Quick detection of a small initial fault is important for an AC motor drive system to prevent a consequent large fault. Papers for the detection have been proposed, for example, by the park vector (PV), the AI techniques, the wavelet analysis, and the negative sequence analysis approaches. This paper proposes a diagnostic method of a stator winding fault of an induction motor by combination of the PV and the negative sequence analysis approaches. The principle consists of two steps. The first step is extraction of a negative sequence admittance component as the diagnosis index because the admittance is almost independent from the slip ratio. The instantaneous positive components are calculated by division between the PVs of the stator voltages and currents. The negative component is extracted as its ripples. The second step is estimation of fault phases and number of turns from the negative component as a complex value. It is possible to find a negative-sequence map which shows a useful relation between fault phases and numbers of turns in multiple phases. Diagnostic simulation results are shown with a real-time simple motor model written in a simulation language of VHDL-AMS. Their experimental results are also shown and the proposed method is investigated and validated.