A new method has been developed for the detection and diagnosis of broken rotor bars faults in three-phase induction motors under no-load conditions. Early detection of faults is made by using a sliding window constructed by Hilbert transforms of one of the phases of the thee-phase currents and the size of a fault is diagnosed by motor current signature analysis (MCSA) of the stored Hilbert transforms of several periods of one-phase current. The information entropy of a symbol tree generated by each sliding window is used as a fault index. The method was tested using healthy and damaged 0.37 kW induction motors under no-load conditions with applied voltages ranging from 220 V to 380 V. One and two broken rotor bars were detected under no-load conditions when supply voltages were 260 V and above. The results indicate that the method yields a high degree of accuracy in fault identification. (C) 2010 Elsevier Ltd. All rights reserved.