Design and verification approaches for reliability and functional safety of analog integrated circuits

New breakthroughs in semiconductor design have enabled a rapid integration of semiconductor chips into systems that affect all aspects of the society. Examples of emerging systems include spacecraft, Internet of Things (IoT), intelligent automotive, and bio-implantable devices. Many of these systems are mission-critical or safety-critical, meaning that failure or malfunction may lead to severe economical losses, environmental damages or risks to human lives. In addition to performances improvement, the reliability and functional safety of the underlying integrated circuit (IC) have attracted more and more attention and have posed grand challenges for semiconductor industries. This dissertation introduces an approach for high performance voltage reference design and investigates two subjects that improve the reliability and functional safety of analog circuits. The first part of this dissertation studies design strategies of a low temperature-coefficient voltage reference generator, which is a fundamental building block and determines the maximum achievable performance of almost all analog/mixed-signal systems. The proposed method is targeted at extracting a physical quantity of the silicon bandgap, and has the potential of designing a voltage reference that has qualitatively better temperature dependence. An implementation of the proposed approach in GlobalFoundries 130nm process shows that the design can achieve temperature coefficients as low as 0.7ppm/°C over a temperature range of -40°C to 125°C over all process corners. The second part of this dissertation focuses on multi-states verification of analog circuits. The multiple DC equilibrium points or multi-states problem traces back to IC design. It is a well-known problem in many basic self-stabilized analog circuits because of the existence of positive feedback loops (PFLs). This work proposes systematic and automatic approaches for locating all PFLs to identify circuits vulnerable to undesired equilibrium states and methods for automatically identifying break-points to break all PFLs in the vulnerable circuits. The proposed methods make it possible to efficiently identify a circuit’s vulnerability to undesired operating points by considering circuit topology only, without the need for finding all possible solutions to a set of simultaneous nonlinear equations which is an open problem with no solution. Moreover, the automatic break-points identification enables easy use of homotopy analysis to guarantee absence of undesired states. The third part of this dissertation focuses on fault coverage simulation of analog circuits. This work describe two methods, one is to reduce the fault coverage estimation time and the other is to improve the fault coverage for analog circuits. The first method incorporates graph theory and sensitivity analysis and leads to dramatic reduction in fault coverage simulation time by 10’s of times for a moderately sized analog circuit. The second method discusses a systematic test-points selection technique to improve the analog fault coverage with simple DC tests and a concurrent sampling technique for monitoring these points. This work could be applied to manufacturing testing or for real-time fault detection