Apologies if the following post might be appear obvious, however I felt it best to share with the community just in case!
I have been reducing some 21cm spectral line data in the CABB 64MHz zoom-band.
For our work we would like to produce a spectrum at the centre of the target source in order to look for absorption lines. We would also like to keep the continuum spectrum if possible since the analysis requires a good estimate of the continuum flux as a function of frequency.
However we noticed that (after full data reduction) the extracted spectrum would sometimes have a regular ripple, and in some cases was a good fraction of the continuum flux. This is a problem if we are interested in detecting weak 21cm lines which could potentially be broad and confused with the ripple feature.
This ripple was found to be real and due to the presence of other significantly bright sources in the field. The "period" of the spectral ripple was seen to be a function of telescope baseline distance and would correctly predict the distances of the other sources from the centre of the field. The amplitude of the spectral ripple would change with time as the synthesised beam picked up the secondary sources when orientated towards them.
Essentially the side lobe contribution from other bright sources in the field contribute a frequency dependent flux to the extracted target source spectrum.
This will primarily be an issue with broad band data since the spectral resolution and width is good enough to demonstrate these ripples. Perhaps coarser channels or narrower bands would not so easily show this.
Bjorn Emonts and I came up with a relatively simple solution to this issue (essentially subtracts the continuum flux at the
positions of sources other than the target source):
1) Generate an multi-frequency synthesis continuum image using INVERT
2) Use MFCLEAN to construct a component model image of the strongest sources (best to go out more than 2xPrimary Beam FWHM).
3) Use IMMASK to blank out the central pixels in the component model containing our target source.
4) Use IMBLR to replace these blanked pixels with zero values.
5) Use UVMODEL to subtract the component model image from the original UV data set
6) Construct a Data Cube from this new UV data set.
7) Extract the spectrum at the data cube centre.
The following spectra show a worst case for some data, before and after the spectral ripple is removed. In this case the data contained a few other significant sources in the field-of-view.
- with_ripple.png (143.05 KiB) Viewed 46791 times
- removed_ripple.png (128.89 KiB) Viewed 46791 times
Cheers,
James