The nonlinear curvature wavefront sensor (nlCWFS) offers improved sensitivity for adaptive optics systems compared with existing WFSs, such as the Shack–Hartmann. The nominal nlCWFS design uses a series of imaging planes offset from the pupil along the optical propagation axis as inputs to a numerically iterative reconstruction algorithm. Research into the nlCWFS has assumed that the device uses four measurement planes configured symmetrically around the optical system pupil. This assumption is not strictly required. In this paper, we perform the first systematic exploration of the location, number, and spatial sampling of measurement planes for the nlCWFS. Our numerical simulations show that the original, symmetric four-plane configuration produces the most consistently accurate results in the shortest time over a broad range of seeing conditions. We find that the inner measurement planes should be situated past the Talbot distance corresponding to a spatial period of r₀. The outer planes should be large enough to fully capture the field intensity and be situated beyond a distance corresponding to a Fresnel-number-scaled equivalent of Z ≈ 50 km for a D = 0.5 m pupil with λ = 532 nm. The minimum spatial sampling required for diffraction-limited performance is 4 to 5 pixels per r₀ as defined in the pupil plane. We find that neither three-plane nor five-plane configurations offer significant improvements compared with the original design. These results can impact future implementations of the nlCWFS by informing sensor design.