We have recently proposed a scheme enabling the collection of arbitrary 2-D NMR data sets within a single scan. This so-called ultrafast methodology operates by replacing the traditional t 1 temporal encoding with a spatial encoding of the indirect-domain spin interactions. Following a mixing period, the spatially encoded I(ω1) spectrum can be repeatedly read out using a train of oscillatory gradients, leading to a single, continuous FID containing an interferogram with the mixed-domain S(ω1, t 2) data. Rearrangement of the resulting data points followed by Fourier processing along the direct-domain axis can thereby lead to arbitrary 2-D I(ω1, ω2) spectra–within a single transient. This article describes the physical principles of this single-scan 2-D NMR procedure, and gives practical guidelines for its implementation in commercial instruments. Attention is paid on how to translate the spectral variables defining traditional 2-D NMR experiments into the timing and gradient parameters involved in single-scan 2-D acquisitions; on how to program the spatial encoding RF manipulations needed for the experiment's execution; and on the special processing aspects involved in this kind of acquisitions. The optimization of the various data collection and processing stages is discussed, and their implementations are illustrated with simple guiding examples and with state-of-the-art experimental results on proteins and nucleic acids.