Power levels in Point of Load (PoL) converters application have seen a rising due to the AI-driven datacenter revolution. The transient requirements have also become more and more challenging, with currents slews approaching the multiple Ampere/ns. This paper proposes a novel control methodology which aim to improve the transient response of such systems, while maintaining a relatively slow switching frequency in the power stages for improved efficiency. The proposed controller is composed of a standard ‘steady-state’ control loop paired with a specialized transient controller, tailored to the system parameters by leveraging machine learning techniques in an analog fashion for minimum latency. This approach allows for better system efficiency, high input voltage capabilities, easier cooling and better transient performance.
