A highly dependable, analog multi-core mixed-signal task distribution system

The objective of this thesis is to develop an analog architecture, which distributes decentralized, highly dependable and self-reliant tasks within a mixed-signal multi-core System-on-Chip. Hence, every step of the design process and their results to develop such an analog system are presented, since the synthesis of analog circuits is still mostly done manually, contrary to the almost fully automatized and formalized design of digital circuits. Especially the validation and long-term satisfaction of the specification of the analog components need a high degree of verification and testing done by hand. The usability of the design is evaluated against known task distribution approaches, which are either highly sophisticated digital circuits or software implementations, which have shown their usability in real-world applications already. Further, evaluations are done against analog approaches, which are capable of being transformed to match the objectives of this thesis. This allows to clearly differentiate the developed and designed architecture from the existing approaches. The design of an artificial, analog hormone system is a bio-inspired replica to distribute information and tasks within a system. The endogenous transmitters are mapped to voltages and currents, which are, properly coordinated, spreading throughout the full system. Those hormones can be applied locally, but are noticed globally at every core. However, the physical laws of electrical engineering have to be taken into consideration as equally as the balance of the hormones to guarantee the reliable and dependable functionality of the task distribution. Within the design process a complete formal description of the analog components of the hormone system is done. Based on the description, solving the differential equations and inequalities, which mirror the behavior of the hormone system, enables to issue a reliability analysis. With this analysis the fail-safety of the components are classified, the dependencies of the circuit parameters are determined and a set of specification, needed for the design process, are derived. Further, a robustness value has been defined, which quantifies the interval of