In accordance with the SA roadmap the overlay requirements for 'state of the art' DUV stepper systems had to be pushed down to 60 nm, which represents one third of the resolution. As a consequence, the mask contribution to this total budget can be estimated to fall below 15 nm on wafer level. Hence, there is a tremendous need for both enhanced mask-writing accuracy and appropriate overlay qualification tools. To qualify the stitching accuracy of e-beam mask-writers e.g., we apply a special mask design, composed by a defined ensemble of dense lines. The overlay information about the pattern generator frozen within such test-masks, will be obtained by measuring the phase alteration effect each 1st order diffracted beam suffers from translation of the grating structure. The overlay reproducibility 3(sigma) is smaller than 5 nm on wafer level. The absolute accuracy of this technique reaches the same order of magnitude. This will be checked by cross-correlating our results with corresponding scanning-based overlay measurements. With this method we take advantage of the fact, that one CCD-snapshot instantaneously delivers the full-field overlay information. Without scanning, i.e. without any stage oscillation or temporal drift effects, we measure the global overlay field. Line-edge roughness effects are averaged out by integration across the pixel area. The paper will present some of our test-results. The demand for 15 nm overlay accuracy could be really critical. For some different mask designs as well as mask-writing systems our investigations indicate the specific necessity for future improvements with respect to stages, temperature drift sensitivity, and possibly local heating effects.