Voting Structures for Cascaded N-Modular Redundant Modules

Abstract

While the classic single-voter and N-voter architectures are well-established, there is more freedom in determining the voting structure for cascaded series of N-modular redundant modules. This paper investigates what kinds of freedom exist, how to identify efficient voting structures, and the effects of voting structures on the overall system reliability using cascaded triple modular redundant modules and cascaded quadruple modular redundant modules. The results reveal new voting structures which provide efficient trade-offs between hardware overhead and system reliability. For example, in the case of cascaded TMR modules, it is possible to combine the small hardware overhead of the classic single-voter architecture and the high reliability of the classic three-voter architecture when the module reliability and the voter reliability are not very different.