The Giant MagnetoImpedance (GMI) eff ect refl ects the strong dependence of the electrical impedance of a ferromagnetic conductor submitted to an external applied magnetic fi eld. The main objective of this thesis consists in increasing the noise performance of GMI-based magnetometers, since it ultimately limits the lower magnetic fi eld which can be detected. We thus present a study of the material's physical properties, which allows us to evaluate the proper sensitivity of a sensing element, using a simpli ed model. The originality of this study lies in the consideration of the sensitivity in units of V/T, which is linked to the optimum excitation current amplitude and frequency, as a criterion of performance. In order to further increase the sensitivity, GMI devices are associated to a detection coil. The field response of the two-port network thus obtained is modeled. The parasitic capacitance, identi ed as a limiting factor, thus implies an optimum excitation frequency associated to each devices. Finally, a complete model of the noise performances, including the electronic conditioning, is developed. It allows us to predict and analyze the performances associated to diff erent electronic conditioning, particularly in relation to the demodulation approach. Those results validate the usefulness of the two-port network structure and shed a new light on the choice of the optimal static fi eld operating point of the sensor. This work thus leads to the design of an optimized magnetometer, exhibiting excellent performances, notably an equivalent noise level lower than 1 pT/sqrt(Hz) in white noise region.
