| With the rapid increase in the amount of information handled by the data center,the performance of its core processor has been greatly improved,the power supply level of the server has increased rapidly,the loss on the power distribution and transmission bus has increased squarely,and the overall power supply system efficiency has dropped significantly.Recent research hotspots and industry trends show that the efficiency of the power supply system will be effectively improved by increasing the power distribution of server racks from the traditional 12 V to 48 V.However,compared to Voltage regulator modules(VRM)with 12 V input,the significant challenge of simultaneously achieving high efficiency,high power density and high dynamics under higher conversion ratio conditions in 48 V VRM is presented.Among the many existing 48 V VRM structures,the sigma converter has obvious advantages in terms of efficiency and power density,but because the structure was proposed for a short time and is relatively complex,there is little research in academia,and there is no dedicated control chip in the industry,which restricts further applications of this structure.Therefore,investing the control strategies of sigma converter comprehensively and deeply becomes meaningful.In this thesis,the core issues such as accurate small-signal model,low-cost and high-accuracy digital current sampling algorithm and high-dynamic control strategy of sigma converter under multimode control are studied.The main innovative research contents are as follows:(1)Aiming at the lack of accurate power stage transfer function model for sigma converters,a small-signal model for voltage mode control is constructed and a decoupling derivation method is proposed to obtain the transfer function,which is more accurate than the model in the existing literature.The method is extended to the small-signal modeling of various current mode control(fixed-frequency peak current mode control and COT control,etc.),and the model accuracy reaches95% in the range of 1/2Fsw.(2)Aiming at the high cost and high noise of the digital sampling system in current mode control,a current estimation algorithm that only uses low-speed ADC and comparator is proposed in this thesis.The algorithm is based on the DCR current sampling principle.A digital high-frequency filter is used to filter the voltage across the inductor to obtain complete inductor current information.At the same time,the source of error in the estimation is quantitatively analyzed,and a new method to improve the estimation accuracy is proposed.Through the prototype verification board test,the DC accuracy error of the estimated current is less than 5%,the AC accuracy error of that is less than 2%,and the estimated delay is within 100 ns.(3)Aiming at the problem of insufficient dynamic performance of the existing loop control methods of sigma converters,based on the small signal model derived in(1)and the current estimation method in(2),various digital loop control strategies,such as voltage mode control with high precision DPWM,peak current mode control and COT control based on current estimation for sigma converters are carried out in this thesis.In the sigma converter,the combined control of COT and AVP is proposed for the first time,which greatly improves the overall dynamic response performance.The prototype test shows that under the conditions of 20 A to 80 A load switching and the AVP voltage window is designed to be 1m V/A,the overshoot voltage is 10 m V,and the corresponding response time is 60μs;without undershoot voltage,the corresponding response time is59μs. |
PDF Full Text Request