Abstract : Nuclear fission is the process in which a heavy nucleus splits into two nuclei called fission fragments. These excited fragments emit prompt neutrons and gammas to reach their ground or a metastable state. The energy released during fission is used in nuclear power plants to provide electricity. The nuclear reactor lifespan and control depend partly on fission observables such as the fragment mass yields, the neutron and gamma multiplicity and energy. The first part of this thesis is dedicated to the study of the fission fragment de-excitation process with the FIFRELIN Monte-Carlo code. This code, containing many nuclear models describing the fragments, predicts the prompt particle properties (multiplicity, energy). During this PhD work the impact of different models on the code predictions was investigated. The models which are studied are those defining the initial angular momentum, the level density and the radiative strength function of fragments. These studies identified those nuclear properties that have a significant impact on the code predictions, and as a consequence permitted selection of the model combination suited to reproduce the maximum number of fission observables, and improving our understanding about the nuclear description of fission fragments. These studies are firstly performed on the spontaneous fission of 252Cf for which numerous experimental data are available allowing constraints to be set on the models. Studies on fast fission (incident neutron with an energy around MeV) of 238U and 237Np are then performed. They are motivated partly by the development of new fast reactor concepts, called Generation IV reactors, that will reduce the quantity of nuclear waste and burn the large amount of natural uranium to provide electricity. Experimental data concerning fast fission are scarce. As a consequence new experimental setups are currently under development. They will measure fission observables over a large excitation energy domain. One of them, called FALSTAFF, is the topic of the second part of this thesis. In the second part of this PhD thesis, the different steps concerning the development, the optimization and the characterization of the first arm of the FALSTAFF setup are presented. This spectrometer, placed at the NFS facility (Neutrons For Science), will study fast fission in direct kinematic for various actinides. The detection of both fragments in coincidence will allow characterization of their energy, their masses (before and after neutron evaporation) and their nuclear charge. The neutron multiplicity will be deduced and will provide information on the energy sharing between the fragments. This data will be later used as input for the FIFRELIN code. The velocity determination of both fragments (2V method), with MWPC-SeD time-of-flight detectors, will provide pre-neutron emission fragment masses. An axial ionization chamber, placed after those detectors, will measure the kinematic energy and the energy loss profile of fragments, which will provide respectively the post-neutron emission mass (EV method) and the nuclear charge. The experimental part of this thesis is dedicated to the development of simulation and data analysis software, the configuration of the data acquisition system, the characterization and the optimization of the detectors. This last item mainly concerns the axial ionization chamber.