The simplest conceptual approach is to make the spectral-line cube, and then subtract from all planes (channels) in that cube an image made from the channels containing continuum only. You can see which channels have line and which have continuum by making a spectrum either with ISPEC\ on the cube, or POSSM on the visibilities (but you only get the spectrum at one location on the sky; see discussion in § 6.2).
This method has two main problems.
If neither of these problems affects your data, you could proceed by first making the spectral-line cube containing all channels (both continuum and line) as described in §15.6. Next, compute the averaged continuum image with SQASH.
SQASH | |
inname,inclass | Input cube |
inseq,indisk | |
outname,outclass | Output averaged |
outseq,outdisk | continuum image |
blc=0,0,N | Specify channels N |
trc=0,0,M | to M to average |
bdrop=3 | Squash axis 3 |
dparm=1,1 | Average and ignore blanks |
It may be that you have continuum at several places in the spectrum. In that case, run SQASH successively on each region, and then average the SQASHed images with COMB. Finally, subtract the continuum image from the cube with COMB.
COMB | |
inname,inclass | Input cube |
inseq,indisk | |
in2name,in2class | Input continuum |
in2seq,in2disk | image |
outname,outclass | Output subtracted |
outseq,outdisk | cube |
blc=0 | Full region |
trc=0 | |
opcode='sum' | Summation operation |
aparm=1,-1 | Negate continuum |
Note that this method requires you to have two copies of your cube; this can be a severe problem if the cubes are large and disk space is tight. You should examine the image for obvious signs of differential beam effects. One way to check this empirically is to make beams from channels as widely separated as possible, and then subtract them with COMB. Examine the output image and find the peak value. See if this is large enough to cause you difficulty for the science you are trying to extract.
An alternative way to achieve the same result, but which requires less disk space is as follows. Instead of making a dirty cube from all the channels, first use AVSPC to average the visibility channels containing continuum only signal into a separate visibility data base (see § 6). Then make the dirty continuum image from this. You then make the dirty spectral-line cube (see § 15.6) from only those channels containing the spectral-line. Then you proceed with COMB as described above.
In the special case that your continuum channels are symmetrically placed about the spectral-line, and the line is sufficiently narrow, this method will cope with a non-zero spectral index, to first order. But this condition is not always the case.