Fault-tolerance is an important design requirement in critical sensor applications. It is conventionally achieved by using redundant components, which increases system’s size, cost and complexity, that are often sacrificed or comprised due to associated limitations. This paper proposes a novel solution to achieve sensor fault-tolerance at the system level instead of the usual approach that targets the component level. The architecture consists of multi-functional sensors which are used to replace conventional single mode sensors, and a data fusion algorithm which provides online test and fault-tolerance. This method has the potential to significantly increase system reliability and supports a reduction in the overheads inherent with the use of redundancy in fault-tolerant systems. A MEMS humidity/pressure sensor has been designed as an example to support the method. The sensor has simple structure, good linearity and sensitivity, and the potential of further integrating a temperature function.