Microchannel-based Calorimeter for Rapid and Accurate Loss Measurements on High-efficiency Power Converters

The difficulty in accurately measuring losses in high-frequency power converters hinders accurate breakdown of losses which is essential for device and circuit optimization. Electrical methods are often not adequate and calorimetric approaches found in the literature require bulky systems, with long measurement times – between several hours to an entire day. Here we demonstrate a compact microchannel-based flow-calorimeter to measure losses down to 20 mW and up to 10 W with an error below 5% over a wide operating range. The close proximity of the microchannel cold-plate, as well as its small heat capacity, enables rapid measurements within 150 seconds, as well as fast calibration. We show the application of this system in measuring losses in a GaN-based class-E resonant converter operating at 10 MHz, by performing fast measurements over a wide range of operating conditions, and providing a precise breakdown of losses between active and passive components. These results highlight the potential for the accurate characterization of next generation power converters, especially at high switching frequencies.

I. Introduction

Wide-band-gap (WBG) semiconductors, such as GaN and SiC, enable highly efficient energy conversion at frequencies reaching the MHz-range. Reduced losses enable more compact cooling systems, whereas higher switching frequencies allow a size-reduction in passive components. These properties make WBG devices prime candidates for a future generation of compact power converters, supporting the electrification of society. However, the low losses and high frequencies of these devices poses a challenge for system-level characterization [1]. The use of electrical measurements to determine the losses in circuit components operating at high frequencies are prone to errors, due to problems including bandwidth limitations of the measurement equipment and signal offsets. Subtracting two similar values even with a relatively small error can result in a measured loss with a large uncertainty [2]. As a result, this causes spurious efficiency measurements, and can cause observed efficiencies to exceed 100% [3].