Observations and processing

In a single-field observation, an interferometer tracks a particular direction of the sky, named the phase center. The portion of the sky which can be image around this direction is directly linked to the size of the primary beam. The easiest way to image field-of-view larger than the primary beam size is to track one direction of the sky after another until the desired field-of-view is filled with small images made around many different tracking directions. This observing mode is called mosaicing and the tracked observations which constitute the mosaic are called fields.

There are many constraints to optimize mosaicing.

Nyquist sampling of the mosaic field-of-view and mosaic pattern

The mosaic field-of-view must at least be Nyquist-sampled to obtain a reliable image. Each observed field can produce a reliable image of the same shape than the primary beam, i.e. a circular Gaussian (This assumes that the short-spacing problem has been solved). Nyquist sampling thus implies that the mosaic fields follow an hexagonal compact pattern as this ensures a distance between all neighboring fields of half the primary beam size. When the total observing time is fixed, Nyquist sampling is the best compromise between sensitivity and total field-of-view. Indeed, the distance between neighboring fields could be less (in which case the mosaic would be oversampled) than half the primary beam size. In this case, the sensitivity on each pixel of the final image would increase with the share of the time spent to observe this direction.

Uniform imaging properties and quick loop around the fields

Getting uniform imaging properties is a desirable feature in the final result. This implies that a $uv$ coverage and a noise level as uniform as possible among the different fields. Quickly looping around the different fields is the easiest way to reach this goal. However, dead time to travel from one field to another must almost be minimized. At PdBI, the compromise is to pause at least 1 minute on each field and to try to loop over all the fields between two calibrations every 20 minutes. Hence, mosaic done in a single observing run is made of at most 20 fields. Larger mosaic must be observed by group of fields in different observing runs.

After calibration and production of an image per field, two different processing flow are possible.

1. The field are deconvolved independently and the clean image are combined to obtain the final image.
2. The dirty images are combined together and a global deconvolution happen on the combined dirty image.

While solution 1 is simpler to implement, solution 2 gives a better deconvolution because 1) information about sources at the edges of and individual field is furnished by the other fields and 2) the signal coming from the different field add up to get a better signal-to-noise ratio. The next two sections describes the particularities of mosaic imaging and deconvolution.