The ideal switch is not enough

Abstract

There is an ongoing demand for increased power density and efficiency along with lower costs of converter systems and shorter development time for specific applications in the field of power electronics. In order to expedite the technology development Google and IEEE initiated the Google Little Box Challenge (GLBC) including $1 million prize money. Aim of the GLBC was to build the worldwide smallest 2 kVA/400...450 VDC/230 VAC single-phase converter with η > 95% efficiency and an air-cooled case temperature of less than 60 °C by using latest semiconductor technology and innovative topological concepts. Out of 2000+ applications 18 finalists have been selected, whose converter systems exhibited power densities mostly in the range of 120...220 W/in3. With this, a clear performance increase compared to the state of the art (ρ <; 50 W/in3) was achieved, but in the end it represented only a limited performance improvement. In this work, a high power density DC/AC converter system, developed by a team of the ETH Zurich, the Fraunhofer Institute for Reliability and Microintegration (IZM) and the Fraza company and presented at the GLBC finale in Golden, Colorado, will be described and further optimized. Given the converter system, it will be clarified which components and technologies are finally limiting an increase in performance. In a first step, the optimum solution will be identified by means of a ηρ-Pareto front obtained from a multi-objective optimization. The analysis will be based on detailed loss and volume models of the utilized GaN GIT power switches, inductors and capacitors as well as on component stresses, resulting for advanced modulation and control techniques. Thereafter, the models of the power semiconductors will be gradually idealized by means of reducing the switching and conduction losses. The resulting shift of the Pareto front reveals the sensitivity of the system performance with respect to the semiconductor technology and ultimately leads to an `absolute' performance limit imposed by the passive components and the cooling system. It is shown that for fully idealized semiconductors a maximum possible performance increase of 50 % regarding power density or losses is feasible, whereby the switching frequencies are limited to & 1 MHz due to the losses in the magnetic components. Thus, for the realization of highly compact systems, high frequency core materials and winding concepts of the magnetic components, new heat management concepts and 3D-packaging will gain further importance in future along with the ongoing improvement of semiconductor technologies.