Boutique 
Model-Based Design Conference 2005
Abstracts
Day 1 Keynote: Organiser la modélisation et la simulation en projet pour renforcer l'efficacité de l'ingénierie véhicule - l'expérience de Renault
Speaker: Jean Marc Crepel, RENAULT
Day 1: Conference Sessions - Model-Based Design for Control
MATLAB/Simulink for real time space simulation projects: the ATV test facilities experience
Speakers: Cristiano Leorato, Dataspazio s.p.a.Jacques Perennec, EADS ST
The usage of MATLAB/Simulink to develop simulation models is more and more frequent in the space industry.
Simulink provides a fast prototyping non-real-time stand-alone environment, as well as the possibility, by means of Real Time Workshop (RTW), to automatically generate ANSI-C code to be embedded into complex real time simulation infrastructures.
In this paper we describe how MATLAB / Simulink is used to develop and maintain the numerical models of the ATV Test Facilities simulators.
The ATV (Automated Transfer Vehicle) is a program funded by the European Space Agency (ESA). The prime contractor is EADS ST Les Mureaux . The ATV is a completely automatic unmanned vehicle. It is launched by Ariane5 and used to supply the International Space Station (ISS) with cargoes (food, water, spare parts, etc.) and propellants. It is also used to re-boost the ISS.
The complexity of the missions and the involved costs implied the development of a set of simulators to be used on different test platforms, dedicated to specific purposes (flight software validation, procedures validation, crew training, electrical and functional equipment validation).
The development of the ATV numerical models takes place at the EADS Space Transportation premises, in Les Mureaux (Paris).This experience has shown that, by using precise Simulink programming standards, the code generated by RTW can meet real time performance requirements and allow the save and restore of the simulation sessions.
Moreover we will see how Simulink is used throughout all the development phases of the ATV simulator and maintenance. This allows keeping a high coherency and quality level on the code and on the documentation of the whole project. Four years after the development of the first ATV models, several evolutions can still today be regularly planned and efficiently implemented even by a reduced development team.
Optimisation de la Conception de simulateurs de mouvement Wuilfert avec MATLAB & Simulink
Speaker: Sophie Ponchaud – SN WUILFERTLe “Model-Based Design” chez RENAULT F1 TEAM: un défi humain et technique
Speaker: Alain Vande Walle, Renault F1Pour Renault F1 Team, 2005 sera l'heure de vérité. Depuis notre retour en Formule 1 en 2002, nous nous sommes toujours fixés des objectifs ambitieux avec pour ligne de conduite la recherche permanente de la performance avec pour maître mot une cohérence « Virtuel / Réel » entre le monde de la simulation et les essais en piste ou au banc moteur.
La modélisation comportementale du moteur et de son environnement piste, le prototypage rapide des lois de contrôle et la génération de code ainsi que la mesure synchrone des phénomènes physiques sont autant d'étapes d'une démarche connue sous le nom de « Model Based Design » qui apporte son lot de performance et représente un défi humain et technique permanent chez Renault F1 Team.
Desktop Simulation in Airbus: Status & Perspectives for Flight Control Design
Speaker: Philippe Ménard, AIRBUSFlight Control Systems on a commercial aircraft are made more and more safe, simple to use, and available for pilots and operators, inducing (as a paradox at first glance) more complex systems, architectures, monitoring devices and control laws. Furthermore, a mandatory target is to provide the Airlines with a more mature aircraft at Entry Into Service, and to guaranty a high level dispatch reliability, which is actually a key point for all aircraft manufacturers.
To achieve these objectives, a promising track is to drastically improve our simulation facilities, combining more and more sub models, performing more and more tests, developing more automated analysis tools, to increase validation coverage and finally enhance the design. But relying more and more on simulation leads to combine efficiently new processes and tools, to get at the end more confidence in testing outputs.
This conference focuses on how desktop simulation may contribute. Thanks to empowerment of tools, information technologies, computers, networks, and thanks of course to experience gained throught previous aircraft development programs, we can probably shift the border between hardware in the loop validation and emulated systems validation, and take advantage from desktop simulation to cut the exponential link between validation stage and validation cost. In other words, we can now perform a fine validation in some fields (handling qualities and control laws optimization for instance), very soon in the development phase, with low cost means but high added value feedbacks, to make our design better and anticipate modifications.
As an illustration, we propose to see how we can, with a real time desktop simulator, assess stability margins in manual control of an aircraft, and stability robustness wrt disturbances added in the control loop. This low cost mean may reveal design traps at any stage of development, from feasiblity and research till support to in service aircrafts. But to take all the benefits, some golden rules have to be carefully settled: some of them will be underlined.
Conception et mise au point sous MATLAB / Simulink des nouveaux algorithmes embarqués dans les directions assistées automobiles
Speaker: Jean-Luc Deville, KOYOLes directions assistées automobiles sont actuellement le centre d'un nombre important d'innovations techniques, basées sur l'utilisation de plus en plus importante de calculateurs électroniques.
KOYO, leader européen et mondial sur ce domaine, travaille avec les plus grands constructeurs automobile et étudie des logiciels embarqués de plus en plus complexes afin de répondre à des besoins:- de performances toujours plus élevées
- de contraintes de sécurité actives et passives toujours plus importantes
- d'exigences de réduction permanente des coûts, par intégration dans le logiciel de fonctions autrefois confiées à des capteurs ou équipements supplémentaires.
Pour répondre à l'ensemble de ces demandes, KOYO a décidé d'utiliser les outils MATLAB / Simulink, depuis les phases amont d'analyse des problèmes et de spécifications du besoin, jusqu'aux phases de validation sur banc puis véhicule avec les clients, des algorithmes conçus en interne.
L'industrialisation de ces logiciels, sous-traitée par la suite à des fournisseurs internationaux, est alors basée sur des échanges de modèles MATLAB / SIMULINK de simulation (spécifications exécutables), ces derniers étant enrichis par des contraintes liées à tout logiciel temps réel embarqué dans un calculateur
Future Calibration Process and Methodologies at Toyota
Speaker: Masoto Ehara, TOYOTA
Day 1: Master class Track - Model-Based Design for Signal Processing and Communications
Les supports des Master Class seront en anglais et les présentations en françaisMaster class 1: Developing Video and Image Processing Applications in Simulink
In this Master Class, you will learn how to develop real-time video processing systems using Simulink and Model-Based Design. The emphasis will be on prototyping video and image processing applications using Simulink, the Signal Processing Blockset, and the Video and Image Processing Blockset, then automatically generating efficient code for a DSP chip using Real-Time Workshop. You will learn about key aspects of system design, floating-point to fixed-point conversion, and interactive visualization and verification of simulation results.
Master class 2: Specification and Development of a Wireless Communication System
This Master Class will present the application of Model-Based Design to communication system design, using the example of the ETSI HiperMAN physical layer standard. We will demonstrate efficient ways to transform the published standard into an executable specification using MATLAB and Simulink. Using the end-to-end model as a test harness, we will show how to elaborate the design of key signal processing subsystems, including techniques for modeling frames and handling rate conversions. The insights you gain from this class will help you to develop clear and comprehensive models to support your design, development, and testing efforts.
Master class 3: Fixed-Point DSP Modelling, Code Generation, and Verification for FPGAs
This Master Class will demonstrate the development of a signal processing application using Model-Based Design by showing, step-by-step, the specification, design, simulation, code generation, and verification of a high-speed digital “front end” for a software-defined radio (SDR) application. We will demonstrate the design of fixed-point, multirate filters for digital down conversion; validation of the design via MATLAB simulation; automatic generation of VHDL; and verification of the HDL model through co-simulation, using Simulink and ModelSim. Fixed-point scaling and hardware pipelining will be utilized to achieve an efficient and practical implementation in silicon.
Day 2 Keynote: Linking Algorithms to Silicon, an actual implementation flow
Speaker: Philippe Magarshack, ST Microelectronics
With the computing capacity offered today by state-of-the-art Silicon processes, many complex algorithms can be implemented efficiently and in a cost effective manner on just a few mm2 of a silicon chip. This capability can enable namy new products in consumer applications, such as W-LAN, UWB, 3G-radios, MPEG4 Codecs, ... and mixing all these functions on a single silicon chip opens up many more possibilities.
However, in order to meet time-to-market demands, an efficient path from Algorithm Models to Silicon is mandatory. Some researchers at ST have successfully implemented a prototype of such a flow and it has already been applied to key products.
Some perspectives will be offered for streamlining and extending the flow in the future.
Day 2: Conference Sessions - Model-Based Design for Signal Processing & Communications
The Application of MathWorks tools in the Architecting of Mobile Phone Radio Frequency ICs
Speaker: Patrick Pratt, Freescale SemiconductorThe architecting, design and verification of radio frequency integrated circuits (RFIC) for the mobile phone market is becoming increasingly demanding due to the shrinkage of both product cycles and margins. The Radio Products Division of Freescale has adopted the MathWorks tools as an essential component in its ‘New Product' flow. This presentation will focus on how the tools have been used during the architecting of the Power Amplifier control loop for the next generation of Freescale's GSM RFICs. After introducing the system and outlining the relevant specifications it will be shown how the MathWorks tools were used to develop a full system model that was used extensively during the architecting and verification phase. In particular, it will be shown how the script based MATLAB tool was central to the early analytical based investigative stage and how Simulink was ideal for the time domain simulations of the system.
Paving the way to 4G Wireless systems with Simulink
Speaker: Thierry Lestable, SamsungDesigning next generation wireless transmission systems requires researchers from the Advanced Technology Group (ATG) in Samsung Electronics Research Institute (SERI) in UK, to deal with leading edge technologies from Physical layer to upper layers.
In order to unify baseband algorithm development, and to establish durable consistency in the physical layer research direction, the ATG decided to migrate all their algorithms to one common platform, Simulink.
As a partner in major European Union cutting edge research projects (Winner & Magnet), Samsung UK wanted to push this platform for adoption by its partners, in order to finalize a software library development to be integrated under Simulink.
As a contributor on physical layer study items, ATG is particularly involved in the coding/decoding group, together with MC-CDMA cluster. The coding-spreading trade-off with major coding scheme candidate for 4G is of high interest.
Within the MC-CDMA study context implementation of generic multi-user environment, enables to study the sensitivity of detection algorithms (SUD, MUD) with respect to the Multiple Access Interference (MAI). Indeed, ATG is investigating new Interference Cancellation (IC) algorithms with hard detection, but also with soft information obtained from several coding schemes (CC, PCCC, LDPC…).
Application of System Generator for model-Based FPGA Design
Speaker: Michel Pecot, XilinxThe purpose of this presentation is to describe how the System Generator tool has been used efficiently to model, develop, test and verify a complete signal processing application on a Xilinx FPGA.
The application lies in the telecommunications area (UMTS base stations), and the system to be integrated inside the FPGA consists of a multiple channel digital down-converter for the receiver part, and a multiple channel digital up-converter associated with crest factor reduction for the transmitter side.
The target FPGA is a Virtex-4 device, with a clock speed chosen around 370 MHz, to minimize the logic size of the design.
The presentation will demonstrate how straightforward it is to develop such a complex system within System Generator. On the one hand, Simulink provides an ideal framework to define the system architecture and parameters, and analyze the performances from a signal processing point of view. On the other hand, the Xilinx blockset allows to directly build the final hardware design with an efficient use of the specific FPGA features (DSP48 slices, RAM blocks and logic), without requiring too much hardware knowledge. Some solutions also exist to simply integrate pieces of Matlab code in the design (simple but even complex functions).
The presentation will show the complete design flow starting from the System Generator model down to the FPGA bitstream, and also indicate how to use hardware co-simulation for design verification or to speed up simulation time during the development phase.
Bluetooth EDR- from system specification to silicon with Simulink
Speaker:Andreï Tudose, Philips SemiconductorsThe gap between the standard specification and the silicon implementation of an integrated circuit/system is filled within all companies with a mixture of hard work, design flow, creativity, reviews and definitely design tools.
Every single component of the mixture may severely impact the time to market or the yield of a solution. The design tools in particular can make the difference between success and failure.
At Philips Semiconductors' Product Line Wireless Personal Area Network we've selected Simulink as the system simulation tool for our Enhanced Data Rate Bluetooth system design and also for checking the bit-true DSP compatibility with the final RTL implementation.
As the project advanced, the modelling and simulation techniques have been refined according to the needs of each step.
We started making use of Simulink's built-in models, continued designing each block in floating-point and finished by converting the design into fixed-point. The very final task was the bit-compatibility check of each RTL-block against its Simulink model.
Each step was accompanied by lots of analyses involving transmit masks, eye diagrams, DEVMs, BERs, PERs, interferer rejection, frequency offset and drift and much more.
Model-Based Design and Wireless Component Verification
Speaker: Mike Woodward, The MathWorksWireless standards have become more and more complex over the last decade or so. It is well-known that this complexity has caused considerable problems for system designers, but it is perhaps less well-known that test engineers have also been significantly impacted. Test engineers now face several issues; many of the new standards specify standards-specific measurements, measurements such as CCDF require sophisticated signal processing to take place, and the waveforms being tested are more complex. Economic forces are encouraging companies to be first to market, but this may put the system and component manufacturer ahead of the test equipment companies, which causes more problems. In the production environment, speed and accuracy of testing are important, but so is the ability to diagnose the cause of failures, which may call for component-specific diagnostic checks.
In this talk we will show how Model-Based Design can be applied to wireless component testing and verification. We will show how Model-Based Design enables the notion of an executable specification, which in turn can form the centerpiece of a verification strategy. We will show how Model-Based Design leads to significant additional flexibility in the testing process, enabling standard-specific and custom measurements and how the MathWorks tools enable diagnostic checks to be made. Finally we will bring these elements together to show how Model-Based Design can be used for verifying wireless components.
Implémentation d’un système de communication sans fil multi-canaux sur architecture mixte DSP/FPGA à partir de modèles Simulink/System Generator
Speaker: Etienne Fiset, LYRTECH/Japan’s National InstituteDomaine de recherche extrêmement actif au cours des dernières années, les systèmes multi-antennes (MIMO – Multiple Input Multiple Output) promettent un accroissement significatif des débits ouvrant la porte à l'expansion de l'offre des fournisseurs de services de télécommunications. Utilisant plusieurs antennes à l'émetteur et au récepteur, les systèmes MIMO éliminent, et tirent même avantage, des problèmes de dispersion spatiale des ondes électromagnétiques présents dans les systèmes à antenne unitaire (SISO – Single Input Single Output) utilisés jusqu'à maintenant.
Afin d'exploiter au maximum le potentiel du concept MIMO, le NICT (Japan's National Institute of Information and Communications Technology) devait à priori, dans le cadre du projet "4G Wireless High Atmosphere Platform System", détenir une bonne compréhension du canal radio MIMO et en effectuer une analyse complète. Entre autre, l'analyse de la direction de signal d'arrivée (DoA) était capitale pour bien caractériser l'environnement de propagation. Cette analyse devait fournir ultérieurement l'information critique sur la facon dont la puissance est distribué selon chaque angle d'arrivée (ie Spectre d'Azimut de Puissance, SAP).
NICT a opté pour une approche utilisant Simulink pour élaborer et implanter l'algorithme de DoA. En effet, le temps attribué à la migration de l'algorithme sur prototype hardware et les ressources humaines disponibles pour cette tâche étant limités, la possibilité qu'offre Simulink de porter un bloc diagramme directement sur hardware sans devoir écrire une seule ligne de code était d'une importance capitale. De plus, la haute cadence et l'importante quantité des traitements nécessitaient une plate-forme de prototypage à la fois puissante et flexible. Dans un tel cas, la génération automatique de code C et VHDL pour architecture mixe DSP/FPGA était idéale.
Day 2: Master class Track - Model-Based Design for Control
Les supports des Master Class seront en anglais mais les présentations en françaisMaster class 1: Automatically Generating Production Code for Control Systems
A key driver of the rapid adoption of Model-Based Design is the ability to quickly develop new algorithms and automatically generate code. Real-Time Workshop Embedded Coder is a key product for rapidly generating fast and efficient code that exactly duplicates the simulation behavior. This Master Class explains the primary workflow and key features of Real-Time Workshop Embedded Coder for developing and deploying code on an embedded controller. An illustrative example with several software-specific requirements will be reviewed and systematically prepared to generate code. Specific topics include data dictionary management, legacy code integration, generation of reusable code, hardware interfacing, operating system integration, and code formatting.
Master class 2: Controller Development for a Mechanical System using Simulink
In this Master Class, you will experience first-hand the Model-Based Design approach to develop a controller for a mechanical system, using a robot arm as an example. We will demonstrate the key steps of the design process, taking the mechanism modeled in a CAD package into Simulink, then designing and implementing the controller in real time. Through this example, you will gain experience with tools for plant modeling, parameter estimation, controller design, optimization, verification, validation, and rapid prototyping on a real-time system.
Master class 3: Electromechanical System Design Using Multi-Domain Modelling and Simulation
This Master Class demonstrates the use of Simulink, SimMechanics, and SimPowerSystems to model and simulate a mechanical linkage driven by an electric motor. A four-bar linkage is modeled using SimMechanics, then enhanced with automated computation of the mass and inertia properties, using MATLAB. Coordinate systems can be defined to speed up the modeling process and improve the model's clarity. The motor is modeled using SimPowerSystems, in this case as a permanent magnet DC motor. The interface between the linkage and motor models reflects the mechanical-electrical interaction, such as the torque that drives the motion of the linkage and the motor shaft velocity that defines the back EMF voltage associated with the motor.
