Modeling And Python Simulation Of Magnetics For Power Electronics Applications PDF Download

Power electronics represent a key technology for improving the functionality and performance, and reducing the energy consumption of many systems. However, the size, cost, and performance constraints of conventional power electronics currently limit their use. This is especially true in relatively high-voltage, low-power applications such as off-line power supplies, light-emitting diode (LED) drivers, converters and inverters for photovoltaic panels, and battery interface converters; a LED driver application serves as a motivation example throughout the thesis. Advances in the miniaturization and integration of energy-conversion circuitry in this voltage and power range would have a tremendous impact on many such applications. Magnetic components are often the largest and most expensive components in power electronic circuits and are responsible for a large portion of the power loss. As operating frequencies are increased, the physical size of the passives can, in theory, be reduced while maintaining or improving efficiency. Realizing this reduction in size and the simultaneous improvement in efficiency and power density of power electronic circuits requires improvements in magnetics technology. This thesis focuses on the challenge of improving magnetics through the analysis, optimization, and design of air-core toroidal inductors for integration into high-efficiency, high-frequency power electronic circuits. The first part of the thesis presents the derivation of models for stored energy, resistance and parasitic capacitance of microfabricated toroidal inductors developed for use in integrated power electronics. The models are then reduced to a sinusoidal-steady-state equivalent-circuit model. Two types of toroidal MEMS inductors are considered: in-silicon inductors (with or without silicon core) and in-insulator inductors. These inductors have low profiles and a single-layer winding fabricated via high-aspect-ratio molding and electroplating. Such inductors inevitably have a significant gap between winding turns. This makes the equivalent resistance more difficult to model. The low profile increases the significance of energy stored in the winding which, together with the winding gap, makes the equivalent inductance more difficult to model as well. The models presented in this thesis account for these effects. In the case of in-silicon inductors, magnetically and electrically driven losses in different regions of silicon are modeled analytically as well. The second part of the thesis focuses on the optimized design of microfabricated toroidal inductors for a LED driver. The models developed in the first part of the thesis allow optimization of inductor designs based on objectives such as minimizing substrate area, maximizing efficiency, and simplifying the fabrication process by maximizing minimum feature size. Because the magnetics size and loss depend strongly on the driver design parameters, and the driver performance depends strongly on the inductance value and loss, the simultaneous optimization of driver components and magnetics parameters is used in the design process. The use of computationally efficient models for both magnetics and other circuit components permits numerical optimization using the general co-optimization approach. Finally, a multi-dimensional Pareto-optimal filtering is applied to reduce the feasible design set to those on the multi-objective optimality frontier. For the case of LED drivers, the current state of the art efficiencies range from 65% to 90%. The co-optimization process results in efficiencies greater than 90% while reducing the size of the LED driver by 10 to 100 times compared to the commercially available LED drivers. This is a significant improvement in both the efficiency and the size of the LED drivers. In the resulting designs, the magnetics are no longer the largest part of the circuit. In the third part of the thesis several numerical and experimental tests are presented. The models developed in this thesis, are verified against results from 2D FEA, 3D FEA, direct measurement of MEMS fabricated devices (for both in-insulator devices for flip-chip bonding and in-silicon devices for direct integration), and in-circuit experimentation of the fabricated devices. These tests show that the equivalent-circuit models presented in this thesis have greater accuracy than existing models. The results also show that these models are good enough to support the LED driver optimization.