Next: Typical deconvolution session Up: The family of CLEAN Previous: Implementation Contents Index
CLEAN (except for MRC) has three outputs: the CLEAN image itself, the RESIDUAL image, and the list of point sources (the Clean Components) that reproduce the observed visibilities, the CCT table.
An aesthetically better results, with better noise properties, can be obtained after CLEAN by removing the Clean Components from the measured visibilities, and re-imaging this to produce the RESIDUAL image. Aliasing at map edges is then minimized, as it only concerns noise if the deconvolution was reasonable. As mentioned previously, this is implicitely done when using MX instead of CLEAN. It can also be done after CLEAN using command UV_RESTORE.
This step can also be used to properly scale the residuals when the synthesized beam is not well fit by a Gaussian, or when all channels do not share the same beam (see Section 4.3). This is done through the use of the so-called JvM factor. This factor, introduced in Jorsater & van Moorsel (1995), estimates the ratio of clean beam area to dirty beam area, and is stored in variable BEAM_JVM and (on a per beam basis) BEAM_VALUES[4] which are computed by FIT /JVM_FACTOR. Residuals are multiplied by this factor (on a channel per channel basis if needed), allowing to first order to scale the residuals to the same unit as the Cleaned data, i.e. Jy/beam area.
In general, this number is fairly close to 1. However, when several observing configurations have been merged together, the synthesized beam often exhibits a central peak over a larger plateau, because of the higher weights of short baselines. Robust weighting limits the effect, but is not able to suppress it totally. In such cases, the JvM factor can be of order 0.5 to 0.6.
UV_RESTORE uses the JvM factor if FIT /JVM_FACTOR has been used before. It also uses the mean Clean beam geometry defined by BEAM_FITTED.