Buck, Boost and Buck-Boost

Switching a supply gives an average of D·V_in with nothing dissipating the difference. The problem is that the instantaneous output is a square wave, and a load generally wants a steady voltage.

An inductor resists changes in current, turning the chopped voltage into a nearly steady current with a triangular ripple. The diode gives that current a path when the switch opens — mandatory, since interrupting an inductor's current produces a destructive spike.

In a boost converter, opening the switch forces the inductor's voltage to reverse and add to the input. Nothing is created — the output current falls in proportion, and energy is conserved exactly.

In continuous conduction the inductor current never reaches zero and V_out depends only on duty cycle. In discontinuous conduction it hits zero and rests there, and the transfer function acquires a load dependence.

A controller tuned for one mode misbehaves in the other, and light load is where the transition happens. Synchronous rectification changes it again, since a MOSFET can conduct backwards where a diode cannot.

Mains safety requires galvanic isolation, which requires a transformer, which requires AC — so the switching frequency becomes the transformer's operating frequency.

A regulating converter measures its output, compares it with a reference and adjusts the duty cycle — with all the stability questions that implies. The output LC filter contributes a pole pair, so compensation is designed with Bode plots and phase margin.

A boost converter additionally has a right-half-plane zero: increasing the duty cycle momentarily *reduces* the output, because the inductor is diverted from the load while it charges. That fundamentally limits how fast its loop can be, regardless of the compensator.