Modeling and Design of Voltage Regulator Control Schemes for Computer Power Applications

Computing devices such as personal computers and laptop computers have become a major market for power supply industry. To meet the various power requirements of a computing device, a variety of DC/DC voltage regulators (VR) are used to power loads such as central processing unit (CPU), graphic processing unit, and I/O power. Due to the advancement of computing devices, the power requirements of VRs have become more and more stringent. Besides, VR efficiency has always been an important research topic. In recent years, the mandate for higher efficiency, not only for the full load but also for the light-load condition, has become even more critical. Conventional control schemes cannot meet such requirements; therefore, novel control schemes are proposed to solve the problems. There are two major trends to meet the high efficiency requirement. One is to use adaptive-voltage positioning (AVP) scheme. The other is to use constant on-time controller, instead of conventional constant-frequency controller, to increase the efficiency at light load. The focus of the dissertation is to develop analytical models and provides the design guideline of two recently-reported control schemes: high-gain peak current control (HGPCC) scheme and the modified ripple-based constant on-time (RBCOT) control. HGPCC scheme has the advantages of phase-current balancing and cycle-to-cycle over current protection. Besides, the scheme can be designed to achieve accurate AVP for CPU power loss reduction. The modified RBCOT control has the advantages of fast transient response, small components count requirement, accurate output voltage, and high efficiency at light-load.
Because of the fundamental difference in the pulse-width modulation schemes used in the two new control schemes, the conventional small-signal average model does not apply. In developing the models, a modified average modeling approach is used for the HGPCC scheme, and the describing function approach is used for the RBCOT scheme. Based on the model, design guidelines are given to achieve prescribed control characteristics. Simulation and experimental results confirm the proposed models and the accuracy of the stability criteria.