Ultrasonic phased arrays have produced major benefits in a range of fields, from medical imaging to non-destructive evaluation. The maximum information, which can be measured by an array, corresponds to the Full Matrix Capture (FMC) data acquisition technique and contains all possible combinations of transmitter–receiver signals. However, this method is not fast enough for some applications and can result in a very large volume of data. In this paper, the problem of optimal array data acquisition strategy is considered, that is, how to make the minimum number of array measurements without loss of information. The main result is that under the single scattering assumption the FMC dataset has a specific sparse structure, and this property can be used to design an optimal data acquisition method. An analytical relationship between the minimum number of array firings, maximum steering angle and signal-to-noise ratio is derived, and validated experimentally. An important conclusion is that the optimal number of emissions decreases when the angular aperture of the array increases. It is also shown that plane wave imaging data are equivalent to the FMC dataset, but requires up to an order of magnitude fewer array firings.