Computation

Task invert is a fairly conventional imaging program, which produces a dirty image from a visibility dataset. It normally does this using a grid-and-FFT approach, although there are switches to use a direct Fourier transform and a median algorithm. Task invert does not require the data to be sorted in any way. Normally any calibration tables are applied by invert on-the-fly (although this can be turned off with the nocal, nopol and nopass options). Both continuum or spectral line observations are handled.

We describe the inputs to invert. For MFS imaging, note options=mfs (and options=sdb) options.

Typical inputs to invert for a continuum experiment are given below.

INVERT
vis=vela.uv.1,vela.uv.2 Name of input visibility datasets
map=vela.imap,vela.vmap Name of the output images - one per Stokes
beam=vela.beam Name of the output beam
cell=1 Pixel size is 1 arcsec square
imsize=256,512 Image size is 256 by 512 pixels
stokes=i,v Image Stokes I and V
sup Leave unset to get uniform weighting,
or  
sup=0 Set sup to zero for natural weighting
fwhm Specify desired resolution - unset gives the max
  resolution
options=mfs Use the MFS option for multi-frequency synthesis, or
options=mfs,sdb use the sbd option as well if using mfclean.
line Controls channel selection and averaging. Default
  is all channels when options=mfs.

For a spectral line observation, typical inputs would be

INVERT
vis=vela.uv.1,vela.uv.2 Name of input visibility datasets
map=vela.imap Name of the output image
beam=vela.beam Name of the output beam
cell=1 Pixel size is 1 arcsec square
imsize=256,512 Image size is 256 by 512 pixels
stokes=i Image Stokes I only
sup Leave unset to get uniform weighting,
or  
sup=0 Set sup to zero for natural weighting
fwhm Specify desired resolution - unset gives the max
  resolution
line=velocity,10,1.5,1.0,3.0 Map channels to velocities. Image 10
  velocities centred at 1.5, 4.5, 7.5 etc km/s

In its computation invert determines the theoretical rms noise. This noise is calculated assuming that the only source of error is the system temperature of the front-end receiver. No account is made of calibration errors, sidelobes or any other `instrumental' effects. The calculation correctly accounts for the weighting scheme used in the imaging process. This theoretical noise is the level you can expect in a detection experiment (assuming no interference or confusion), and it is the best one can hope for in high dynamic range work (usually instrumental effects will limit you before the noise in these sorts of experiments).

The noise calculation of invert (and all other Miriad tasks that compute the variance of a correlation) is based on values of system temperature, system gain, integration time and bandwidth stored in a dataset. Unfortunately data loaded into Miriad using fits will have only nominal system temperatures and system gains, and an educated guess is made for the integration time. Data loaded using Miriad atlod, the system temperatures are those measured on-line, and the integration time will be correct. See Chapter 8 for a discussion of setting you dataset up so that noise estimates are correct. The system gain, however, is still a nominal figure. If system temperature, system gain or integration time are incorrect by some factor, then the theoretical rms noise will also be wrong by the same factor.

In continuum mode (i.e. the 33 channel/128 MHz mode), when all 33 channels are retained after atlod (i.e. options=birdie was not used) there is another effect which will cause invert's noise estimate to be a factor of $\sqrt2$ too pessimistic (i.e. the actual noise is a factor of $\sqrt2$ lower than the printed value). With the ATCA correlator in this mode mode, the channel bandwidth is twice as large as the separation between channels (i.e. the channels are oversampled). Unfortunately invert assumes that the bandwidth is the same as the separation. This will only affect you if you are imaging individual correlator channels (i.e. no frequency averaging) and if you keep all channels. It will not affect ``channel 0'' or multi-frequency synthesis imaging.

Miriad manager
2016-06-21