Multi-Objective Direct Quasi-Sliding Mode Control of a Two Phase Buck Converter

Abstract

Multiphase converters are the state of the art for high efficiency point-of-load conversion for modern processors and FPGAs with wide operating load conditions. However, there are a variety of control issues that must be addressed, such as dynamically adjusting the number of active phases, robustness to disturbances, and phase current regulation. To achieve the highest efficiency conversion, the phase currents should not necessarily be split evenly between phases. Unfortunately, most of the literature concerning multiphase buck converters implements equal current sharing between phases.

Sliding mode controllers are known for their robustness and insensitivity to disturbances and plant parameter variations. Sliding mode control is particularly suited for power electronics due to the inherent switching between two system structures. The control effort of the sliding mode controller depends solely on which side of the surface the system state currently resides. This effort can be implemented in two ways: direct and indirect.

This work designs, simulates, and implements a direct sliding mode controller for a two-phase buck converter. The technical approach extends a single-phase sliding mode controller to control a two-phase buck converter. Since there are two phases, two switching surfaces are necessary. Thus, suitable surfaces must be developed to address the challenges of phase delay in addition to the overall goal of output voltage regulation. This work is tested in simulation with good results and is implemented with some modifications on a hardware test platform consisting of a TI TMS320F28335 microcontroller and two buck converter phases. The proposed controller can be extended to support regulating to desired phase's currents and dynamically change the numbers of active phases based on load conditions. This work confirms that the sliding mode controller is a suitable controller design for multiphase buck converters and is the basis for future work improving current sharing, implementing phase shedding, and expanding the maximum number of active phases.