A system-level supervisory approach to mitigate single event functional interrupts in data handling architectures.

This thesis examines the effective mitigation of the Single Event Effects (SEEs) in commercial State-Of-The-Art (SOTA) data handling devices to provide high performance and fault-tolerant data handling architectures for space missions. It concentrates upon Single Event Functional Interrupts (SEFIs), whereby a single particle hit in sensitive device cross-section leads to unexpected device behaviour. Reports of SEFIs are increasing in all key data handling technologies, e.g. memories, microprocessors, field programmable gate arrays (FPGAs) and on-board local area networks (LANs). Constructing a high performance on-board data handling (OBDH) architecture will therefore require a large number of resources to cope with problem of SEFIs. This research proposes an architectural/system-level approach to SEFI mitigation, where a global supervisor is added into the architecture to monitor heterogeneous OBDH units. In the proposed OBDH architecture, all units are connected through a router-based Space Wire network. A supervisor is a radiation hardened microprocessor, which is part of the router unit. The supervisor expects to receive special detection and diagnosis (DAD) packets from the underlying units. Health information collected through these packets is compared with designer's input to produce any fault signature. The supervisor intervenes when the state of a unit does not match expectations or DAD packets stop arriving. In such an event, the supervisor will apply a recovery procedure based on fault signature observed and any previous recovery record for that unit Theoretical and experimental analyses are presented to establish practicality of the scheme. The outcome of this thesis is a SEFI-tolerance methodology aimed at applications that demand for SOTA commercial technologies and increased availability but cannot afford high cost and resources associated with traditional redundancy-based mitigations. This is particularly useful for small satellites where very limited mass, volume and power resources preclude the use of multiple-redundant system-based architectures. Therefore, it promises a measurable increase in small satellite utility across range of mission performance requirements