P10.28 Sebastiaan van der Tol (Astron)Theme: Data processing pipelines
Estimating continuous direction-dependent gain screens from radio interferometric visibilities and aTo obtain a high dynamic range image with a wide field of view from radio interferometeric data.
it is crucial to take direction dependent (DD) gain variations into account.
Correcting for DD variations is usually done by either partitioning the image into facets or
including the DD variations in the convolution kernel while (de)gridding, i.e. A-projection.
The projection method has as advantage over faceting that it effectively applies a continuous (sinc interpolated) correction. The computation of the kernels for A-projection used to be prohibitively expensive. Image domain gridding solves this problem by circumventing the need to compute kernels.
Up until now, the only feasible methods for finding gain screens for large sky models rely on facets or patches. While it is possible to fit a continuous screen to a discrete set of solutions, such a two-step process does not take full advantage of the ability of applying continuous screens with A-projection.
Here we present a method, based on image domain gridding, to directly fit continuous gain screens. The method is efficient even for very large sky models. Although the method can be used to fit full Jones matrices, we show here results for the simpler case of a scalar complex gain described by second order 2D polynomials for phase and amplitude. We demonstrate the feasibility of this approach on a model image of 20k x 20k pixels with 10^6 non-zero pixels (clean components)