Charge Pumps

Quick Answer

A charge pump moves energy with a capacitor rather than an inductor. Switches charge it from one place, then reconnect it somewhere else and let it discharge, so its stored voltage adds to or inverts the supply. It needs no magnetics at all, and it delivers only small currents.

A canal lock does not push water uphill. It takes a boat sitting at one level, closes a gate behind it, fills the chamber, and opens a gate at the other end. Nothing has been pumped in the sense of being forced along a pipe. A quantity was isolated at one level and released at another, and the work was done by opening and closing gates in the right order.

A charge pump is that, with charge instead of water. A capacitor is connected across the supply and charges to it. Then it is disconnected, both of its terminals are reconnected somewhere else, and it discharges into whatever is waiting there. Because a capacitor holds a voltage difference rather than a voltage, moving its lower plate up to the supply rail takes its upper plate to twice the supply — and moving its upper plate to ground takes its lower plate below ground, which is how the same circuit makes a negative rail.

That is the whole idea, and it explains both of the reasons to use one. There is no inductor, so there is no magnetic field, no core to saturate, no winding to radiate, and nothing that has to be laid out with care. And the output is not limited to something below the input the way a linear regulator's is: a charge pump can hand back more voltage than it was given, or a voltage of the opposite sign, from a supply that has neither.

The diode ladder is the same principle with the gates made of diodes rather than switches, and it is where the arrangement is easiest to see: each stage carries charge one rung further up, and the multiplication is just the count of rungs. Replacing those diodes with MOSFETs removes their forward drops, which is what makes the integrated version practical at low supply voltages where a diode drop is a large fraction of everything available.

What it will not do is deliver much current, and the reason is structural rather than a limitation of any particular part. Each cycle transfers exactly the charge the flying capacitor holds, so the output current is that charge times the switching rate and nothing else. Push harder and the output voltage falls, because the capacitor has not finished charging before it is asked to move — which appears in a datasheet as an output resistance and behaves like one. Wanting more current means a bigger capacitor, a faster clock, or an inductor, and past a few tens of milliamps the answer is usually an inductor and a switching regulator.

Two things follow that are worth carrying. The output ripples at the switching rate, because the transfer is discrete, so a charge pump is a poor choice for anything sensitive sitting next to it unless it is filtered or its clock is synchronised to something. And the flying capacitor is switched between rails thousands of times a second, so it carries the full transfer current in short pulses — a job for a low-loss ceramic, not for whatever was in the drawer.