The paper presents methodologies to detect and locate short-circuit faults on the stator winding of VSI-fed PM brushless dc motors. Normal performance characteristics of the motor are obtained through a discrete-time lumped-parameter network model. The model is modified to accommodate shortcircuit faults in order to simulate faulty operation. Fault signatures are extracted from the waveforms of electromagnetic torque and phase-voltage summation using wavelet transform. Three independent detection techniques are introduced. Experimental measurements agree acceptably with simulation results, and validate the proposed methods. This work sets forth the fundamentals of an automatic fault detector and locator, which can be used in a fault-tolerant drive.