White matter diffusion MRI enables non-invasive estimation of the axon diameter distribution, which is of interest as it modulates communication speed and delays between brain regions, and changes during development and pathology. Distribution mapping is challenging: current methods simplify it by either estimating the mean diameter, imposing parametric distributions, or combining non-parametric approaches with Double Diffusion Encoding. We present a non-parametric framework based on a PGSE protocol. Simulations show robust reconstruction of unimodal and bimodal distributions. The method is sensitive to population specific changes within bimodal distributions, as long as the underlying populations are separated by a minimum distance.