One potential method of improving the reliability and accuracy of instrumentation is to employ a microprocessor as an intelligent voter, or combiner, within a redundant measurement system. Several measurement channels are used to monitor the same quantity and the processor forms a single, more accurate and reliable, output from the information gained from all of the channels. A microprocessor makes it possible to implement combining techniques that are not feasible using conventional hardwired circuitry. A number of such combining algorithms are identified and their performance, in the presence of varying levels of additive Gaussian noise, is analysed. Several methods are proposed for reducing systematic error by using the information that can be gained from comparing channels. It is shown that the probability of being in an undetected failed state can also be improved. The trade offs between performance and the resources that are required to implement the proposed techniques, are thoroughly investigated. Empirical performance is compared with theoretical predictions and practical difficulties are identified.