38 November/December 2023 | E-Mobility Engineering inputs on a cycle-by-cycle basis, making adjustments in real time to small, forced-resonant transistors, enabling perfect soft-switching in harsh changing environments. Variations in system temperature, device degradation, changing input voltages and abrupt current swings are all accounted for and optimised within the ML algorithm. The algorithm controls the current that charges the capacitors for around 100 ns at the zero-voltage, zero-current point. This adjusts the timing of the switching with an accuracy of 2 ns every cycle. Other algorithms can do this for certain points in power, but it’s the infinite variety of changing conditions that creates the problem, which is why ML is used for the prediction. This takes in data that’s incomplete and in a noisy environment, and calculates where the switching will be. It monitors the device’s temperature and switching speeds, and measures how they react and degrade over time to allow the algorithm to compensate for these changes. There is some pattern matching, but the algorithm calculates the individual responses of each transistor on a cycle-by-cycle basis. The evaluation data shows the switching technology hits a peak of 99.57%, with 98.5% efficiency at 5% load in the 200 kW inverter design with 100 kHz switching. The efficiency at low loads and the low distortion for the high-frequency sine wave to the motor are key to improving the WLTP range performance. Gate drivers IGBT and SIC MOSFETs have similar requirements for motor control. IGBTs need -8 V in the Off state and +15 V in the On state, while the SiC MOSFET voltage varies slightly, from +18 or +20 V in the On state to -5 V in the Off state. These are close enough for a single gate driver to drive either type of transistor if designed appropriately. GaN transistors on the other hand need drivers that range from -2 to -5 or -6 V, and so need different gate driver designs. The gate driver IC has to turn on the SiC FETs as efficiently as possible, while minimising switching and conduction losses that include the energy used during the ‘turn on’ and ‘turn off’ process, which has to be as fast and efficient as possible. The ability to control and vary the gate-drive current strength reduces switching losses, but at the expense of increasing transient overshoot at the switch node during switching. Varying the gate-drive current controls the slew rate of the SiC FET. Real-time variability of the gate-drive current enables transient overshoot management as well as design optimisation throughout the highvoltage battery energy cycle. A fully charged battery with a state of charge of 80-100% should use low gate-drive strength to minimise the SiC voltage overshoot. As the battery charge drops from 80% to 20%, using high gate-drive strength reduces switching losses and increases traction inverter efficiency. These scenarios are possible during 75% of the charging cycle, so the efficiency gains can be quite significant. A 20 A isolated real-time variable gate driver adds a highly configurable adjustable slew-rate gate driver targeted to drive high-power SiC MOSFETs and IGBTs, and squeeze as much current out of the system as possible. Power transistor protections such as shunt resistor-based overcurrent, temperature sensors and desaturation (DESAT) detection include selectable soft turn-off or two-level soft turn-off during these faults. To further reduce the size of the motor control system, an active Miller clamp and an active gate pull-down can be integrated into the driver. An integrated 10-bit ADC enables monitoring of up to two analogue inputs, A traction inverter design showing the motor control stage (Courtesy of Texas Instruments)
RkJQdWJsaXNoZXIy MjI2Mzk4