A new method to estimate the complex pupil-plane field from a functional derivative of the optical transfer function (OTF) has been described by Codona.1 It is in principle a diversity technique that uses two focal-plane images with a spatially localized difference introduced in the pupil. The difference can be in phase, amplitude, or both, and need not be well known provided that its spatial extent is small compared to the required resolution of the field. Unlike other diversity methods, however, the dOTF wavefront estimation algorithm is non-iterative and very fast. The technique may be exploited in a straightforward way for tomographic wavefront sensing from a random scene of unresolved sources whose angular extent exceeds the isoplanatic limit. A single camera is arranged to capture the whole scene. An exposure is recorded, a modification is made to the pupil, and a second exposure recorded. These two images are then sufficient to derive the 3D information to characterize the non-shift-invariant point-spread function over the full field. While offering nothing revolutionary in tomography, the technique is a simple way to characterize the aberrations in an optical system and localize them along the optical axis.