There are several tasks which may be used to convolve spectral-line data to a rectangular grid and then Fourier transform and Clean that grid to make an image. The old tasks, which we no longer recommend, are called UVMAP, and APCLN and may be used if you insist. Today, IMAGR should be used for all normal imaging and Cleaning. Read §5.2 for basic information about this task and §5.3 for information about using it to deconvolve your images. It can read either single- or multi-source data, can apply calibration, and writes all channels into an image cube. IMAGR uses much more flexible and correct methods for data weighting (§5.2.3) and uses the superior multi-field method of Cleaning in which source components are subtracted from the visibility data and the images recomputed in every “major cycle.” IMAGR grids each channel in exactly the proper place in the uv plane and can be used to do frequency smoothing (“averaging”) in this most correct of ways. It also offers options to correct your data for various effects of importance over wide fields and bandwidths. It even has experimental variations on Clean (§5.3.4) to deal with extended sources and the Clean bias.
To produce a cube of spectral line images with IMAGR from channels n1 through n2 use:
> IND m ; GETN n C R | to specify the input uv file. |
> CLRONAME C R | to use the default output names and disk. |
> STOKES ’I’ C R | to make total intensity (unpolarized) images. |
> NCHAV 1 C R | to avoid averaging channels. |
> CHINC 1 C R | to do every channel, setting the channel increment. |
> CHANNEL 0 C R | do not restart. |
> CELLSIZ Δx , Δy C R | to set the image cell dimensions in arc seconds. Cells do not have to be square. |
> IMSIZE Nx , Ny C R | to set the image size in pixels (must be powers of two). |
> NITER 0 C R | to do no Cleaning. |
This makes “dirty” images and beams of the specified channels. If Cleaning is needed, set the Clean adverbs NITER, NBOXES, CLBOX, GAIN, BOXFILE, etc. Note that you must image one IF at a time in order to avoid combining multiple frequencies. The task NOIFS might be useful depending on your data.
Because of the upgrade of the VLA in 2010, IMAGR is more careful with image units. Unless you force the Clean beam size with BMAJ, BMIN and BPA, IMAGR will image each channel at its “natural” (frequency-dependent) resolution. It now carefully scales each image plane so that it is actually in units of Jy per the Clean beam listed in the header. The Clean beam for each frequency is recorded in a CG table and used by MCUBE and CONVL among other tasks.
At this point, it is a very good idea to determine the noise in your output images and to compare it to the theoretical noise you expect. If your images are significantly noisier than expected, it is a very good idea to stop processing, to think about what may have gone wrong, and then to check and correct that. The noise may be determined using task IMEAN:
> IND m ; GETN n C R | to specify the input image file. |
> NBOXES 200 C R | to use a significant number of boxes in the plot. |
> PIXRANGE = -x , x C R | to limit the histogram to the range x, where x should be about 5 times the expected noise. |
> GO C R | to run the task, plotting on the TV. |
This will print two rms’s on your message screen one computed using all the data and one done by fitting a histogram to the noise portion of the signal. Especially if the latter fails, the plot will allow you to estimate the true rms, ignoring those pixels significantly above and below the Gaussian noise part of the histogram. Using this plot, it may be better to include all of the data (BLC 0 ; TRC 0 C R) rather than to limit the number of pixels contributing to the histogram. The AIPS verb IMSTAT does the all-pixel rms computation without the useful plot, while the verb TVSTAT allows you to mark one or more non-rectangular regions on an image on the TV over which the rms is computed. Task RSPEC will plot robust rmses computed on a per plane basis.