This thesis describes the design of multivariable excitation controllers for synchronous generators, and their practical implementation on a laboratory machine system. The main objectives of the investigation were aimed at extending the steady-state stability limit of the generator and improving its general controllability. The present work is principally concerned with the implementation and test of the designs. Instrumentation has been developed to measure and combine feedback signals; this instrumentation and the laboratory generator system are described in detail. The performances of the controllers are predicted by computer simulation and evaluated by practical testing; attention is focussed on comparison of predicted and experimental performances. The theoretical basis of the design strategy is described, in which an optimal control algorithm applies the system states as feedback signals to the open-loop system, thereby shifting certain of the system eigenvalues to more desirable, predetermined locations. The state variables are selected as physically measurable quantities which obviates the need for implementing state estimation techniques.