In this study, the application of S-transform with a frequency-dependent window is proposed for distinguishing internal incipient faults in transformers. The S-transform is applied to the recorded input current waveforms during the impulse test. Then, time–frequency contours of the S matrix are plotted and a suggested index is computed, accordingly. Faulty and healthy windings can be discriminated using time–frequency contours and proposed decision index. A 66-kV/25-MVA interleaved transformer winding is modelled in a MATLAB environment, also an arc discharge model is used and the effectiveness of the proposed method is studied. Laboratory tests are performed on this transformer winding and the performance of the method in noisy environments is investigated. According to the laboratory test and simulation results, it can be said that the suggested S-transform-based method is capable of detecting incipient faults conveniently.