Predictive efficiency optimization for DC-DC converters with highly dynamic digital loads

This paper presents a novel technique and system for increasing the efficiency of dc-dc converters that supply dynamic electronic loads, such as modern audio and video equipment and other devices whose power consumption largely depends on the digital data they process. The optimization does not require a current-measurement circuit and is well-suited to portable applications. It is based on a real-time prediction of the dc-dc converter output current from easily accessible digital data streams present in the targeted loads. The result of the prediction is used for dynamic adjustment of the power-stage transistor size and/or for switching into pulse-frequency-mode of output voltage regulation, in order to maximize the instantaneous converter efficiency on-the-fly. The use of a segmented power-stage allows the effective power-transistor size to be changed on-the-fly, and the tradeoff between the gate-drive and rms conduction losses is continuously optimized over the full range of operation. The effectiveness of the optimization is demonstrated on an experimental system, including a I-W digitally controlled 4-MHz, 3.6 V-1.8 V buck converter with an integrated segmented power-stage and a digital high-fidelity class-D audio amplifier acting as the digital load. The results show a good agreement between the digitally predicted and actual dc-dc converter load current, as well as a reduction in total energy consumption of up to 38%.