In this paper a phase detection method for fault diagnosis of the induction motors has been presented. The proposed method has a powerful environmental noise suppression capability. It has been shown in literature that the performance of the previously used fault detection method (instantaneous power analysis) was affected by the environmental noise, switching disturbances and other low order harmonics. The instantaneous power analysis yields erroneous results under low load conditions of the motor where fault signature was buried in the noise. It has been theoretically and experimentally shown that the proposed phase detection method can detect fault signatures in the noisy environment without use of any extra hardware. The accuracy of the proposed phase detection method was compared with the instantaneous power analysis method for bearing ball defects and the results on the real hardware implementation confirm the effectiveness of the proposed approach. © 2014 IEEE.