This paper deals mainly with the real-time implementation of a model-based Fault Detection and Isolation (FDI) scheme on a laboratory scale electric power system to detect various faults in the synchronous generator excitation control loop. The following types of faults have been considered: sensor faults, actuator faults, and load faults in the synchronous machines. The implementation of FDI has been carried out on a TMS320-C30 based real-time multi-channel processor. Satisfactory results have been obtained and reported in this paper. Furthermore, the paper has concluded that for the purpose of fault detection, a simplified synchronous machine model is sufficient. However, in order to isolate the fault, or to obtain the detailed information about the fault, or to estimate the severity of the fault, more detailed mathematical and physical models have to be used.