#41 Problem with -joinchannels -channelsout and flagged data

[André Offringa]

When some 'output' channels are completely flagged, the created MFS images are not valid.

Reported by R van Weeren:

We are running against a problem using the wideband mode of wsclean. For a dataset with 128 channels we run with "-joinchannels -channelsout 32". (It's a wideband, 120-180 MHz, LOFAR dataset). Four neighboring channels of this dataset are completely flagged (precisely comprising a 4 channel block that is treated "separately") due to RFI. This results in weird MFS output images and the clean components are not restored. (also beams are set to 0.0 in the individual images, and also the sources are not restored there as well).

I'm quite convinced the problem is related to the 4 flagged channels (where there is no data to image). Would it be possible to fix this issue?
(the deconvolution itself seems to go completely fine (for the unflagged channel groups), until it hits the point were it finishes the deconvolution and tries to make the MFS/final images, the code is probably getting confused with the image from the channel group with no data (and thus no beam or clean components))

I can provide a small test dataset if needed (a 2 channel dataset, with one completely flagged channel should be enough to reveal the problem using "-joinchannels -channelsout 2". Let me know.

[André Offringa]

Recreated a test case in which the first output channel (out of two) was completely flagged; all went fine, so this doesn't always happen. It might have to do with #42 : if the last output channel is completely flagged, wsclean would not have proper beam information after fitting.

[André Offringa]

Bug was reportedly solved, except that the MFS image won't get beam information, but thas is reported in bug #42.