Integration-aware Modeling, Simulation and Design Techniques for Smart Electronic Systems

Smart electronic systems represent a vast category of energy-autonomous and ubiquitously connected systems that incorporate analog, digital and MEMS components, combined with various kinds of sensors, actuators, energy storage devices and power sources. Smart systems generally find applications in the worldwide market for “Monitoring & Control” products and solutions, hence they are used in a broad range of sectors, including automotive, healthcare, Internet of Things, ICT, safety and security, and aerospace. In order to support such wide variety of application scenarios, smart systems integrate a multitude of functionalities, technologies, and materials. The design of smart systems is therefore a complex and major multidisciplinary challenge, as it goes beyond the design of the individual components and subsystems. New design and simulation methodologies are fundamental for exploring the design space in order to find the most efficient trade-off between performance and involved resources, and for evaluating and validating system behavior taking into account the interactions between closely coupled components of different nature. Current system level design methods must indeed accurately manage increasing system complexity and interaction effects between the environment and the system and among the components. Nevertheless, the involved components are usually described using different languages, relying on different models of computation, and need to be jointly simulated at various abstraction levels. This dissertation aims at bridging this gap focusing on novel integration-aware solutions for different aspects of a smart system: the design of digital subsystems and components, the modeling of batteries, and the power estimation of smart systems at system level of design abstraction. Although the design flow of digital components is well consolidated and highly standardized (e.g., commercial, fully automated synthesis & optimization tools, technology libraries, etc.), additional integration-aware design constraints arise due to the interaction of components of different technological domains and to the harsh environment where smart systems typically operate. This work presents a methodology for addressing these new constraints, thus enhancing the design of digital components. As a partial fulfillment of such constraints results in a global degradation of performance, the proposed methodology focuses on the effects rather than the physical sources of the constraints. This allows to move from the typical RTL to a system level of abstraction, i.e., SystemC TLM, obtaining a faster validation of the performance of digital subsystems. Energy efficiency is becoming increasingly important for self-powered smart electronic systems, as the amount of energy they can gather from the environment or accumulate in storage devices cannot be considered constant over time. Power supplies have therefore a very heterogeneous nature: depending on the application, more than one type of power source (e.g., photovoltaic cells, thermoelectric or piezoelectric energy generators) and storage device (e.g., rechargeable and non-rechargeable batteries, supercapacitors, and fuel cells) could be hosted onto the system. As a matter of fact, no single power source could provide the desired level of energy density, power density, current, and voltage to the system for all possible workloads. Batteries are being significantly used in smart electronic systems due to the their increased energy capacity, improved production process, and lower cost over the last years. However, a battery is an electrochemical device that involves complicated chemical reactions resulting in many non-idealities of its behavior. Therefore, a smart system designer has to characterize these non-idealities in order to accurately model how the battery delivers power to the system. This dissertation introduces a systematic methodology for the automatic construction of battery models from datasheet information, thus avoiding costly and time-consuming measurements of battery characteristics. This methodology allows generating models for several battery charge and discharge characteristics with tunable accuracy according to the amount of the available manufacturers’ data, and without any limitation in battery chemistry, materials, form factor, and size. Finally, this work introduces a modeling and simulation framework for the system level estimation of power end energy flows in smart systems. Current simulationor model-based design approaches do not target a smart system as a whole, but rather single domains (digital, analog, power devices, etc.), and make use of proprietary tools and pre-characterized models having fixed abstraction level and fixed semantics. The proposed methodology uses principles borrowed from the system level functional simulation of digital systems and extends them for simulating the behavior of subsystems whose functionality is to generate, convert, or store energy (e.g., power sources, voltage regulators, energy storage devices, etc.). This has been done at system level using standard open-source tools such as SystemC AMS and IP-XACT, which allow to explicitly represent current and voltage similarly to digital logic signals. The implemented approach facilitates virtual prototyping, architecture exploration, and integration validation, with high flexibility and modularity.