| This dissertation addresses several design and control issues of modern, complex switch-mode power converters: controller optimization, the effects of parametric uncertainty on converter small-signal dynamics, and converter and input filter interactions.; A new software design tool, which uses genetic algorithms and PSpice to optimize converter performance, is presented. Optimizations with concurrent time domain simulations with switched converter power stages and frequency domain simulations with averaged converter models are possible with GenSpice. Several optimization examples are given, including design of a controller that minimizes converter closed-loop output impedance.; Two solutions to the problem of parametric uncertainty are considered. Adaptive model following control (AMFC) based on hyperstability and positivity concepts is applied to buck, boost and buck-boost converters. A buck converter with AMFC is implemented in hardware. It is shown that AMFC reduces variations in performance characteristics, such as open-loop gain, audiosusceptibility, and input and output impedances, when power stage components change value.; mu-analysis and synthesis is the second solution investigated. Parametric uncertainty in converter power stages is isolated with a linear fractional transformation. Robust stability and performance of closed-loop converters are evaluated with structured singular value calculations. A mu-synthesis technique known as D-K iteration is used to design a controller to optimize converter performance characteristics.; Finally, converter and input filter interactions are mitigated with controllers designed to increase converter input impedance. Three controller designs are considered: a 2-stage, phase-lead controller with an integrator optimized with GenSpice, a novel adaptive model following controller (mAMFC), and an optimum controller designed with D-K iteration. The mAMFC design is implemented in hardware and compared to simulation results with the GenSpice and mu-synthesized controllers. |
