Maximum sensitivity is obtained with natural weighting, and no taper. Because the phase noise may increase with baseline length, however better image quality may be obtained with a smaller taper, at the expense of spatial resolution. Maximum resolution is obtained using uniform weighting, and no taper. The uniform cell size can control to some extent the beam shape. Bigger cells will tend to increase noise, however. Uniform weighting enhances the contribution of (and the noise from) the long baselines. Task UV_MAP allows a compromise between uniform and natural weighting, by arguments UV_CELL. Best compromise can be determined using task UV_STAT.
In using CLEAN, avoid the following traps: cleaning to deep, undersampling the beam, cleaning more than the inner quarter of a map, or restoring with too small a beam.
Many astronomers believe that MX produces better results than UV_MAP and CLEAN. This is not true: UV_MAP and CLEAN will produce results very similar to MX provided they are used on images twice bigger. Despite this greater storage use, using UV_MAP + CLEAN is usually much faster than MX, and provides the advantage of checking the aspect of the dirty image.
In practice, never use MX at the beginning. MX is only worth (for Plateau de Bure data) for ``final'' deconvolution, if at all.