Analysis and Design of Magamp Post-Regulator for Flyback Converters

• Main Authors: Chau-Chun Wen, 溫兆均
• Other Authors: 陳秋麟
• Format: Others
• Language: en_US
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/43459758836359957730

博士 === 國立臺灣大學 === 電機工程學研究所 === 93 === Applications of the magamp technique are mainly focused on multiple-output forward converters. However, the drawbacks of forward converters are more part counts and higher cost compared with flyback converters. In this dissertation, a new magamp technique for flyback converters is proposed. Using the current sharing approach, the outputs are powered in different time slots. In this way, the multiple output voltages can be regulated. In this dissertation, the detailed studies of the operation principle and system dynamics are proposed. Unlike other traditional topologies, the input/output relationships are depended on the average currents. Both steady-state model and small-signal model is derived with state-space averaging approach. Since the multiple outputs share the magnetizing energy of the coupled inductor, the interactions are found among the outputs. The existence of right half plane zero in single output flyback converter is also moved to the interaction loops. However, the cross interactions among outputs are small under a specified condition. A dominant function compensation method is presented. Signal flow graph technique is utilized to design the compensator. Experimental verifications on a 20W two-output flyback converter are conducted. Recordings illustrate the effectiveness of the proposed magamp approach, the small-signal modeling and compensator design. At last, the operation area of practical circuit will be bounded by the non-ideal component characteristics. An analytical model for studying the phenomenon is proposed. Based on the model, the mechanism of boundary condition that causes the converter out of regulation is explored. Finally, experimental results are provided to show the accuracy of the proposed analytical model.