We report an experimental investigation of aeolian sand ripples, performed both in a wind tunnel and on stoss slopes of dunes. Starting from a flat bed, we can identify three regimes: appearance of an initial wavelength, coarsening of the pattern and finally saturation of the ripples. We show that both initial and final wavelengths, as well as the propagative speed of the ripples, are linear functions of the wind velocity. Investigating the evolution of an initially corrugated bed, we exhibit non-linear stable solutions for a finite range of wavelengths, which demonstrates the existence of a saturation in amplitude. These results contradict most of the models. The surface of aeolian sand dunes is not smooth but is usually formed into regular patterns of ripples, transverse to the wind [1]. Their wavelength ranges from the centimeter to the meter with a constant aspect ratio (≃ 4%) [2]. Although many different models have been proposed to explain the formation and evolution of aeolian ripples [3,4,5,6,7,8,9,10,11], few field observations [1, 2, 12] and controlled experiments [4,13,14] have been performed so far. By contrast with subaqueous dunes or ripples which result from a hydrodynamic instability induced by the interaction between shape and flow [1], aeolian ripples are of different nature and result from a screening instability. When the 'saltons' -high energy grains -collide the bed, they eject grains of smaller energy, 'reptons'. The windward slope of a small bump is submitted to more impacts than the lee slope, so that the flux of reptons is higher uphill than downhill and makes the bump amplify.Most of the models agree for the linear stage of the instability -see [5] for a pedagogical review. They assume that the reptons remain trapped by the bed after a single hop of length a, distributed according to a distribution P (a). The saltons are considered as an external reservoir which brings energy into the system. On this basis, the most unstable wavelength λ 0 is found to scale on the reptation lengthā = da a P (a). Besides, the most recent investigations of sand transport, both experimental [15,16] and theoretical [17,18], indicate that P (a) is independent of the wind shear velocity u * , which implies thatā scales on the grain diameter d. The initial wavelength of the ripples λ 0 is thus expected to be independent of the wind strength.