The buck converter
Every page on this site that says "use a buck converter instead" has been pointing at four parts — a switch, a coil, a diode and a capacitor, arranged so that the regulator never has the input voltage and the load current in the same place at the same time.
A linear regulator makes 5 V from 12 V by sitting in the path and dropping the other 7 V across itself. Whatever the load takes, it takes through the regulator, so the waste is (12 − 5) × the load current and no better part can change that. At 1 A it is 7 W of heat to get rid of, for 5 W of useful output.
A buck converter does the same job without ever dropping a voltage across anything. Watch the coil's current rather than the circuit.
The switch is only ever fully on or fully off
That is where the efficiency comes from, and it is the same argument as a MOSFET as a switch: a part with no voltage across it dissipates nothing, and a part with no current through it dissipates nothing, and this switch is always in one of those two states.
What makes it possible to run a load off a switch that is off more than half the time is the inductor. Its current cannot stop, so during the off half it keeps pushing — pulling the switching node below ground until the diode conducts, and circulating through that loop instead. The capacitor fills in what is left.
Duty cycle is the control
Vout = D × Vin
Where D is the fraction of each cycle the switch is on. 5 V from 12 V is 42 %. 3.3 V from the same input is 28 %. That is the entire regulation mechanism: the chip measures the output, compares it against a reference, and adjusts how long it stays on. There is no analogue pass element and nothing burning the difference.
What it costs
The efficiency is not free, and the price is paid in four places.
- Noise. The switching node moves between 12 V and −0.4 V in a few nanoseconds, hundreds of thousands of times a second, and everything near it hears that. An ADC reading a thermocouple next to a buck converter reads the buck converter. Decoupling stops being optional.
- Layout. The loop carrying the switching current has to be physically small, because it is an aerial. This is the one circuit where a breadboard version genuinely behaves differently from a PCB version.
- Ratings that are easy to get wrong. The inductor has to survive the peak of the triangle, not the average — see the last page.
- Light loads. Many converters skip pulses when almost nothing is drawn, which drops the switching frequency into the audible band. That is the whine.
Below about two volts of drop, a linear regulator is still the right answer. It is two capacitors and a three-legged part, it makes no noise, and the heat is nothing to think about. 5 V to 3.3 V at 300 mA wastes half a watt — see what an AMS1117 is for. The buck converter is for the case where that arithmetic stops being survivable.
The variations
Same four parts, rearranged. A boost converter puts the inductor in series with the input and shorts its far end to ground to charge it; when the switch opens, the coil's voltage adds to the input's and the sum goes through the diode to the output, which is how you get more volts out than you put in. A buck-boost does either, which is what a lithium cell needs when it starts above 3.3 V and finishes below it. The inductor does the same job in all of them: it is the bucket the energy sits in between one switch closing and the next.
Edit this page — content/fundamentals/electricity/the-buck-converter.mdx
Discuss this article
Ask about this page. The answer stays here, on the page it belongs to, for whoever hits the same wall next.