<title>Effect of adaptive interference supression on radio astronomical image formation</title>

Abstract

ABSTRACT The increasing use of the electromagnetic spectrum and the need for more sensitive radio telescopes spures wide interest inadaptive RFI suppression techniques, such as spatial filtering. We study the effect of spatial filtering techniques on radioastronomical image formation. Current deconvolution procedures such as CLEAN are shown to be unsuitable to spatiallyfiltered data, and the necessary corrections are derived. To that end, we reformulate the imaging (deconvolution/calibration)process as a sequential estimation of the locations of astronomical sources. This leads to an extended CLEAN algorithm andgives estimates of the expected image quality and the amount of interference suppression that can be achieved. Some of theeffects are shown in simulated images.Keywords: Spatial filtering, synthesis aperture, image formation, CLEAN, radio frequency interference, Square KilometerArray, parametric methods. 1. INTRODUCTION The future of radio astronomical discoveries depends on achieving better resolution and sensitivity while maintaining immunityto terrestrial interference which is rapidly growing. The last two demands are obviously contradicting as improved sensitivityimplies receiving more interfering signals. One possible track is to switch to massive phased array technology. If instead ofthe huge dishes which became the trademark of radio astronomy, we use phased array radio-telescopes comprised of tens ofthousands of small elements, then we gain both in terms of resolution and sensitivity while increasing the flexibility to mitigateinterference. The international effort in this direction is coordinated under the framework of the Square Kilometer Array project(SKA). In this paper we try to analyze the effect of on-line interference rejection on the image formation process in such aninstrument as well as on existing radio telescopes such as WSRT.All existing radio interferometers, as well as future ones have elements sparsely located, due to both physical and econom-ical limitations. The sparse coverage of the visibilty domain, imposes a need to reduce the sidelobe response in the imagethrough a process of deconvolution. Two principles dominate the astronomical imaging deconvolution. The first method wasproposed by Hogbom' and is known as CLEAN. The CLEAN method is basically a sequential Least-Squares (LS) fittingprocedure in which the brightest source location and power are estimated. The response of this source is removed from theimage and then the process continues to find the next brightest source, until the residual image is noise-like. Although it hasbeen partially analyzed (e.g., by Schwartz2) full analysis of the method is still lacking due to its iterative nature. A second rootproposed by Jaynes3 is maximum entropy deconvolution (MEM). The above mentioned algorithms assume perfect knowledgeof the instrumental response (point spread function). Due to various internal and external effects this assumption holds onlyapproximately. One way to overcome this problem is the use of calibrating sources. An unresolved source with known param-eters is measured, and by relating the model errors to the array elements a set of calibration equations is solved. A much moreappealing solution is to try to improve the fitting between the data and the sky model by adjusting the calibration parameters.A good overview of the various techniques is given by Pearson and Redhead.4A major problem facing radio astronomy is the accelerated use of the electro-magnetic spectrum. Even in bands which arereserved to radio astronomical observation one can find interference, e.g. ,