Re: What is DC to DC Converter
Hi daraemon_liu,
Yes, there certainly is more noise present in the output of a switching regulator than from a linear regulator, and you are correct - this arises as a consequence of the switching action of the power transistor (BJT, MOSFET or IGBT). The switching action induces noise via 3 main mechanisms:
1. Discontinuous power transfer from input->output, as the energy is tranferred through the converter (through inductors/transformers/capacitors/etc) in "bursts". This is reduced by the output filter, typically a few-pole LC configuration, with the corner frequency as far below the switching frequency as practical (given constraints of cost, size and saturation of the magnetics). Residual noise may be (typically) a few hundred millivolts in a crappy converter, to microvolt levels in a careful, low power design.
2. Capacitive coupling. Stray capacitance between switching and output nodes can couple energy during the switching transitions of the transistor. This is reduced by limiting switching rise/fall times, reducing the voltage excursions on switching nodes (via snubber networks, low-self inductance winding techniques etc) and minimising stray capacitance (via electrostatic shielding, geometry etc).
3. Magnetic coupling. This is particularly insidious as effective magnetic shielding is hard to accomplish, particularly at low frequencies. In this case, flux from switched inductors couples into output filter inductors/PCB traces, inducing ripple voltages. Shielded inductors (potcores, toroids etc), geometry and careful PCB layout reduce these effects.
Some of these mechanisms are bigger/lesser problems in the various converter topologies, but are fairly universal nonetheless.
While switching noise is absent from a linear regulator, they can have noises of their own which may or may not matter, depending upon their application. Be aware that even the ubiquitous LM7805 can oscillate horribly (2Vpp output ripple!) if they are inadequately decoupled (yes, you _do_ need the input/output 'stability' capacitors!). Input ripple will generally be attenuated (and converted to heat) and substantially reduced with a linear regulator, but only if its frequency lies within the loop bandwidth of the (active) regulator. The datasheet will give you some clues as to how much attenuation you'll get - look for something like "ripple rejection" - values of ~50+ dB (at 100Hz) are typical. In the case of a zener diode, the ripple is (to first order) reduced by the voltage divider formed by the series resistance and the diode's equivalent series resistance (and further filtered by any shunt capacitance).
Cheers!