The Design Of Buck DC-DC Autotuning Algorithm Based On System Identification

With widespread applications of Artificial Intelligence(AI)and Big Data,the server cluster develops rapidly and a mass of data payload needs a better dynamic performance on its Power Management IC(PMIC).However,the problem due to non-optimal configuration of the loop,aging of power stage devices and deviation of manufacturing process will deteriorate the dynamic performance,which will even influence the loop stability.In response to the above problems,an autotuning algorithm of buck DC-DC converters based on system identification(SI)is designed in this thesis.Firstly,the basic working principles of the designed circuit are introduced.Then the following main work is carried out:(1)The transfer function of power stage is derived by using the state space averaging method,and the corresponding form of compensation transfer function is designed.(2)The system identification algorithm is designed,including the compensation scheme of sample delay,the optimization scheme of system identification and the extraction and modifying scheme of power stage parameters which successfully realizes the extraction of high precision power stage parameters.(3)The autotuning algorithm is proposed,including the judgement of phase margin(PM),multiple crossover frequency(MCF),limit cycle oscillation(LCO)and closed-loop output impedance.And the best compensation parameters are sifted out throuth proposed autotuning algorithm,which achieves high performance of dynamic response.Finally,according to the proposed algorithm,the Simulink simulation platform is built to achieve algorithm level and Register Transfer Level(RTL)simulation verification and the test prototype is built.The Field Programmable Gate Array(FPGA)verification platform is used to verify the autotuning algorithm based on system identification.The simulation and measured results show that the SI algorithm designed in this thesis realizes the extraction of power stage parameters.The error of power stage parameters is 4.42%and the duration of system identification is 100 ms.The autotuning algorithm which is based on closed-loop output impedance realizes the sifting of best compensation parameters,which lasts for 124 ms.With a load change from 5A to 0A,the overshoot and response time are 70 m V and80?s respectively,which is a decrease of 36% and 20% compared to traditional compensation.With a load change from 0A to 5A,the undershoot and response time are 80 m V and 80?s respectively,which is a decrease of 38% and 33% compared to traditional compensation.The test results meet the requirements of design specification.