Embedded Systems - Exam - PDF

Introduction Recordings available on webeep. Prof. William Fornaciari Exam: last lecture of December, written test. EMBEDDED SYSTEMS Embedded systems are specialized computing systems designed to perform specific tasks or functions within larger mechanical or electrical systems. Unlike...

Introduction Recordings available on webeep. Prof. William Fornaciari Exam: last lecture of December, written test. EMBEDDED SYSTEMS Embedded systems are specialized computing systems designed to perform specific tasks or functions within larger mechanical or electrical systems. Unlike general-purpose computers, they are typically tailored to perform a particular operation or a set of operations, often with real-time constraints. These systems are embedded as part of a complete device, often hidden from the user, and control the functions of hardware components. In the last few decades, general public computing has switched from stand-alone desktop computers to mobile devices connected to the cloud, addressing the new usages and needs created by smartphones and similar devices that allow people to be connected at all times and get access to a huge amount of information. Given that Dennard scaling has ended (transistor performance improves as transistors shrink in size), that Moore's law is reaching an end (the number of transistors on a microchip doubles approximately every two years, leading to a consistent increase in computing power and a corresponding decrease in cost per transistor), and that the cost involved in developing new technology nodes is skyrocketing, new directions should be investigated to continue improving the performance of storage and communication units. This represents a brand- new opportunity for Europe to demonstrate its creativity and to invent solutions that break away from current advances relying on technology improvements. At a more global level, we observe a shift within the ICT domain from compute centric to data centric, from applications (triggered by big data and also deep learning) down to hardware: we are beginning to realize the cost of moving data, hence the striving for "computing in memory" or having computing near data, which translates at all levels, from chips to systems with edge computing. The current era can be defined as "More computing at the edge for improved safety, privacy, and cost of ownership". Computing is becoming a continuum, from sensors, data fusion, processing, storage, communication where data is progressively refined into useful information. In the computing continuum, embedded systems form the backbone of edge computing, handling specific, low- power, real-time tasks. They provide critical local processing while feeding data to more powerful systems for deeper analysis and long- term storage. Embedded systems are fundamental to bridging the physical world with the digital world, making them a key component in modern computing ecosystems. In short, nowadays applications are pushing and technology is invading all aspects of our life, but we also are about to reach a plateau in technological evolution: - CMOS technology is arriving to the limit. Let's start considering the first element of the compute-communications-storage computing systems triangle: computing units. With CMOS scaling now reaching its absolute limits, new technologies are on the horizon, but these are more likely to exist alongside CMOS than replace it. In fact, improvements thanks to shrinking transistors had already slowed down or disappeared by 2005, as there is a physical limit for which junctions can no longer be controlled in standard structures. Instead, in order to achieve even smaller dimensions, major changes have been made to the structure of circuit elements, with the result that they are far more complex. - Cost of new basic technologies is sometimes unaffordable. - Power/energy wall. Because of the ever increasing number of transistors integrated per unit-area, demanding larger power consumptions and higher thermal dissipation. The limitation in power also means that not all the circuit elements can work at the same time, leading to the concept of dark silicon (areas of the chip temporarily off at any given time). - Data proliferation. Refers to the rapid and widespread growth of data generated, collected and stored by individuals, businesses and systems. Handling the enormous volume, variety and velocity of data has become increasingly complex and organizations face difficulties in storing, categorizing and retrieving data efficiently. - Hard to move from invention to innovation. - Design methodology is exploiting humans (fortunately). Application needs They will be compute-intensive. They will require efficient hardware and software components, irrespective of their application domain: embedded, mobile or datacentre. They will be connected to other systems, wired or wireless, either always or intermittently online. In many cases they will be globally interconnected via the internet. They will be entangled physically, meaning that they will not only be able to observe the physical environment that they are operating in, but also to control it. They will become part of our environment. The virtual, digital world and the real, physical world are being connected in the Internet of Things (IoT) and in Cyber-Physical Systems: we have entered an era where the traditional computing system, recognizable by the keyboard and screen as interfaces, is being complemented and to some extent supplanted by mobile computing models characterized by machine-to-machine communication, and comprising a vast array of sensors and actuators. ○ In the model of the IoT, smart sensors in the environment communicate via gateways, or specialized computing devices, with remote servers in the cloud. They generate an enormous amount of data, which is analysed to extract information to provide new and better services. An IoT system is a distributed system composed of a number of physically separated, communicating devices which do not usually involve a human in the loop. Cyber-physical systems take the integration with the physical world a step further by directly interacting with the physical Embedded Systems Pagina 1  ○ Cyber-physical systems take the integration with the physical world a step further by directly interacting with the physical world based on the results of data analytics. They are characterized as having an actuator that directly affects the physical world. They will be smart, able to interpret data from the physical world even if that data is noisy, incomplete, analogue, remote… cognitive computing is making the interface, often driving big-data analytics and data mining. We are now in the third era of Computing, and over the years, driving metrics have changed according to the different eras. The evolution in the industry driving metrics could be: Functionality: up to 1980s, supercomputers and mainframes; Cost: 1990s, personal computer; Power and cost: 2000s, notebooks; Portability and energy: 2010s, mobiles. The dark silicon problem refers to the phenomenon where, due to power and thermal constraints, a significant portion of a chip's transistor cannot be powered on simultaneously at their maximum frequency. As a result, part of the silicon must remain inactive or "dark" to avoid overheating or exceeding power budgets. This problem has become more prominent as transistor density continues to increase, while the ability to dissipate heat and manage power consumption has not kept pace. This is cutting the evolution of single core processing units and is opening the era of multi-core processors. Once we understand that the path to take is that of multi- Core processors, we need to figure out how to make them communicate with each other at high speed, allowing a large data exchange between them. Network-on-Chip is a network-based communication infrastructure designed for integrated circuits, using network-like principles to connect various components, such as processors, memory blocks, and peripherals on a single chip. It replaces traditional bus-based or point-to-point interconnect architectures with a scalable and efficient communication network, improving performance and scalability. Embedded systems 10 years from now are expected to be: Networked: from working in isolation towards communicating, networked, distributed solutions. Networked embedded systems will be interconnected through various communication protocols, forming part of the Internet of Things. Secure: systems are threatened by enormous security issues, challenging its technical and economic viability. Security in embedded systems will be paramount, addressing vulnerabilities from increased connectivity and data sharing. Complex: embedded systems will become increasingly sophisticated, incorporating advanced algorithms and capabilities to handle diverse applications. They will achieve giga-complexity through nanotechnology, enabling unprecedented performance and functionality. They will be increasingly heterogenous, integrating diverse components such as sensors (including biosensors, MEMS, and NEMS), actuators, and advanced interactive displays. These systems will also incorporate novel input devices like speech and handwriting recognition. With more emphasis on software-driven solutions and reconfigurable architectures, embedded systems will become tailored to specific application domains. Finally, advancements in communication protocols, RF technologies (use of radio frequency signals to wirelessly transmit and receive data over a large range of frequencies), and standards will ensure seamless connectivity and data exchange across devices. Low power: prioritize energy efficiency, crucial for battery-operated and energy-sensitive applications. MEMS: Micro-Electro-Mechanical Systems. MEMS technology integrates mechanical and electrical components on a microscale, allowing for the creation of compact, smart devices. MEMS are three-dimensional structures created using integrated circuit fabrication technologies and micromachining, typically on silicon or glass wafers. MEMS devices include transducers, microsensors, microactuators, and mechanically functional structures like microfluidic valves, pumps, and microengines. Integrated microsystems combine MEMS components with circuitry to autonomously perform tasks or interface a host computer. MEMS components provide interface to the non-electrical world: ○ Sensors capture inputs from the physical world; ○ Actuators provide outputs to non-electronic events. Integrated: combining multiple functionalities and components into single packages. Embedded Systems Pagina 2