Power topologies

Linear vs switching

Every DC-to-DC converter you meet is one of two families. The trade-off is heat versus complexity — and the choice is almost always obvious once you name it.

A linear regulator drops the difference between input and output voltage as heat. Convert 12 V down to 5 V at 1 A and you dissipate (12 − 5) × 1 = 7 W as heat in a package the size of a pencil eraser. That is a small camping stove. Efficiency is fundamentally Vout / Vin — nothing you can do improves it.

A switching regulator instead chops the input into fast pulses (100 kHz to a few MHz), stores each pulse's energy briefly in an inductor or capacitor, and reassembles the output at the level you want. Because it never dissipates the voltage difference, efficiency of 85–95 % is routine, and the same 12 → 5 V 1 A conversion wastes half a watt instead of seven.

The catch: switching regulators need an inductor, a diode or synchronous FET, a control loop, and careful layout to keep the fast edges from radiating. Linear regulators need one resistor and one capacitor. For low-current, low-drop, quiet loads (op-amp rails, RF LO supplies), the extra noise from switching often outweighs the efficiency benefit — that is why analog benches still ship with linear regulators.

Buck — step down efficiently

The buck is the most common switching topology on Earth. It takes a higher voltage in, delivers a lower voltage out, and does it 85–95 % efficiently.

A buck converter has three key parts: a high-side switch (usually a MOSFET), a low-side switch or a Schottky diode (the 'catch' element), and an inductor feeding a large output capacitor. The high-side switch turns on for a fraction D of each period, letting current ramp up through the inductor into the output. When it opens, the inductor keeps that current flowing through the low-side element, discharging into the output.

The average output voltage lands at Vin × D, where D is the duty cycle. A buck feedback loop measures Vout, compares it to a reference, and nudges D up or down every cycle to keep Vout locked in — that is why a buck ignores changes in load and input voltage until you exceed its control-loop bandwidth.

The inductor is the whole reason buck efficiency is so high: it stores energy in a magnetic field with almost no loss (a good inductor has a few tens of milliohms of DC resistance) and hands it back a moment later. The design trade-offs are switching frequency (higher = smaller inductor but more switching loss) and inductor ripple current (higher = smaller inductor but larger output ripple).

Boost — step up from low voltage

A boost converter takes a low voltage in and gives you a higher one out. It is how a 3.7 V lithium cell drives a 5 V USB or a 12 V LED strip.

A boost converter is a buck with its topology flipped. The inductor sits at the input; a switch grounds the inductor's output, letting current ramp up (energy stored in the magnetic field). When the switch opens, that magnetic energy has to go somewhere — it forces current through a diode into the output cap at whatever voltage is needed to keep going.

The result: the output voltage lands at Vin / (1 − D), where D is the duty cycle. At D = 0.5 the output is 2 × Vin; at D = 0.75 it is 4 × Vin. The theoretical maximum boost is infinite, but in practice inductor resistance and switch losses cap real-world gain to about 5×.

One consequence surprises everyone: boost converters have no way to stop conducting. Even when the switch is off, the inductor is still connected through the diode to the output. Short the output of a running boost and you draw as much current as the input can supply, straight through — a fault that requires an upstream fuse to survive.

Buck-boost — for batteries that pass through the rail

When a battery starts above and ends below your output rail, you need a topology that can go both ways. Buck-boost is that topology.

There are two families of buck-boost topology: inverting (a single inductor, but the output is negative relative to the input) and four-switch (two MOSFETs on each side of the inductor, so the converter can operate as either a buck or a boost depending on which pair is switching). The four-switch version is what almost every modern battery-powered device uses.

The four-switch chip watches Vin against Vout continuously. When Vin > Vout by a comfortable margin, the input pair PWMs and the output pair holds still — pure buck mode. When Vin drops close to or below Vout, the input pair holds still and the output pair PWMs — pure boost mode. A narrow band in between blends the two.

Buck-boost is more complex, slightly less efficient than either pure mode, and a bit more expensive. In return you get a rail that stays locked over the entire useful battery range, and a device that neither cuts out early nor stresses the pack by drawing too hard when it is nearly empty.

LDO dropout & current sensing

Two topics beginners meet at the same moment: how low can a linear regulator's input go before it stops working, and how do you measure how much current is flowing.

Every linear regulator needs some headroom between Vin and Vout to work. A classic LM7805 needs about 2 V — feed it 6.5 V and the output starts drooping. A modern LDO ('low-dropout') might need only 100–300 mV, which is why they took over battery-powered designs where every millivolt counts.

Dropout is not free — an LDO trades off dropout against quiescent current and load-transient response. A 100 mV-dropout LDO usually has higher quiescent current and slower response than a 500 mV-dropout part. There is no free lunch; the datasheet's dropout number at your operating current is the one that matters, not the marketing headline.

Current sensing puts a small, precise resistor (a shunt) in series with the load. Ohm's law: the voltage across it is proportional to the current. A 10 mΩ shunt at 1 A drops 10 mV — perfectly manageable, but too small to read reliably with a plain ADC, so a purpose-built current-sense amplifier (INA-family or similar) amplifies it by 50–500× first.

The two topics come together because current sensing is the feedback path that lets a regulator know when it is being overloaded. Every modern LDO has an internal current-sense loop that folds back the output when the shunt sees too much — that is why an LDO's short-circuit behaviour is graceful while an unprotected switcher's is dramatic.

Related