Moving Towards Analog Functional Safety

Over the past century, the exponential growth of the semiconductor industry has led to the creation of tiny and complex integrated circuits, e.g., sensors, actuators, and smart power systems. Innovative techniques are needed to ensure the correct functionality of analog devices that are ubiquitous in every smart system. The standard ISO 26262 related to functional safety in the automotive context specifies that fault injection is necessary to validate all electronic devices. For decades, standardizing fault modeling, injection and simulation mainly focused on digital circuits and disregarding analog ones. An initial attempt is being made with the IEEE P2427 standard draft standard that started to give this field a structured and formal organization. In this context, new fault models, injection, and abstraction methodologies for analog circuits are proposed in this thesis to enhance this application field. The faults proposed by the IEEE P2427 standard draft standard are initially evaluated to understand the associated fault behaviors during the simulation. Moreover, a novel approach is presented for modeling realistic stuck-on/off defects based on oxide defects. These new defects proposed are required because digital stuck-at-fault models where a transistor is frozen in on-state or offstate may not apply well on analog circuits because even a slight variation could create deviations of several magnitudes. Then, for validating the proposed defects models, a novel predictive fault grouping based on faulty AC matrices is applied to group faults with equivalent behaviors. The proposed fault grouping method is computationally cheap because it avoids performing DC or transient simulations with faults injected and limits itself to faulty AC simulations. Using AC simulations results in two different methods that allow grouping faults with the same frequency response are presented. The first method is an AC-based grouping method that exploits the potentialities of the S-parameters ports. While the second is a Circle-based grouping based on the circle-fitting method applied to the extracted AC matrices. Finally, an open-source framework is presented for the fault injection and manipulation perspective. This framework relies on the shared semantics for reading, writing, or manipulating transistor-level designs. The ultimate goal of the framework is: reading an input design written in a specific syntax and then allowing to write the same design in another syntax. As a use case for the proposed framework, a process of analog fault injection is discussed. This activity requires adding, removing, or replacing nodes, components, or even entire sub-circuits. The framework is entirely written in C++, and its APIs are also interfaced with Python. The entire framework is open-source and available on GitHub. The last part of the thesis presents abstraction methodologies that can abstract transistor level models into Verilog-AMS models and Verilog- AMS piecewise and nonlinear models into C++. These abstracted models can be integrated into heterogeneous systems. The purpose of integration is the simulation of heterogeneous components embedded in a Virtual Platforms (VP) needs to be fast and accurate.