GaN E-HEMT based high flatness broadband power noise source for dynamic identification of mechatronic systems and batteries

Still today the dynamic identification of mechatronic systems - in particular electrical machines - is often based on stimulating them with deterministic signals, being single sinusoidal tones, single steps or periodical pulse trains of fixed frequency the most common ones. Such methodologies were the most, when not the only, practicable ones when the instrumentation was limited to low frequency generators and analog oscilloscopes, at most. The modern metrological scenario allows to concretize way more sophisticated approaches nowadays, which can bring to Mechatronics more holistic and easily executable identification methods; like those based on flat noise sources traditionally used in Radio Frequency (RF). More in general, the adoption of SiC and GaN fully controllable switches demands that also Power Electronics become eventually permeated with the very different way of thinking peculiar of RF. Within the framework of such cultural and methodological evolution that Power Electronics must embrace, the paper describes the unconventional use of converters based on GaN switches as powerful high-flatness noise sources for broadband dynamic identification, here tailored for modern electrical machines and battery technologies. Indeed, owing to the much shorter raise and fall times, as well as much higher switching frequencies enabled by GaN, it is possible to greatly extend the upper boundaries of the frequency bands where desired spectra can be synthesized by commutating such fully controllable switches according to proper time-based patterns. The authors illustrate the use of a randomly PWM-modulated H-bridge employing GaN E-HEMTs as high-flatness noise source up to at least 200 kHz thanks to 1 MHz PWM carrier. Various selected experimental results from identifications of BLDC Motors and Lithium-Iron-Phosphate (LiFePo4) cells are shown. The details in the figures are fully visible by magnifying the electronic version of the paper, which represents the same approach when Microfilms were widely used.