Model-based dual-mode controller for low-power buck converters

This paper describes the design of a digital, modelbased dual-mode controller for a low-power, synchronous buck converter for battery-powered applications. In feedforward mode, the controller drives to new operating points as fast as possible without exceeding the maximum allowed battery current. The desired output voltage can be determined digitally and can be changed during operation time for dynamic voltage scaling (DVS) capability. In feedback mode, a significant reduction of the analog to digital converter (ADC) sampling rate is achieved by using a model-based approach. The model also estimates the load current, thus no current sensing is required. Furthermore, the model preserves the output voltage from limit cycles, although a low-resolution pulse width modulation (PWM) is used. A gain-scheduling scheme of the PI controller optimizes the control performance of the nonlinear behavior of the converter and keeps the adjustment energy after a load step in a range the battery can accomplish. Known load changes of the application are additionally taken into account to further improve voltage stability. The control algorithm has been implemented in a field programmable gate array (FPGA) and is tested together with a buck converter. ADC sampling rate is chosen 10x smaller than switching frequency of the converter and a battery with a maximum current drain of 40mA is used. The transient load response shows a maximum voltage drop of 10mV for known and 80mV for unknown load steps with a recovery time of 70μs. Efficiency of low-power switching regulators is improved over conventional digital designs by using slow ADCs and a low-resolution PWM.