Here's an edge at the SW node when it transitions from the low side turning off to high side turning on. Blue trace is the PWM HI input of the GaNFET. Green trace is the SW node at the inductor.
The STM32 HRTIM hardware is configured to about 5ns of deadtime on the complementary outputs, which has some propagation delay. The roughly 10 ns negative interval occurs while both GaNs are off. The inductor current must continue flowing, so it pulls the SW node negative through third quadrant conduction until the high-side FET turns on. GaNs do not have a body diode like a MOSFET, and reverse voltage is higher.. in this case nearly -5V.
On the spectrum analyzer with a near field probe sitting over top of the hot loop, you can see the energy of the ringing at around 180MHz.
Going to see if I can pull off a CNC machined digital SMPS using an STM32G474 and LMG5200 GaN half bridge as a power distribution board for my hobby robot.
The GaN stage will be driven by the G474's HRTIM, with ~184 ps (~5.44 GHz equivalent) timing resolution.
Inductor current is measured with a 4 terminal kelvin shunt (much more reliable on CNC boards, as kelvin is internal to the resistor) and high side INA sense amp, feeding both an ADC and an on chip comparator.
Planning to run peak current-mode control.. an internal DAC set by the outer voltage loop defines the cycle by cycle current threshold on the comparator, which asynchronously terminates the PWM pulse through HRTIM hardware.
The outer voltage loop will use the G474's FMAC accelerator for 3P3Z IIR compensation from a voltage divider buffered using an internal opamp follower routed to another ADC channel. Also connected to a comparator for HRTIM fault events for fast hardware based over voltage lockout.
Includes a INA254 bidirectional sense amp with integrated 400uOhm shunt for monitoring battery current.
Added CAN bus to broadcast voltage, current and status telemetry across the platform.