Characterization of electromagnetic interference (EMI) between two powerelectronic systems is an important activity in electromagnetic compatibility(EMC). However, EMC problems are currently treated after the production ofa prototype, which causes additional costs and significant delays if therequired standards are not reached. In order to consider the compliance toEMC standards from the very beginning of the design phase, a method hasbeen developed at Laboratoire Ampère to predict near-field coupling betweencomplex components. This predictive method is based on the multipolarexpansion in spherical harmonics of the field close to the element. Thiselement can then be represented by an equivalent punctual source composed ofthe elements of a multipolar expansion, which allows the computation of thenear-field coupling with other elements.The mutual inductance, which represents the magnetic coupling between twoobjects, can be computed once the coefficients of the multipolar developmentare known. To this purpose, a dedicated measuring bench has been designed inLaboratoire Ampère: the multipolar expansion is obtained by the measurementof magnetic field all around the source. Ideally, these field should be measuredin the real environment, with all other elements present, but not excited.However, this is not always practical. In the proposed approach, all the usefulinformation is obtained from measurement of the isolated element. Then themutual inductance can be computed for any set of element positions using thevector-spherical-harmonic addition theorem. Nevertheless, as a result of theabsence of the second source during measurements, the mutual inductancecomputed in this method can possibly differs from the theoretical value. Thisis mainly due to the coupling of second order in the system and the conductingmetal in the secondary element.In this paper, we first expose how the induced currents can sometimes affectthe computed mutual impedance. We propose to study a canonical problem inwhich two systems, composed of a loop source and a passive one, arepositioned close one from the other. The study of the error between thecomputed and the theoretical value of the mutual impedance show that theerror depends on several parameters such as the position of each element, theshape of the conducting metal, etc., which means that the proposed approachcannot always be applied. In the end, a corrective method using the samemeasuring bench is proposed to take into account the effects of inducedcurrents.