An effective and efficient diagnosing scheme is essential for mechanical fault detection. At the same instance simple, robust and cost effective platform is necessary for developing the diagnosing schemes. However, regardless of being expensive and not feasible to perform always, vibration measurement is widely used to detect the fault conditions. This paper presents a low-cost laboratory based research platform for induction motor mechanical fault diagnosis from stator current using fast Fourier transform (FFT). In this work normal, natural cage defective and natural ball defective bearings are considered to detect faults using spectral analysis and verified with the results of theoretical models proposed in literature. In addition, a comparative spectral analysis is performed for the signals acquired at different sampling rates. The bearing fault frequencies are defined to identify the faults at given load. Experimental results showed that amplitudes of harmonics in the side bands increase with respect to increase in load. The bearing fault frequencies are investigated with actual and theoretical frequencies to demonstrate actual frequencies could improve detection performances. © 2017 IEEE.