M.A. Holdaway [1], S.M. Foster [1]
October 19, 1994
Keywords: OTF mosaicing, on the fly mosaicing, Fourier plane coverage, UV coverage, observing modes, slew rate
In mosaic observations consisting of many (250-10,000) interometric pointings, non-adjacent pointings may have significantly different (u,v) coverage hence different synthesized and restoring beams. The advantages to having similar synthesized beams for each pointing include: the image is easier to interpret, deconvolution errors will have similar character across the image, a single clean beam can be used for the entire image, residuals are in the same units across the mosaic, and the inexpensive linear mosaic algorithm can be used. In the same units across the mosaic, and the inexpensive linear mosaic algorithm can be used. In order to make the beams for all pointings similar to the 1 % level, each pointing in the mosaic must be observed every 12 minutes. For mosaics in which beam shape differences of 5% can be tolerated, each snapshot must be observed every 45- 60 minutes. Mosaics smaller than 250 pointings can achieve very similar (u, v) coverage and beams similar to 1 % with the traditional stop- and-go mosaic observing strategy with on-source integrations of about 3 seconds per snapshot with 1 second setup time. Mosaics smaller than 900 pointings can achieve 5% beam differences with stop- and-go mosaicing. Larger mosaics may require continuous slewing across the source, reading out the data on time scales which are not short compared to the beam crossing time. We call this technique on-the-fly (OTF) mosaicing. The Ekers and Rots short spcing time scheme places restictions on the maximum averaging time in terms of the beam crossing time. The short integration times required by OTF mosaicing place extreme demands on the correlator and on-line system. The mosaicing algorithms currently used will work with OTF mosaicing, but the assumed rotationally-symmetric primary beam must be replaced with the effective primary beam, which is the true primary beam convolved with a 1-D boxcar. The very high data rates may require rethinking the typical data processing pathway.
[1] NRAO/Socorro
Last modified: 09 December, 1999
kweather@nrao.edu