ABSTRACTFuture global astrometry missions have targeted the determination of positions, parallaxes, and annual proper motions to the 10 μas level. This can be achieved through the use of fringe imaging interferometers, possibly featuring wide fields of view (e.g., Fizeau configurations). The basic location information is to be extracted from the fringe pattern by proper implementation of the detection system and proper exploitation of the focal plane data. The sampling resolution requirements are a key trade‐off issue between science and engineering: therefore, fringe acquisition by means of realistic detectors and the resulting accuracy in photocenter location is discussed herein. The location performance is described as a χ2 minimization problem; the resulting expressions are then evaluated in analytical form and by means of a Monte Carlo simulation, whch provide good agreement. In order to achieve the limiting interferometer accuracy, 8–10 pixels per fringe period are required, whereas a sampling resolution of 4–5 pixels per period provides a 30% degradation. We evaluate the location accuracy degradation induced by progressively reduced fringe visibility and increasing noise level. The former provides a smooth performance reduction, acceptable to a wide extent; read‐out noise is critical because the fringe pattern signal is recorded over many pixels, each providing a comparable contribution to the overall noise.