National Radio
Astronomy Observatory

5.3.4.6 The threshold revisited

For mosaics, the specification of the threshold is not straightforward, as it is in the single field case. This is because the different fields can be observed to different depths, and get different weights in the mosaic. We now provide internal rescaling (based on scaletype) so clean does its component search on a properly weighted and scaled version of the sky.

For ftmachine=’ft’, the minor cycles of the deconvolution are performed on an image that has been weighted to have constant noise, as in ’SAULT’ weighting (see § 5.3.4.5). This is equivalent to making a dirty mosaic by coadding dirty images made from the individual pointings with a sum of the mosaic contributions to a given pixel weighted by so as to give constant noise across the image. This means that the flux scale can vary across the mosaic depending on the effective noise (higher weighted regions have lower noise, and thus will have higher “fluxes” in the ’SAULT’ map). Effectively, the flux scale that threshold applies to is that at the center of the highest-weighted mosaic field, with higher-noise regions down-scaled accordingly. Compared to the true sky, this image has a factor of the PB, plus a scaling map (returned in the .flux image). You will preferentially find components in the low-noise regions near mosaic centers.

When ftmachine=’mosaic’ and scaletype=’SAULT’, the deconvolution is also performed on a “constant noise image”, as detailed above for ’ft’.

ALERT: The intrinsic image made using ftmachine=’mosaic’ is equivalent to a dirty mosaic that is formed by coadding dirty images made from the individual fields after apodizing each by the PB function. Thus compared to the true sky, this has a factor of the PB² in it. You would thus preferentially find components in the centers of the mosaic fields (even more so than in the ’ft’ mosaics). We now rescale this image internally at major-cycle (and interactive) boundaries based on scaletype, and do not have a way to clean on the raw unscaled dirty image (as was done in previous released versions).

More information about CASA may be found at the CASA web page

Copyright © 2010 Associated Universities Inc., Washington, D.C.

This code is available under the terms of the GNU General Public Lincense