AbstractSimulations of nuclear magnetic resonance (NMR) signal from fluids contained in porous media (such as rock cores) need to account for both enhanced surface relaxation and the presence of internal magnetic field gradients due to magnetic susceptibility contrast between the rock matrix and the contained fluid phase. Such simulations are typically focussed on the extraction of the NMR T₂ relaxation distribution which can be related to pore size and indirectly to system permeability. Discrepancies between such NMR signal simulations on digital rock cores and associated experimental measurements are however frequently reported; these are generally attributed to spatial variations in rock matric composition resulting in heterogeneously distributed NMR surface relaxivities (ρ) and internal magnetic field gradients. To this end, a range of synthetic sediments composed of variable mixtures of quartz and garnet sands were studied. These two constituents were selected for the following reasons: they have different densities allowing for ready phase differentiation in 3D μCT images of samples to use as simulation lattices and they have distinctly different ρ and magnetic susceptibility values which allow for a rigorous test of NMR simulations. Here these 3D simulations were used to calculate the distribution of internal magnetic field gradients in the range of samples, these data were then compared against corresponding NMR experimental measurements. Agreement was reasonably good with the largest discrepancy being the simulation predicting weak internal gradients (in the vicinity of the quartz sand for mixed samples) which were not detected experimentally. The suite of 3D μCT images and associated experimental NMR measurements are all publicly available for the development and validation of NMR simulation efforts.