Abstract Improving the compactness and efficiency of switching converters is a central issue in power electronics ; even more so in satellites where every gram and every watt counts. Each of the many radio-frequency emitters and receivers onboard telecommunications satellites need to be powered by various voltages, converted in an isolated way from the main power distribution bus. Due to the strong thermal, reliability and radiation hardness constraints applying to electronic components in space applications, available degrees of freedom for improvement of power supplies are limited - at least with current qualified semiconductor technologies (which are both expensive and far behind state-of-the-art performance). The recent commercialization of gallium nitride (GaN) normally-off power transistors, having superior electrical characteristics compared to the best silicon power MOSFET, is promising on that regard. Indeed, their intrinsic radiation hardness seems to allow their use in space-grade converters. The aim of this work is the evaluation of how this technology can help improve the design of isolated DC/DC power supplies for typical hardware units of telecommunications satellite payloads. Operation at higher switching frequencies with these better performing components should, in principle, reduce converters’ footprint while keeping the same (or better) efficiency level and still obeying each application’s specific requirements. The accuracy of this hypothesis as well as the most adequate implementation architecture have been explored for the low power supply of a RF receiver, including realization and comparison of several demonstration boards. In order to approach higher power converters, a theoretical and experiment study of switching losses in GaN transistor bridge legs has been performed. A performance computation software has been developed in Python and used to identify the global optimum of the design of a Dual Active Bridge converter for a power RF amplifier (250W DC). A prototype board has been built and demonstrated the interest of both the topology and GaN devices in this application, while clearly showing that high-frequency losses in magnetic components dominate total converter loss. This last issue happens to be the main limitation of the approach - precious to the engineer - of optimum design by computation : currently existing models for power loss estimation in magnetic elements are not satisfactory to predict performances of this type of converter.
LINK ORIGINAL
https://core.ac.uk/download/pdf/78385176.pdf
