework with Raptor codes. 3.3.5 Other Codes Also more variants for packet level coding exists like Irregular Repeat and Accumulate (IRA), Generalized IRA (GeIRA), Tornado and Protograph codes shortly explained in [3]. 3.4 Comparative Performance Analysis Under this section is carried out a comparative between the LL-FEC design options depicted in 3.3 in order to choose the most suitable code for the LL-FEC implementation in ANTARES. To this aim we will compare the capacity of the codes to solve a burst of erasures (eciency) and the diculty to encode/decode the information of each of them (complexity in arithmetic operations). 3.4.1 Information from literature review 3.4.1.1 Eciency Regarding to code eciency, in reference [4] a thorough study is presented to compare the eciency in burst erasure channels of several types of codes. This eciency is measured as the ratio of the maximum resolvable length burst, L max , divided by N-K , being N the encoded bits and K the original source bits. N-K is the maximum resolvable erasure length indicated by the Singleton bound. For an RS code, interleaved or virtual interleaved the eciency can be measured as: η=(N−K) · I · m−m+ 1 (N−K) · I · m= 1 −m−1 m · I · N(1 −r) (3.2) Where r is the code rate K/N , m is the number of bits per symbol that in RS must accomplish that 2 m -1=N and I is the block interleaving depth. From the 45 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION expression above in the form η= 1−µ it can be seen that µ=m−1 m · I · N(1−r) is the ineciency of the code. For RS we can also dene the total transmission length in bits of an interleaved code as: Tranmission length =N · I · m (3.3) This is the total number of bits sent during all the interleaving. Note that if I is set to one we obtain the number of encoded bits of the code. For LDPC codes nding L max must be done by applying the algorithm in reference [7] to the parity check matrix of the LDPC codes. In this same reference and in [13] are provided results for dierent types of LDPC codes such as MacKay, extended Irregular Repeat and Accumulate (eIRA), array, Euclidean Geometry (EG) or Progressive Geometry (PG), under the burst erasure channel. These results will be used for an eciency code comparison. For Raptor codes eciency no results in the literature have been found for the burst erasure channel. 3.4.1.2 Complexity With reference to the complexity issue, it is well-known that the major drawback of RS codes is the exponential encoding and decoding complexity increase in terms of operations. The complexity on RS encoding/decoding can be expressed in terms of Galois Field Multiply and Accumulate (GFMAC) operations that the encoder/decoder performs for computing the RS codeword. Complexity performance is written in function of N , K , and T=(N-K)/2 . For the encoding process just one option is possible, while for the decoding process several algorithms are available. Encoding algorithm Operations - O(N · 2T) Table 3.8: RS encoding algorithm complexity Decoding algorithm Operations Syndrome computation O(N · 2T) Berlekamp-Massey O(2T · T) Chien Search O(N · T) Forney Algorithm O(T · T/2) Table 3.9: RS decoding algorithm complexity As we can see in Table 3.8 and 3.9, complexity is basically a function of T=(NK)/2 , i.e. a function of the redundancy added to the code. The lower the term T is, for a xed N,the bigger is K; and lower is the encoding/decoding complexity. 46 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION For LDPC codes also exist several encoding/decoding algorithms, which in general require less arithmetic operation than in RS. Some of those algorithms are listed in Table 3.10 and in Table 3.11 and extracted from references [14] and [15]. Encoding algorithm Operations LDPC 1 O(2q) LDPC 2 O(2(K · (K-q))) Table 3.10: LDPC encoding algorithm complexity Decoding Algorithm Operations Belief Propagation O(2 · q) Extended Mini Sum O(q · log(q)) Table 3.11: LDPC decoding algorithm complexity Being q the number of codewords of the LDPC code, i.e. for an LDPC code (N,K), q is N-1 . With regard to Raptor codes the most common used encoding/ decoding algorithms are shown in Table 3.12 and in Table 3.13 and extracted from reference [5]. Encoding algorithm Operations - O(K) Table 3.12: Raptor encoding algorithm complexity Decoding algorithm Operations - O(K) Table 3.13: Raptor decoding algorithm complexity 3.4.2 Simulations on comparative analysis 3.4.2.1 Eciency simulations To obtain comparison between LDPC, Raptor and RS codes regarding to the gure of merit of the eciency, the codes listed in Table 3.14 have been used. Note that several types of LDPC codes have been simulated, as well as several r=K/N ratios and lengths of the encoded block N in order to obtain a fair comparison with the RS codes. In order to show the results in a more understandable way, 47 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION in the graphics are plotted the ineciencies of the codes in function of the total transmission length dened as in equation (3.2) and equation (3.3) respectively. Code (N,K) L max (bits) LDPC MacKay 4095,3358 515 LDPC EG1 4095,3367 376 LDPC eIRA1 4095,3367 507 LDPC eIRA2 4550,4096 300 LDPC Array1 4095,3367 500 LDPC Array2 4550,4096 299 LDPC EG2 255,175 70 LDPC EG3 1023,781 157 LDPC PG1 273,191 75 LDPC PG2 1057,813 124 RS 255,85 Dependent on I RS 255,127 Dependent on I RS 15,5 Dependent on I RS 15,7 Dependent on I Table 3.14: Simulated codes From the curves in Figure 3.23 we can extract the following conclusions:  Note that for any interleaving length of an RS code, including non-interleaving ( I=1 ) RS codes are less inecient that any of the LDPC codes in the graphics. Hence in terms of solving errors in burst erasure channels; RS codes are more ecient than LDPC codes .  We can observe that for a same interleaving length, a shorter RS code is less inecient.  Codes with lower coderate, r , are less inecient for the same N length. 3.4.2.2 Complexity simulations In order to show the encoding complexity two length N codes have been chosen, N=15 and N=255 . For each of these lengths the encoding complexity is plotted for the several algorithms available per code in function of T=(N-K)/2 . This allows us to see the dependence of the complexity algorithm with T and K . The following conclusions can be extracted from Figure 3.24:  Raptor and LDPC encoding algorithms present the lowest encoding complexity. Besides, LDPC 1 algorithm presents a linear encoding complexity increase with N, which is desirable. 48 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION 101102103104105106107108109 10−8 10−7 10−6 10−5 10−4 10−3 10−2 10−1 100 Transmission length in bits Inefficiency RS(255,127) RS(255,85) RS(15,7) RS(15,5) MacKay EG1 eIRA1 Array1 eIRA2 Array2 EG2 EG3 PG1 PG2 Figure 3.23: Ineciency comparison vs total transmission lengths in bits for pure burst erasure channels Figure 3.24: Encoding complexity at N=15 and N=255 respectively for several algorithms  RS codes present two drawbacks when encoding, complexity increases when increasing T , i.e. the redundancy of the code. Complexity increases exponentially when increasing N as we can see when comparing both graphics, the increase in N from 15 to 255 is not the same increase in number of operations. For the decoding complexity we will compare rst the dierent available algorithms for the RS codes at N=15 and N=255 as a function of T . After that we will compare the least complex algorithms. From Figure 3.25 the following conclusions can be observed:  Forney algorithm presents the best performance, since its complexity increase with T is very slow and its complexity increase with N it is no as 49 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION abrupt as the other algorithms (although it is still exponential).  Berlekamp Massey algorithm presents a similar performance to Forney algorithm however its complexity increase with T is higher. Figure 3.25: Decoding complexity for several RS algorithms at N=15 and N=255 respectively So, in order to compare the RS decoding complexity with the complexity of other codes we will take as reference the Forney algorithm. Figure 3.26: Decoding complexity at N=15 and N=255 for several algorithms So a similar behavior to the encoding can be observed in Figure 3.26, specically:  LDPC BP and EMS algorithms present very low complexity and almost a linear increase with N and remain constant when only K varies.  We can see that for short lengths of the code, Forney algorithm has very low decoding complexity (lower than LDPC codes).  Raptor codes present low complexity. 50 CHAPTER 3. ANALYSIS OF THE ERASURE CHANNEL AND PROPOSED SOLUTION 3.4.3 Conclusions on code comparative  In terms of eciency, as it can be seen in Figure 3.23, RS codes, whether interleaved or not, clearly outperform the rest of the codes.  From the complexity point of view LDPC are the best codes, however it is shown in Figure 3.25 that some some RS decoding algorithms such as Forney algorithm reduce considerably the number of operations to perform and in Figure 3.26 that for short lengths have a similar performance to LDPC algorithms. Being known that, if the channel behaves as a pure erasure channel, we recommend using interleaved/virtual interleaved short length RS codes . The major drawback would only rely on the encoding complexity. Figure 3.27 helps to clear out this issue where we can see the RS decoding complexity for a short code (N=15) is very low while the major drawback relies on the encoding. Figure 3.27: Encoding/Decoding comparison for short LDPC, Raptor and RS codes However in Figure 3.28 we can see this drawback is compensated by much higher eciency of the code, two order dierence, for the range of total transmitted bits, from 1e3 to 1e4 depending on the delay-aware introduced. The region is marked with a red ellipse. 51 CHAPTER 4. PAYLOAD DEPENDENT SYSTEM MODEL Polarization Splitter Uplink Filter L N A Splitter If another polarization is used DOCON TYPE A DOCON TYPE A DOCON TYPE B DOCON TYPE B Channel Filter Channel Filter Channel Filter Channel Filter LTWTAs Groups OFLTR Diplexer RTN Rx FWD Tx IDMUXes xfRfactor LTWTAs Groups OFLTR Diplexer RTN Rx LTWTAs Groups OFLTR Diplexer RTN Rx LTWTAs Groups OFLTR Diplexer RTN Rx Rx Section xGW FWD Tx FWD Tx FWD Tx DOCON chains Ins & OUTs: xGW and POL. 2 for 1 xfRfactor LTWTAs groups Ins & OUTs: Depending on fRfactor Number: Depending on fP xfRfactor xfRfactor xfRfactor | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Figure 4.2: Conventional forward link payload 4.1.3 Payload Architecture 4.1.3.1 Forward Link Architecture After uplink signal ltering of each polarization output, the antenna elements are connected to a 2 for 1 redundant LNA to amplify the signal introducing as less noise as possible. Depending on the frequency plan, f P , more than one type of DOCON could be needed, so the splitter performs the action of sending the signal to the correct DOCON. Then the DOCONs down converts each of the frequency segments of each polarization FWD uplink to the frequency segments of the FWD downlink. Depending on the number of gateways and the number of polarizations the number of inputs and outputs for the DOCONs changes. After the frequency down-conversion the IDMUXes separates the channels assigned to each user link beam, at least are needed as many IDMUXes as the frequency reuse factor, f R factor. Groups of LTWTAs are used to provide the nal amplication of the channels and OFLTRs are used to limit the intermodulation and harmonics high amplication eects. An scheme of this payload can be seen in Figure 4.2. 58 CHAPTER 4. PAYLOAD DEPENDENT SYSTEM MODEL Diplexer Diplexer Diplexer FWD Tx FWD Tx FWD Tx FWD Tx RTN Rx RTN Rx RTN Rx RTN Rx Uplink Filter L N A 2 for 1 Channel Filter Channel Filter Channel Filter Channel Filter IMUXes xfRFactor Channels depending on fP DOCON TYPE A DOCON TYPE A DOCON TYPE B DOCON TYPE B IMUXes Second stage if necessary depending on the fP DOCON chains Ins & OUTs: xGW and POL Types depending on fP LTWTAs Groups OFLTR Polarization Splitter If another polarization is used Uplink Filter L N A 2 for 1 Uplink Filter L N A 2 for 1 L N A 2 for 1 Reception Chain xfRfactor Reception Chain xfRfactor Reception Chain xfRfactor Reception Chain xfRfactor | | | | | | | | | | | | | | | | | | Diplexer Uplink Filter Figure 4.3: Conventional return link payload 4.1.3.2 Return Link Architecture In the RTN link after uplink ltering, each output of the receive elements is connected to a 2 for 1 redundant LNA, then the IMUXes combine the channels in order to accomplish the frequency plan for the RTN downlink, as many IMUXes as the frequency reuse factor are needed. Then the DOCON downconverts the frequency RTN uplink segments into the RTN downlink frequency segments. Depending on the frequency plan the DOCONS chains needed could vary and a second stage of IMUXes could be needed. Also depending on the frequency plan more than one type of DOCON could be needed. LTWTAs are used to high amplify several channels and OFLTRs are used to limit the eect of intermodulation in the adjacent channels and to limit the harmonics. At last the lters can be connected to two polarization transmit ports if needed. An scheme of this payload can be seen in Figure 4.3. 59 CHAPTER 4. PAYLOAD DEPENDENT SYSTEM MODEL 4.2 Flexible Payload Model 4.2.1 Payload Elements The conventional payload elements for both forward FWD and return RTN links are referred in Table 4.2. Forward Link Return Link LNAs LNAs DOCONs DOCONs Intermediate Frequency Filters (IFLTRs) IFLTRs On Board Processor (OBP) Return Link OBP Up-Converters (UPCONs) UPCONs Channel Filters (CFLTRs) CFLTRs Hybrid Matrix Power Amplier (HMPAs) LTWTAs OFLTRs OFLTRs Table 4.2: Flexible Payload Elements 4.2.2 Antenna Design In the exible payload are used separated transmission/reception antennas with antenna Array Fed Reector (AFR) conguration, so each beam is generated using a determinate number of elements of the array. As the feed horns are too smalls one antenna for the transmission and one antenna for the reception is needed. The main advantage of AFR is that the number of apertures can be reduced with respect to the SFBN conguration. 4.2.3 Payload Architecture 4.2.3.1 Forward Link Architecture In the forward link the signal follows the next process, rst each polarization output is low noise amplied by the LNA, then the DOCONS down-convert the received signals to the C-band frequency used by the OBP, the IFLTRs after the DOCONs limit the out of band spurious emissions. The signals inputs into the OBP, which performs the following actions:  Spectral isolation of the individual modulated user channels that compose each FDM multiplexed, multicarrier, gateway signal.  Routing and steering of the complex samples that compose the uplink carriers signals received on FWD uplink to the destined FWD downlink 60 CHAPTER 4. PAYLOAD DEPENDENT SYSTEM MODEL Polarization Splitter Uplink Filter L N A Splitter If another polarization is used Rx Section xGW 2 for 1 DOCON chains INs & OUTs: xGW and POL Types depending on fP DOCON DOCON IFLTR IFLTR OBP UPCON UPCON CFLTR HMPA OFLTR Diplexer … RTN Rx FWD Tx xN elements of the AFR UPCON chains INs & OUTs: xGW and POL IFLTR UPCON UPCON CFLTR HMPA OFLTR Diplexer RTN Rx | | | | | | | | | | | | | | | | | | | | | | | | | | | xGW and POL xGW and POL xGW and POL xGW and POL xGW and POL xGW and POL HMPAs INs & OUTs: xGW and POL OBP INs & OUTs: xGW and POL FWD Tx | | | Figure 4.4: Flexible forward link payload Digital Beam Forming Network (DBFN) in order to generate the subsequent FWD downlinks.  Spatial ltering of the complex samples that compose the uplink carriers signals to generate the subsequent constituent beam signals to be applied to the antenna elements.  Frequency synthesis of the spatially ltered element beam signals to generate the FDM multiplexed, multicarrier element signal to be applied to each of the antenna elements that compose the transmission antenna array. The signals from the output of the OBP are then upconverted with the UPCONs to the downlink frequencies and channel ltered to limit the out of band spurious emissions with the CFLTRs. HMPAs composed of LTWTA are used to amplify the signals that feed the antenna elements and before are transmitted the signals are ltered to limit the noise in the receive frequency band and to limit the spurious emissions with the OFLTRs. An scheme of the FWD link exible payload can be seen in Figure 4.4. 61 CHAPTER 4. PAYLOAD DEPENDENT SYSTEM MODEL 4.2.3.2 Return Link Architecture In the return link the following process is performed, rst the uplink output of each receive element is connected to a 2 for 1 LNA to low noise amplify, then this signal is down-converted through the DOCONs to get the C-band used by the return link OBP and ltered with IFLTRs to limit the out of band spurious emissions. The OBP for the return link performs the following actions on the signal:  Spectral isolation of the frequency segments that compose the FDM multiplexed, multicarrier, co-channel RTN uplink signal derived from the output of each of antenna elements.  Spatial ltering of each spectrally isolated frequency sub-band to generate the corresponding constituent beam signals.  Spectral isolation of the individual user or channel bands assigned to each RTN downlink beam.  Routing and steering the complex samples that compose each individual user channel received on the RTN uplink to the RTN downlink.  Frequency synthesis of the spatially ltered user beam to generate the FDM multiplexed multicarrier gateway signals. The signal from the output of the OBP is up-converted with the UPCONs to get the RTN downlink frequency and CFLTRs are used to limit out of band spurious. In this case LTWTAs are used to amplify the full frequency band of one polarization and output ltered with OFLTRs and then ready to be sent to the antenna feed elements. An scheme of this payload can be seen in Figure 4.5. 4.3 Beam Hopping Payload Model 4.3.1 Payload Elements The elements of the beam hopping payload for the FWD and RTN link are listed in Table 4.3. 62 CHAPTER 4. PAYLOAD DEPENDENT SYSTEM MODEL OBP Diplexer FWD Tx Uplink Filter L N A 2 for 1 Reception Chain: xfRfactor DOCON DOCON IFLTR IFLTR UPCON UPCON UPCON chains INs & OUTs: xGW and POL LTWTAs Groups IFLTR OFLTR Diplexer If another polarization is used RTN Tx # of chains: Depending on the number of beams and the AFR antenna Diplexer FWD Tx Uplink Filter L N A 2 for 1 Reception Chain: xfRfactor DOCON DOCON IFLTR … | | | | | | | | | | | | | | | | | | | | | | | | xGW and POL xGW and POL xGW and POL LTWTA groups INs & OUTs: xGW and POL OBP OUTs: xGW and POL DOCON chains: INs & OUTs: xfRfactor xfRfactor | | | | | | Figure 4.5: Flexible return link payload 63 CHAPTER 4. PAYLOAD DEPENDENT SYSTEM MODEL Forward Link Return Link LNAs LNAs DOCONs DOCONs IFLTRs IMUXes OBP UPCONs CFLTRs HMPAs LTWTAs OFLTRs OFLTRs Table 4.3: Beam-hopping payload elements 4.3.2 Antenna Design The antenna used in this payload is the same used in the conventional one, combined transmission and reception antennas are used in the satellite, with a SFBN antenna conguration using 4 apertures. 4.3.3 Payload Architecture 4.3.3.1 Forward Link Architecture In the forward link the process is as follows, the signals goes through the 2 for 1 LNAs to amplify it, then is down-converted with the DOCONs to the OBP C-band and the IFLTRs limit the out of band spurious emissions. In the OBP the following actions are performed:  Spectral isolation of the individual, phase modulated carriers signals that constitute each FDM multiplexed, multicarrier gateway signal.  Grouping the carriers received on the FWD uplink into FWD downlink sets.  Frequency synthesis of the FWD downlink carrier sets to generate the sub-sequent FDM multiplexed, multicarrier signals. These synthesized multicarrier signals are identied as beam-hopping signals.  Application of the beam hopping signals to the antenna elements. The signal at the output of the OBP is upconverted with the UPCONs from the OBP C-band to the FWD downlink frequency, channel ltered and amplied with the HMPA matrix. At last before sent to the antenna feed elements is ltered to limit noise in the receive frequency and to limit harmonic distortion with OFLTRs. The scheme of the FWD link beam-hopping payload can be seen in Figure 4.6. 64 CHAPTER 4. PAYLOAD DEPENDENT SYSTEM MODEL Polarization Splitter Uplink Filter L N A Splitter If another polarization is used Rx Section xGW 2 for 1 DOCON chains INs & OUTs: xGW and POL DOCON DOCON IFLTR O B P IFLTR UPCON UPCON CFLTR HMPA OFLTR Diplexer UPCON chains INs & OUTs: depending on fP Diplexer Diplexer FWD Tx RTN Rx Diplexer IFLTR UPCON UPCON CFLTR HMPA OFLTR # of chains depending on fRand fP xN Shared Elements … xGW and POL OBP INs: xGW and POL # depending on fP # depending on fP # depending on fP HMPAs INs & OUTs: depending on fP | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | FWD Tx FWD Tx FWD Tx RTN Rx RTN Rx RTN Rx Figure 4.6: Beam-hopping forward link payload 65 CHAPTER 4. PAYLOAD DEPENDENT SYSTEM MODEL 4.3.3.2 Return Link Architecture For the return link the process followed is the same than in the conventional payload in section 1.1 and the payload scheme can be seen in Figure 4.3. 66 Chapter 5 Analysis of co-Channel Interference The purpose of this chapter is obtain analytic expressions for the co-channel interference at beam level for the forward downlink of satellite systems. To this aim is obtained the received signal in the form y=Hx+n , where H is a matrix compound of the satellite antenna gains matrix G , which depend on the angle θ dened in section A.1, and the rest of the link budgets contributions in matrix A . To perform the analysis in a systematic way we will rst analyze a two beam model to extract at the end a general expression for a variable number of beams. Considered geometric aspects, antenna models and link budget computations are extracted from Appendix A. Besides, as we have seen in Chapter 4, the computation of the link budget depends on the payload model used, and this will aect to matrix A . So for dierent payloads we will obtain dierent levels of interference. The notation used within this chapter for the system model denition is described here below:  Vectors are set in bold lowercase letters.  Matrix are set in bold uppercase letters.  Subindex (.)T denotes the transpose. 5.1 Two beam model Consider the following scenario, a satellite is transmitting information using a determinate number of antennas, 1 to N , to a determinate number of beams, 1 to N . Two of these beams are using the same frequency, hence because of the 67 CHAPTER 6. OVERALL PERFORMANCE  Lower delay bounds for sending a packet can be extracted by assuming t SCH is close to zero and that t a is going to be the transmission time of the most ecient MODCOD. This number gives us an idea of which would be the minimum delay for each of the architectures.  As upper delay bounds are xed by the CoS imposed delays, we will take the most restricting TD 95 delay (from Table 1.7, DG-C=1.4s) and suppose t a is going to be the transmission time of the less ecient MODCOD and obtain t SCH . This number will give us an idea of the maximum time a packet must be in queues, i.e: how good must be the scheduling we design in each of the architectures for the most restricting CoS. As higher is the t SCH number, more relaxed is the constraint while if negative it is impossible to achieve. To these aim, the system parameters in Table 1.2 have been taken and obtained the transmission times (t a , for a 4096 bits block) specied in Table 6.2. Although messages are smaller they must be encapsulated within a block of this size and we assume the entire block must be received to start decoding. MODCOD t a (1) QPSK 1/3 46ms (2) QPSK 1/2 31ms (3) QPSK 2/3 23ms (4) 8PSK 1/2 20.7ms (5) 8PSK 2/3 15.5ms (6) 16APSK 2/3 11.6ms Table 6.2: Obtained transmission time for a packet t a For extracting lower delay bounds 16APSK t a will be used, while for knowing how good must be the scheduling QPSK 1/3 t a will be used. Type of architecture Type of WAN Lower delay for a packet d min Distributed SATBASED dmin(s)=0.7616 TERRBASED dmin(s)=0.2616 Centralized SATBASED dmin(s)=0.5232 TERRBASED dmin(s)=0.2616 Table 6.3: Lower delay bounds 74 CHAPTER 6. OVERALL PERFORMANCE Type of architecture Type of WAN t SCH (s) Distributed SATBASED 1.104 TERRBASED 1.104 Centralized SATBASED 0.408 TERRBASED 1.104 Table 6.4: Scheduling complexity From Table 6.3 and Table 6.4 the following conclusions can be extracted:  Both architectures using TERRBASED-WAN, whether it is centralized or distributed get the minimum lower bounds in end-to-end delay.  The worst lower bound is achieved by the distributed SATBASED-WAN since three satellite hops are introduced, one for requesting resources, another to receive the allocation of the resources and the last one to send the information.  Centralized SATBASED-WAN oers a mid value lower bound delay.  Although lower delay bounds, the time in which the scheduling must be performed is very similar for all the architectures. Particularly the distributed SATBASED-WAN architecture can start the scheduling process while the GESs and the NCC are exchanging resources information (requests petitions and allocation assignments) to minimize this time. Being known this the architecture that would achieve the best trade-o between delay bounds and scheduling complexity are the distributed and centralized architectures with the a terrestrial based WAN. 6.1.2 Simulations and Performance for MF-TDMA design In this section are presented the simulation and performance results for the MFTDMA design done in Chapter 2 section 2.3. As it is explained in the related chapter, two allocation algorithms are compared, the xed T S algorithm and the dynamic T S algorithm. As the baseline options for the number of GES in the system is 1, 3, 5 these will be the cases simulated. The process used for evaluating the MF-TDMA SF allocation algorithms is described next:  Choose a number of GES for the system (1 GES case is obviated since is straightforward to see all resources would be allocated to the unique GES).  Choose a MODCOD per GES supposing a real distribution of the AES per GES, as it can be seen in the table below: 75 CHAPTER 6. OVERALL PERFORMANCE Number of GES MODCOD distribution Users assigned per GES 3 [6 2 2] [6000 2000 2000] 5 [6 4 2 2 1] [3500 2500 1334 1333 1333] Table 6.5: Table of MODCODs and users distributions used in the simulations (MODCOD indexes are extracted from Table 2.2)  Obtain a vector of possible demands between a minimum and a maximum per GES using the characteristics of the ANTARES trac shown in subsection 1.2.5.  Compute the xed timeslot allocation algorithm and the dynamic timeslot allocation algorithm.  Plot the obtained gures of merit using both algorithms in function of the global requested bitrate and plot the total allocated timeslots vs the total demanded timeslots. 6.1.2.1 Simulations for a system with 3 GES The obtained system performance with 3 GES is as shown in Figure 6.1 and 6.2. Figure 6.1: BO and TC performance 76 CHAPTER 6. OVERALL PERFORMANCE Figure 6.2: Mean of Allocated timeslots vs Demanded timeslots 6.1.2.2 Simulations for a system with 5 GES The obtained system performance with 5 GES is as shown in Figure 6.3 and 6.4. Figure 6.3: BO and TC performance 77 CHAPTER 6. OVERALL PERFORMANCE Figure 6.4: Mean of Allocated timeslots vs Demanded timeslots 6.1.2.3 Conclusions on simulation results The following conclusions can be extracted from the graphics in sections 6.1.2.1 and 6.1.2.2 above.  For a system with 3 GES:  When bitrates requests are under the mean (around 6Mbps), both algorithms perform very similar, the results for the BO and TC gures of merit are very similar and both algorithms achieve to allocate all the demanded timeslots.  When bitrates requests are over the mean the dynamic timeslot algorithm performs clearly better than the xed timeslot algorithm. The improve on the BO and TC gures of merit is at least of the 10%. Besides it also gets to allocate at least 10% more demanded timeslots than in the xed timeslot algorithm.  For a system with 5 GES:  When bitrates requests are under the mean, both algorithms perform very similar, the results for the BO and TC gures of merit are very similar and both algorithms achieve to allocate all the demanded timeslots.  When bitrates requests are over the mean the dynamic timeslot algorithm performs slightly better than the xed timeslot algorithm. The improve on the BO and TC gures of merit is around 5%. It also gets to allocate 5% more demanded timeslots than the xed timeslots algorithm. Being known this, we propose to use the dynamic timeslot algorithm since for high bitrate requests it performs clearly better than the xed timeslot algorithm. 78 CHAPTER 6. OVERALL PERFORMANCE Besides as it is explained in the introduction to this work, it is expected than in the following years air trac is going to increase exponentially and therefore control communications. Hence the need of an algorithm to allocate trac demands properly when requests are elevated. The major drawback of this algorithm is its computational complexity, however in the literature are explained several methods to reduce this computational complexity, such as the one in reference [18]. 79 CHAPTER 6. OVERALL PERFORMANCE 6.2 Simulations and Performance for LL-FEC design Within this section are presented the simulations and performance results for the LL-FEC design. The chosen scenarios in the following subsections correspond to each of the erasure models detailed in Chapter 3, section 3.2.2. Results are plot as the CFR using RS vs the CFR without using RS. CFR gure of merit is dened in subsection 3.5.3. 6.2.1 Simulations for the Erasures produced by airplanes in the same line of sight with the satellite scenario The duration of the erasures for this scenario have been extracted from Table 3.2, minimum, mean and maximum values. However the probability of having an erasure, i.e. of one airplane being in the same line of sight with the satellite and blocking the signal, is taken as an input since no information to obtain this value has been found in the literature. Simulation parameters are summarized in Table 6.6. Duration of LL-FEC Probability RS(N,K) codes the erasure (s) delay aware(s) of erasure 0.04 0.68 [1e −4 ,1e −2 ] RS(255,127), RS(255,85), RS(15,7), RS(15,5) 0.3 0.68 [1e −4 ,1e −2 ] RS(255,127), RS(255,85), RS(15,7), RS(15,5) 1.15 1.5 [1e −4 ,1e −2 ] RS(255,127), RS(255,85), RS(15,7), RS(15,5) Table 6.6: Simulation parameters for the current scenario The results for the dierent erasure durations can be seen in Figure 6.5. The conclusions that can be extracted are:  Erasure durations of 0.04 have not been simulated since if the RS code is able to overcome with CFR=1 the 0.3 seconds erasure a shorter with same probability can too.  Codes with N=255 perform slightly better than codes with N=15 for both link layer frame sizes.  Codes using bigger link layer frame sizes perform better. However it must be taken into account that for some bitrates the ADT table is not big enough to t one link layer frame accomplishing the delay requirement then it is impossible to use this mode, i.e. for 0.6s delay at 16kbps ADT size is 3200 bits and link layer frame size is 8192 bits. 80 CHAPTER 6. OVERALL PERFORMANCE  Codes with coderate 1/3 get 1 or close to one CFR in all the cases, while codes using coderate 1/2 does not perform so well with long erasures. It is possible to achieve even better performance if the LL-FEC delay-aware is increased, however, if done this, some of the CoS delay requirements could not be met. 6.2.2 Simulations for the Erasures produced by the airplane itself when maneuvering scenario The simulation parameters for this scenario are extracted from section 3.2.2.2. The probability of performing a maneuver, δ, and the probability of the elevation angle with the satellite, ε , being lower than zero are extracted from Figure 3.8 and resumed in Table 6.7. As no duration for the erasure has been obtained in the literature, as an approximation it will be considered 0.5 seconds erasure. Duration of LL-FEC Prob. of Prob. of Flight RS (N,K) erasure (s) delay aware δ ε < 0 Duration (s) Codes RS(255,127), 0.5 0.68 0.1-0.3 0.2 7200 RS(255,85), RS(15,7), RS(15,5) Table 6.7: Simulation parameters for the current scenario As we can observe from the values in the Table 6.7, this eect is really unreliable to happen since the probability of erasure is under 1e −9 . Simulations showed that no erasures occurred during the simulation time. However if that happened the performance should be similar than in the case above for mean erasure duration since the erasure duration values are similar. 6.2.3 Simulations for the Erasures produced by the blades scenario Although this is not a general case, in the helicopter scenario, as it can be seen in Figure 3.19, most of the erasures are produced because of the blades (any probability interferes in the process). When the erasures are produced by the blades the case is deterministic since the probability of the signal being blocked is always constant. The parameters used in the simulations are shown in Table 6.8 and extracted from section 3.2.2.2. 81 CHAPTER 6. OVERALL PERFORMANCE Duration of LL-FEC Duration Duty cycle RS(N,K) the erasure (s) delay aware (s) no erasure (%) codes RS(255,127), 0.078 0.68 0.2 30% RS(255,85), RS(15,7), RS(15,5) Table 6.8: Simulation parameters for the current scenario Results for this scenario in terms of CFR using RS vs CFR without using RS are shown in Figure 6.6. The conclusions that can be extracted are:  Both RS codes with N=255 and N=15 are able to correct all the errors for the dierent bitrates.  512 bytes link layer frame with link layer FEC delay 0.6 seconds it is not possible since the ADT table is too small to t a frame. 6.2.4 Simulations for the Erasures produced by blades and buildings scenario This scenario takes as a base the scenario where erasures are produced because of the blades, and adds the probability of nding a building during the ight, high enough to block the signal from the satellite. The parameters used in the simulations and extracted from section 3.2.2.2 are listed in Table 6.9. Duration of LL-FEC Probability of RS(N,K) buildings erasure (s) delay-aware (s) building codes 0.9 1.5 1e −4 -1e −2 RS(255,127), RS(255,85), RS(15,7), RS(15,5) Table 6.9: Simulation parameters for the current scenario Results for this scenario are shown in Figure 6.6. The conclusions that can be extracted are the following:  RS codes with N=255 perform slightly better than codes with N=15.  RS codes with bigger link layer frame size (512 bytes) perform better than those with 256 bytes frame size. However as it has been said it must taken into account that there are some cases where the link layer FEC delay is very low that results into a very small ADT table that can not t even an entire frame, hence it is impossible to use this mode. 82 CHAPTER 6. OVERALL PERFORMANCE Figure 6.5: RS performance in commercial aviation scenario wit dierent erasure durations. Top row: RS(15,K) codes for 256 bytes and 512 bytes link layer frames. Bottom row: RS(255,K) codes for 256 bytes and 512 bytes link layer frames 83 CHAPTER 6. OVERALL PERFORMANCE 6.3 Simulations and Performance on the Interference Analysis study In order to simulate the interference in multibeam satellite systems in function of the θ angle we have depicted two dierent scenarios.  A rst scenario composed of 70 beams covering Europe with a half angle beamwidth θ−3dB = 0.245 º , in the Ka band. Parameters for this system are listed in Tables A.1, A.2, A.3 and A.4. Beam lattice is depicted in the left part of Figure 6.7.  A second scenario depicting the baseline option for the ANTARES system, using 3 beams which covers the entire ECAC ight area and with a beamwidth of θ−3dB = 2.90 º . Parameters for this system are listed in Tables A.1, A.2, A.3 and . Beam lattice is shown in the right part of Figure 6.7. For both scenarios the satellite is located at 0 º Latitude, 0 º Longitude. Figure 6.7: Beam lattice for 70 beam scenario and ANTARES scenario respectively For each of the scenarios listed above we will analyze the received SINR within the beam of interest (red colored beam) changing the number and/or location of the interfering beams. Mention that the exible payload has not been simulated since it has been impossible to implement an easy-computing AFR antenna model. For the 70 beam scenario:  We will analyze the SINR produced by an increasing number of adjacent interfering beams (dashed green ring) and the SINR produced by all nonadjacent beams (dashed lilac ring) in Figure 6.7, for each of the payloads. 90 CHAPTER 6. OVERALL PERFORMANCE For the ANTARES scenario:  We will analyze the SINR received in the beam of interest supposing none, one or two interferers for the conventional and beam-hopping payload. 6.3.1 70 Beam system Within this section we will analyze the eect of the interference in a 70 beam scenario, for the conventional and beam-hopping payloads. The following assumptions are made:  For the conventional payload it is assumed a 4 colored bandwidth scheme, i.e. bandwidth is divided in four portions. This means interferers has assigned the same portion of bandwidth as the beam of interest.  For beam-hopping payload, k-interferers mean that k beams are illuminated at the same time sharing the entire bandwidth. Besides, gures from now on are plotted as follows:  Marker shows the beam center of the beam of interest.  First contour indicates the 3dB fall with respect to the maximum of the plotted variable.  Green label shows the rst 10dB with respect to plotted variable for the beam of interest.  Grey points mark the original edge of the beam, i.e. 3dB fall without interferers. 6.3.1.1 Dierences between payloads The received SINR for a user in the beam of interest for each of the payloads and when there are no interferers, i.e. the SNR, is shown in Figure 6.8. In the mentioned gure it can be observed the following:  Received SNR for the conventional payload is higher than in the beamhopping payload since in the conventional payload the entire bandwidth is divided in parts while in beam-hopping the entire bandwidth is assigned to beams. Although this allows to receive more information it can make closing the link budget more dicult. 91 CHAPTER 6. OVERALL PERFORMANCE 9.09863 X= 14.25 Y= 50.75 Level= 19.0986 Longitude in degrees Latitude in degrees SNR (dB) in function of θ for the conventional payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 10 1.32864 X= 14.25 Y= 50.75 Level= 11.3286 Longitude in degrees Latitude in degrees SNR (dB) in function of θ for the beam−hopping payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 Figure 6.8: SINR received by the beam of interest for 0 interferers and for the conventional and beam-hopping payload respectively 6.3.1.2 Interference generated by adjacent beams In this section we will study the eect of the interference generated by an increasing number of adjacent beams (green ring in Figure 6.7) for both payloads. To this aim we will plot the SINR as a function of the number of the adjacent interfering beams for both payloads as seen in Figure 6.9. The increase in the received SINR for the conventional payload with respect to the beam-hopping one is given in the rigth part of the gure. 0 1 2 3 4 5 6 7 8 −5 0 5 10 15 20 Number of interfering beams beams SINR (dB) SINR evolve as a function of the number of interferers Conventional payload Beam−hopping payload 0 1 2 3 4 5 6 7 8 0 10 20 30 40 50 60 70 80 90 Number of interfering beams beams SINR increase in % SINR increase of the conventional payload with respect to the beam−hopping payload Figure 6.9: SINR as a function of the number of interferers We can observe that as the number of adjacent interfering beams increase, the SINR decays very fast for both payloads, even for a small number of interferers. Moreover We the trend is that dierences in the received SINR between both payload decrease when the number of interfering beams increase. Results lead to the following rules when designing:  For the conventional payload it should be avoided to assign the same frequency bands to adjacent beams. 92 CHAPTER 6. OVERALL PERFORMANCE  For the beam-hopping payload it should be avoided to illuminate at the same time adjacent beams. 6.3.1.3 Interference generated by non-adjacent beams In this section is plotted the SINR caused in the beam of interest by all the beams that cross the lilac dashed ring in Figure 6.7. This case is of special interest in the conventional payload because when using the typical algorithm for assigning frequencies these would be the beams with the same frequency band. Results for the beam-hopping payload will help us to compare performances. 6.9487 X= 14.25 Y= 50.75 Level= 16.0643 Longitude in degrees Latitude in degrees SINR (dB) in function of θ for the conventional payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 10 0.837068 X= 14.25 Y= 50.75 Level= 10.6506 Longitude in degrees Latitude in degrees SINR (dB) in function of θ for the beam−hopping payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 Figure 6.10: SINR in the beam of interest when all beams crossing the dashed lilac rings in Figure 6.7 interfere As we expect other further beams using the same frequency band will not in- uence in the SINR, this can be a realistic example of the real received SINR 93 CHAPTER 6. OVERALL PERFORMANCE within a beam. In Figure 6.10 we can observe that all the interferers produce a 3dB fall in the the SINR for the conventional payload and a 1.3dB fall for the beam-hopping payload with respect to the case without interferers. Although dierences in the received SINR are now smaller, conventional payload still receives a higher SINR than the beam-hopping payload. The increase in the SINR of the conventional payload with respect of the beam-hopping is of the 70%. 6.3.1.4 Interference generated by far away beams Finally, we can observe which is the SINR decrease within the edge of coverage of the beam of interest, when the interfering beam is placed far-away from the beam of interest. The eect can be seen for both payloads in Figure 6.11. 9.09847 X= 14.25 Y= 50.75 Level= 19.0983 Longitude in degrees Latitude in degrees SINR (dB) in function of θ for the conventional payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 10 1.32859 X= 14.5 Y= 51.75 Level= 10.8703 Longitude in degrees Latitude in degrees SINR (dB) in function of θ for the beam−hopping payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 Figure 6.11: SINR decrease caused by far-way interfering beams What can be observed for both payloads is that far-located interfering beams almost don't cause any decrease in the received SINR since the values obtained are similar than for no interferers (Figure 6.8). 6.3.1.5 Conclusions for the 70 beam system Election between conventional payload or beam-hopping payload is a trade-o since:  Conventional payload achieves much higher SINR values, increases reach the 80% in some cases. This makes closing the link budget easier than in the beam-hopping payload. The drawback of this scheme is that is not very exible since we work with a xed frequency reuse pattern.  Beam-hopping payload assigns the entire bandwidth to each beam which in theory allows higher information uxes, besides it is more exible since we can decide which beams will be illuminated at a time. It is demonstrated 94 CHAPTER 6. OVERALL PERFORMANCE in reference [24] that beam-hopping payload adapts better to real trac uxes despite its lower SINR, specially under heavy loaded conditions. It achieves average increases on the Satisfaction Factor (dened as the ratio of served over demanded trac) of the 10%. However closing the link budget under this case could be more dicult. 95 CHAPTER 6. OVERALL PERFORMANCE 6.3.2 ANTARES system Within this section we will analyze the eect of the interference in the ANTARES system for the conventional and beam-hopping payloads. Figures from now on are plotted as follows:  Marker shows always the center of the beam of interest.  First contour indicates the 3dB fall with respect of the maximum of the plotted variable.  Successive contours indicate additional 10dB fall with respect of the maximum of the plotted variable and its values is indicated with a green label.  Grey points mark the original edge of the beam, i.e. 3dB fall without interferers. 6.3.2.1 Analysis of the SINR In Figure 6.8 is presented the received SINR in the beam of interest for the conventional and beam-hopping payloads when there are no interferers, i.e. the SNR of the system. It can be observed in the contour the characteristics of the antenna pattern,strong red contours are produced by the main lobe of the antenna while orange and yellow contours are produced by side lobes (such lobes cause the interference) nally dark and soft blue contours depict the zeros of the antenna. −32.2146 −32.2146 −12.2146 −12.2146 −2.21465 −2.21465 17.7854 17.7854 17.7854 17.7854 17.7854 24.7751 24.7751 24.7751 24.7751 −32.2146 −32.2146 −12.2146 −12.2146 −12.2146 X= −20 Y= 40 Level= 27.6683 Longitude in degrees Latitude in degrees SNR (dB) in function of θ for the conventional payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −70 −60 −50 −40 −30 −20 −10 0 10 20 −41.7455 −41.7455 −21.7455 −21.7455 −11.7455 −11.7455 8.25445 8.25445 8.25445 8.25445 8.25445 15.2442 15.2442 15.2442 15.2442 −41.7455 −41.7455 −21.7455 −21.7455 −21.7455 X= −20 Y= 40 Level= 18.1374 Longitude in degrees Latitude in degrees SNR (dB) in function of θ for the beam−hopping payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 10 Figure 6.12: SINR received by the beam of interest for 0 interferers and for the conventional and beam-hopping payloads respectively As we can observe there is a big dierence in the received SNR between the conventional and the beam-hopping payload, the reason beyond that is: 96 CHAPTER 6. OVERALL PERFORMANCE  In the conventional payload the fact that there are no interferers means that bandwidth is shared among beams, hence noise bandwidth is reduced and the received SINR is bigger.  In the beam-hopping payload the entire bandwidth is destined to each beam; the fact that there are no interferers means that only one beam is illuminated at a time. Hence for this case the noise bandwidth is bigger and the received SINR is lower. Now, we will set one interfering beam in the system. In Figure 6.13 we can observe the received SINR for each of the payloads in the left part of the gure, while in the right part we can observe the power level of the interference. −58.1939 −58.1939 −58.1939 −38.1939 −38.1939 −18.1939 −18.1939 −8.19393 −8.19393 −8.19393 11.8061 11.8061 11.8061 11.8061 −38.1939 −38.1939 18.7958 18.7958 18.7958 X= −20 Y= 40 Level= 12.2227 Longitude in degrees Latitude in degrees SINR (dB) in function of θ for the conventional payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −70 −60 −50 −40 −30 −20 −10 0 10 −178.4424 −178.4424 −158.4424 −158.4424 −148.4424 −148.4424 −128.4424 −128.4424 −128.4424 −128.4424 −121.4527 −121.4527 −121.4527 −158.4424 −158.4424 X= −20 Y= 40 Level= −127.7066 Longitude in degrees Latitude in degrees Interference (dB) in function of θ for the conventional payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −210 −200 −190 −180 −170 −160 −150 −140 −130 −62.6961 −62.6961 −62.6961 −42.6961 −42.6961 −22.6961 −22.6961 −12.6961 −12.6961 −12.6961 7.30391 7.30391 7.30391 7.30391 14.2936 14.2936 14.2936 −42.6961 −42.6961 X= −20 Y= 40 Level= 11.5388 Longitude in degrees Latitude in degrees SINR (dB) in function of θ for the beam−hopping payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 10 −180.1918 −180.1918 −160.1918 −160.1918 −150.1918 −150.1918 −130.1918 −130.1918 −130.1918 −130.1918 −123.2021 −123.2021 −123.2021 −160.1918 −160.1918 X= −20 Y= 40 Level= −129.456 Longitude in degrees Latitude in degrees Interference (dB) in function of θ for the beam−hopping payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −220 −210 −200 −190 −180 −170 −160 −150 −140 −130 Figure 6.13: SINR received by the beam of interest for 1 interferer and for the conventional and beam-hopping payloads respectively The following conclusions can be extracted:  For the conventional payload one interfering beam means that the assigned bandwidth to each beam has increased, hence does it noise power level. Therefore besides of the interference power, noise power level has increased which leads to the big loss in the SINR level. 97 CHAPTER 6. OVERALL PERFORMANCE  For the beam-hopping payload one interfering beam means that the beam of interest and the interfering beam have assigned the entire frequency at the same time. Hence besides of the interference generated noise power level is maintained and the SINR loss is lower than in the conventional payload. Although dierences between the SINR level have been reduced, conventional payload achieves a mean higher value as we can deduce by the values of the contours. Finally if we set that all beams interfere between them, we can see the pattern in Figure 6.14 for the beam of interest.Note that now conventional and beamhopping payload have assigned the entire bandwidth to each of the beams, hence the noise power is the same for both of them. From there the similar SINR level obtained, conventional payload obtains a higher SINR value since its output power is higher. −62.4993 −62.4993 −42.4993 −42.4993 −22.4993 −22.4993 −12.4993 −12.4993 −12.4993 7.50068 7.50068 7.50068 −42.4993 −42.4993 14.4904 14.4904 14.4904 −22.4993 X= −20 Y= 40 Level= 2.8636 Longitude in degrees Latitude in degrees SINR (dB) in function of θ for the conventional payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 10 −144.0163 −144.0163 −124.0163 −124.0163 −124.0163 −124.0163 −117.0266 −117.0266 −117.0266 −117.0266 X= −20 Y= 40 Level= −117.9876 Longitude in degrees Latitude in degrees Interference (dB) in function of θ for the conventional payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −190 −180 −170 −160 −150 −140 −130 −120 −47.3009 −47.3009 −27.3009 −27.3009 −17.3009 −17.3009 −17.3009 2.69907 2.69907 2.69907 2.69907 −47.3009 −47.3009 9.68877 9.68877 9.68877 −27.3009 −67.3009 X= −20 Y= 40 Level= 2.7645 Longitude in degrees Latitude in degrees SINR (dB) in function of θ for the beam−hopping payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −80 −70 −60 −50 −40 −30 −20 −10 0 −145.7657 −145.7657 −125.7657 −125.7657 −125.7657 −125.7657 −118.776 −118.776 −118.776 −118.776 X= −20 Y= 40 Level= −119.737 Longitude in degrees Latitude in degrees Interference (dB) in function of θ for the beam−hopping payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −190 −180 −170 −160 −150 −140 −130 −120 Figure 6.14: SINR received by the beam of interest for 2 interferer and for the conventional and beam-hopping payloads respectively It can also be interesting to plot only the received SINR within the edge of the beam of interest and within the edge of the interfering beams and observe the results. Obtained SINR is shown in Figure 6.15 and results show the following: 98 CHAPTER 6. OVERALL PERFORMANCE  As the satellite is located at 0 latitude and 0 longitude it exists one case that the SINR of both considered beams is the same because beams are symmetrically located with respect the satellite.  However as it can be seen in the two last images of Figure 6.15 the level of interference that a beam a causes to a beam b need not be the same as the one that beam b causes to beam a . 11.8061 11.8061 11.8061 11.8061 11.8061 11.8061 18.7958 18.7958 18.7958 18.7958 11.8061 11.8061 11.8061 18.7958 18.7958 18.7958 Longitude in degrees Latitude in degrees SINR (dB) within the edge of the beam of interest and of the interferers in function of θ for the conventional payload X= −20 Y= 40 Level= 12.2227 X= 20 Y= 40 Level= 12.2227 −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 2 4 6 8 10 12 14 16 18 20 6.40298 6.40298 6.40298 6.40298 6.40298 13.3927 13.3927 13.3927 4.98976 4.98976 4.98976 11.9795 11.9795 Longitude in degrees X= −20 Y= 40 Level= 3.3341 X= 0 Y= 60 Level= 4.7127 Latitude in degrees SINR (dB) within the edge of the beam of interest and of the interferers in function of θ for the conventional payload −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −2 0 2 4 6 8 10 12 7.50068 7.50068 7.50068 7.50068 7.50068 7.50068 7.50068 7.50068 Longitude in degrees Latitude in degrees SINR (dB) within the edge of the beam of interest and of the interferers in function of θ for the conventional payload −0.33734 −0.33734 −0.33734 X= 0 Y= 60 Level= 1.7454 X= −20 Y= 40 Level= 2.8636 X= 20 Y= 40 Level= 2.8636 −60 −40 −20 0 20 40 60 0 10 20 30 40 50 60 70 80 −4 −3 −2 −1 0 1 2 Figure 6.15: Received SINR within the edge of the beam of interest and within the edge of the interferers for conventional payload This is produced because the θ from the interfering beam with the user is not the same for dierent beams of interest, it depends on the position of the satellite. As seen, if beams are symmetrically located from the satellite the interference is the same from a to b than from b to a , however, if not, generated interference levels are dierent. This must be taken into account when designing a system since we must consider worst cases for closing the link budget. Figure 6.16 shows the mean received SINR in the beam of interest for each of the payloads as a function of the number of interferers. In the image we can observe how as the number of interfering beams increase dierences between the received SINR in both payloads become smaller. 99 APPENDIX A. LINK BUDGET A.2.2 Forward Downlink For the forward downlink the following acronyms have been dened:  EIRPsat =Psat · OBOhpa · Lsat · Gtx(θ) is the satellite EIRP, with Psat the on board saturated power, OBOhpa is the Output Back-O that applies to that satellite payload, Lsat are the satellite repeater output losses and Gtx is the antenna gain that depends on the angle formed between the satellite beam center for the considered ground terminal and the ground terminal position.  Ldown are are the free space losses and the additional rain, polarization, atmospheric and scintillation losses for the downlink.  Ggt is the ground terminal antenna gain.  Tnoise is the noise temperature of the receiver considered, in this case the ground terminal. The carrier power computation C and noise power per unit bandwidth N0 is given by the following equations: C=Psat · OBOhpa · Lsat · Gtx(θ) · Ldown · Ggt (A.6) N0 = kTnoise (A.7) So we can express the overall C/N0 for the forward downlink as: C N0FWD−DOW N =Psat ·β · Gtx(θ) kTnoise [dB ·Hz] (A.8) Where β=OBOhpa · Lsat · Ldown · Ggt . A.2.3 Return Uplink Besides we can substitute the subindex gw by the subindex gt in equations A.3 and A.4 to obtain the Return Uplink link budget, from ground terminal to satellite: C=Pgt ·Ggt ·Lgt ·Lup · Grx(θ) (A.9) N0 = kTnoise (A.10) 106 APPENDIX A. LINK BUDGET And we can express the overall C/N0 as: C N0RT N−UP =Pgt ·γ · Grx(θ) kTnoise [dB ·Hz] (A.11) Where γ=Ggt ·Lgt ·Lup A.2.4 Return Downlink If we substitute the subindex gt by the subindex gw in equation A.6 and A.7 we can obtain the computation of the Return Downlink link budget, from satellite to gateway where: C=Psat · OBOhpa · Lsat · Gtx(θ) · Ldown · Ggw (A.12) N0 = kTnoise (A.13) And we can express the overall C/N0 as: C N0RT N−DOW N =Psat ·δ · Gtx(θ) kTnoise [dB ·Hz] (A.14) Where δ=OBOhpa · Lsat · Ldown · Ggw . A.3 Link Budget and Payload Values The system and receiver parameters for the Ka band system and for the ANTARES system are given in the following Table (A.1). System and Receiver parameters Parameter Ka band ANTARES Modulation 8PSK QPSK, 8PSK, 16APSK Downlink frequency 19.950GHz 1500MHz Bandwidth 500MHz 14MHz Half angle Beamwidth ( θ−3dB ) 0.245 º 2.90 º G gt [21] [23] Tnoise,gt [21] [23] Table A.1: System and receiver parameters 107 APPENDIX A. LINK BUDGET Since we are strongly interested in the forward downlink, here below are given the references where the payload parameters for the forward downlink can be found, in Table A.2, A.3 and A.4. Conventional Payload Parameter Ka band ANTARES Psat [21] Table 2-14 [21] Table 3-19 Gmax [21] Table 2-14 [22] OBOhpa [21] Table 2-14 [21] Table 3-19 Lsat [21] Table 2-14 [21] Table 3-19 Ldown 211.1dB 188dB Table A.2: Satellite parameters for conventional payload Flexible Payload Parameter Ka band ANTARES Psat [21] Table 3-19 [21] Table 3-19 Gmax [21] Table 3-19 [22] OBOhpa [21] Table 3-19 [21] Table 3-19 Lsat [21] Table 3-19 [21] Table 3-19 Ldown 211.1dB 188dB Table A.3: Satellite parameters for exible payload Beam-hopping Payload Parameter Ka band ANTARES Psat [21] Table 4-9 [21] Table 4-9 Gmax [21] Table 4-9 [22] OBOhpa [21] Table 4-9 [21] Table 4-9 Lsat [21] Table 4-9 [21] Table 4-9 Ldown 211.1dB 188dB Table A.4: Satellite parameters for beam-hopping payload The constant β for the system conguration chosen and for the three dierent payloads is detailed here after: System Conventional Flexible Beam-hopping Ka band 2.97e −22 (-215.2dB) 3.58e −22 (-214.46) 2.64e −22 (-215.77) ANTARES 5.92e −20 (-192.2) 7.14e −20 (-191.46) 5.28e −20 (-192.77) Table A.5: β constant for the both systems and the three dierent payloads 108 APPENDIX A. LINK BUDGET The computed noise densities level per unit bandwidth, N0 for the Ka band system and the ANTARES system respectively is: N0Ka = 10 log10(1.38e−23 ·188.3K·B) = −205.85dB ·Hz (A.15) N0ANT = 10 log10(1.38e−23 · 164.3K·B) = −206.4dB · Hz (A.16) 109 Appendix B Guidelines for LL-FEC implementation in ANTARES In this appendix is explained, for each of the scenarios found, the method used to implement and nd the most suitable RS codes in function of the parameters dening each scenario. For the four dierent scenarios found, the parameters dening the erasure channel model are basically the duration of the erasure and the probability of the erasure. Depending on this values and the t delay wanted congurations and performance of the RS codes will vary. Besides in some of the scenarios not all the needed information is available and congurations and performance rely on assumptions that must be made. Also nal performance will rely on nal decisions of link layer frame and physical layer packet sizes. B.1 Undened parameters and general procedure for RS implementation in Table B.1 is presented a relation of the undened parameters of the ANTARES project and the values assumed that are needed for obtaining the RS performance in each scenario. Parameter Values assumed Bitrate R b 16kbps, 32kbps LL frame size 256 bytes, 512 bytes PHY packet size 1024 bits, 2048 bits Overhead L1/L2 10% Table B.1: Undened parameters and values assumed 110 APPENDIX B. GUIDELINES FOR LL-FEC IMPLEMENTATION IN ANTARES For Reed Solomon ADT and RSDT tables congurations the next process is followed:  Fix the delay-aware, t delay , of LL-FEC constrained to t delay >t erasure , otherwise all packets could be erased. Besides t delay should be set into a lower value than the most stringent of the trac CoS TD 95 if possible. If performance under this value is too bad or t erasure >TD 95 , t delay should be set into a lower value than ET.  Choose an RS(N,K) code, a physical layer bitrate, a link layer frame size and an encapsulation overhead.  Compute the number of rows, Nrows, needed for the ADT and RSDT table being known that 1 : tdelay =m · N · Nrows +OH(bits) Rb  Build and ll the ADT Nrows×K and RSDT Nrows×T tables.  Extract information in the tables column wise and add link layer and physical layer headers.  Select a scenario and generate the erasure channel model.  Send the information through the channel and de-encapsulate.  Evaluate CFR performance without using LL-FEC.  Fill the ADT and RSDT reception tables, read information row wise and correct the errors in the row if less than T=(N-K)/2 erasures are produced.  Evaluate CFR performance using LL-FEC and compare with CFR without using LL-FEC. In Figure B.1 is shown a ux diagram of the process followed. 1 Detailed explanation on the parameters involving t delay can be found in section 3.5.3 111 APPENDIX B. GUIDELINES FOR LL-FEC IMPLEMENTATION IN ANTARES Fix tdelay TD96>tdelay>terasure Choose an RS(N,K) code Assume values for undefinedAssume values for undefined parameters: Rb, LL frame size, PHY packet size, PHY OH Build tables and Build tables and extract information column wise adding headers and overhead Compute Nrows using: b rows delay R bitsOHNNm t)(··   Correct the errors in the row Generate erasure model for an scenario Send informationSend information through channel and deencapsulate Build receptionBuild reception tables and read information row wise Evaluate CFR without LL-FEC Evaluate CFR using LL-FEC Compare perfomance If not OK set lower Tdelay 2 KN T  Yes No Figure B.1: Flux diagram B.2 Characteristics of the Erasures produced by airplanes in the same line of sight with the satellite scenario Parameter Expression Values obtained/assumed t erasure x cos(θ) · vrelative 0.04s, 0.3s, 1.15s p erasure - 1e −4 -1e −2 delay-aware - 0.68s, 1.5s Table B.2: Channel parameters Where x is the length of the aircraft, θ the horizontal angle between the aircrafts and v relative the relative speeds between the two airplanes. Detailed explanation of the t erasure obtention is given in section 3.2.2.2. 112 APPENDIX B. GUIDELINES FOR LL-FEC IMPLEMENTATION IN ANTARES B.3 Characteristics of the Erasures produced by the airplane itself scenario Parameter Expression Values obtained/assumed t erasure - 0.5s p erasure - 3.4e −12 -3.4e −10 delay-aware - 0.68s Table B.3: Channel parameters A detailed explanation of the parameters involving t erasure can be found in section 3.2.2.2. B.4 Characteristics of the Erasures produced by the blades scenario Parameter Expression Values obtained/assumed t erasure 1 RP Mblades/60 · (bladeswidth) 0.078 Duty cycle terasure 2πdantenna 4−bladeswidth · 1 RP Mblades/60 30% of the erasure delay-aware - 0.68s Table B.4: Channel parameters Where the term d antenna is the distance from the rotation axis of the blades to the location of the helicopter antenna. A detailed explanation of the parameters involving t erasure can be found in section 3.2.2.2. B.5 Characteristics of the Erasures produced by blades and buildings scenario Parameter Expression Values obtained/assumed t erasure l vhelicopter 0.9 p building erasure can be obtained 1e −4 -1e −2 from real city maps delay-aware - 1.5 Table B.5: Channel parameters 113 APPENDIX B. GUIDELINES FOR LL-FEC IMPLEMENTATION IN ANTARES Where l is the length of the building and v helicopter the speed in m/s of the helicopter. A detailed explanation of t erasure obtention can be found in section 3.2.2.2. 114 Bibliography [1] Multi-Star, Payload Systems, Beam Hopping Techniques for multibeam satellite systems. [2] Draft version of System Simulator Architecture: Non Beam Hopping System, Beam Hopping Techniques for multibeam satellite systems. [3] G.P. Calzolari, M. Chiani, F. Chiaraluce, R. Garello and E. Paolini, Channel coding for future space missions: New requirements and trends. Procedings of the IEEE, 95(11):2157-2170, Nov. 2007. [4] Jon Hamkins, Optimal Codes for the Burst Erasure Channel. IPN Progress Report 42-174 , August 2008 . [5] Amin Shokrollahi, Senior Member, IEEE , Raptor Codes. IEEE Transactions on information theory, vol. 52, no. 6, pp. 2551-2567, June 2006 . [6] Jiang Lei, Student Member, IEEE , María Ángeles Vázquez-Castro, Senior Member, IEEE , and Thomas Sotckhammer, Senior Member IEEE , LinkLayer FEC and Cross-Layer Architecture for DVB-S2 Transmission With QoS in Railway Scenarios. IEEE Transactions on vehicular technology, vol. 58, no. 8, pp. 4265-4276, October 2009 . [7] M. Yang and W. E. Ryan, Performance of eciently encodable low-densityparity-check codes in noise bursts on the EPR4 channel. Magnetics IEEE Transactions, vol. 40, no. 2, pp. 507-512, March 2004. [8] E. paolini and M.Chiani, Improved low-density-parity-check codes for burst erasure channels. Communications,2006. ICC'06. IEEE International Conference on 3:1183-1188, June 2006. [9] Gokul Sridharan, Abishek Kumarasubramanian, Andrew Thangaraj and Srikrishna Bhsayam, Optimizing burst erasure correction of LDPC codes by interleaving. Information Theory, 2008. ISIT 2008. IEEE International Symposium on, pp1143-1147, July 2008. [10] Luigi Rizzo, Eective erasure codes for reliable computer communication protocols, ACM SIGCOMM Computer communication review, vol. 27, pp. 24-36, 2007. 115