Master's Thesis : Analysis, simulations and prototyping of a 1 kW wide bandgap semiconductor based DC/DC converter

Abstract

Power converters are widely used over the world and are implemented in several electronic applications. This thesis was realised in partnership with CE+T POWER, one of the leaders on the power management market. The principal goal of this project was to revise their current DC/DC converter implemented inside one of their main products, the ``Sierra 10". To do so, the idea of replacing MOSFET by new GaN transistors was investigated.

First, the converter operations were studied and successfully checked on LTspice simulations. The wide bandgap semiconductors technology was summarised and it was shown that there were several advantages of using GaN instead of silicon transistors. GaN transistors are easy to use, allow new capabilities, are reliable, and will be at least as cost-effective as the silicon within few years.

Then, it was shown that the transistors of the converter were controllable by sensing the magnetising current. In practice, the drain current of the primary transistor is almost an image of the magnetising current (without considering the resonance part of the drain current). The drain current would be sensed. It was possible to fix the needed output power to compute the corresponding peak magnetising current values and switching frequency to impose. At first glance, there were several operating points for a given output power. However, it was shown, under assumptions, that an operating point which induces the lowest power losses inside the primary transistor existed. This could be translated into a simple optimisation problem. The mathematical programming results corresponded to the analytical results. The model suggested to decrease the switching frequency around 30 kHz for a peak magnetising current of 81 A. This operating point might not be the most practical one in terms of transformer sizing and cost. Supplementary manufacturing constraints could be added to the model to shift the minimum losses operating point.

The obtained results showed that a possible minimum losses operating point exists and could be tracked under a simple model of losses computation that could be sharpened in function of the technical constraints.