Three Ways for MCU to Solve Common Design Challenges of Traction Inverter for 800V Electric Vehicle

Figure 1: Control loop signal chain

    To achieve excellent real-time control performance, you must optimize the entire signal chain, including hardware and software. The time spent from ADC sampling (input from the motor) to writing PWM (output to control the motor) is the basic measure of real-time control performance. Starting from ADC sampling, the inverter system needs accurate and fast sampling, that is, to achieve high sampling rate, at least 12 bit resolution and low conversion time. Once it can be sampled, it needs to be transmitted to the processor and read by the processor through interconnection, and the optimized bus and memory access architecture reduces latency. In the processor, the kernel needs to use FOC algorithm to calculate the next PWM step according to the phase current, speed and position of the motor.

    In order to reduce computing time to a greater extent, the kernel needs a higher clock rate and must efficiently execute a specific number of instructions. In addition, the kernel needs to execute a series of instruction types, including floating-point, trigonometric, and integer math instructions., The kernel uses the low latency path again to write the updated duty cycle to the PWM generator. Applying dead-time compensation on PWM output will prevent short circuit when switching high side and low side FETs, and apply it at hardware level to reduce software overhead.

    The delay of traction inverter control loop of TI MCU is as low as 2.5s, and the delay of AM2634-Q1 is less than 4s. This level of control loop delay will be targeted at future designs including SiC architecture,.