Large Multibeam Array Antennas with Reduced Number of Active Chains

Direct radiating arrays (\´DRA\´) have been proven to be an interesting solution for reconfigurable multibeam transmit antennas, as spreading naturally the RF power to be radiated over the whole aperture, and avoiding cold redundancies thanks to graceful degradation. DRA are in general designed following two main constraints : (1) Antenna diameter is determined by directivity and isolation specifications (2) Grid lattice is constrained by grating lobe rejection outside a given domain (typically outside the Earth, for geostationary satellite antennas considered here) As high directivity beams are mostly required, adding these two constraints leads to a prohibitive number of antenna elements (so of active chains). In order to reduce the number of active chains without affecting antenna pattern characteristics, two solutions are studied here : (1) Array thinning, relies on suppressing part of the radiating elements in the regular grid lattice, (2) Non regular aperture sampling consists in dividing the radiating aperture into non-regular subarrays. Industrial constraints lead to gather small identical elements in rectangular groups with various size. The basic elements are small enough to avoid any grating lobe on the Earth disk; as the 2^nd-step aperture meshing (by non-regular groups) is non- periodic, no other grating lobes appear on the Earth. Various kinds of mathematical algorithms have been compared for both these arraying methods. They are classified in 2 main categories: a) part I presents so-called "global optimisation algorithms based on random searching": \´genetic\´ algorithm and \´simulated annealing\´ are assessed to perform array thinning and subarray division. As a result, such algorithms are well- suited to array thinning, but not to gathering elements in non-regular groups, providing good performances for all numerous beams; so we went to a new category of much different methods b) part II presents a new analytical method, built especially - or this problem by UPS/MIP. It associates: (1) an optimised choice of the cost function, able to warrant convergence of a gradient-type method (2) combining solutions found for each beam in a single power distribution on the aperture is performed using the "singular value decomposition" (SVD) method (3) then the obtained distribution is sampled into amplitude values that can be provided by gathering elements by 1, 2, 3, or 4. And the best rectangles arrangement is found in an iterative process using "topologic gradient" method. A clever association of these various steps leads to a non-regular subarrays distribution saving nearly 50% of the initial elements number, while complying for all beams with typical requirements on gain and isolation, and using equal-power feeding, so better efficiency and lower cost for a single amplifiers class.