When you need a negative rail or a voltage higher than your supply but don't want the cost, size, or EMI of an inductor-based switcher, reach for a charge pump. These circuits use only capacitors and switches to shuttle charge between nodes, doubling, inverting, or dividing voltages with surprising efficiency at low currents.

The classic Dickson charge pump uses two phases. In phase 1, a "flying capacitor" C1 charges to Vin through switches. In phase 2, those switches flip — C1's bottom plate is lifted to Vin while its top plate dumps charge into the output capacitor C2. The result: Vout ≈ 2×Vin minus diode drops or switch losses. Invert the connections and you get a negative rail instead.

Real-world example: The MAX1044 / ICL7660 has been generating −5 V from +5 V in op-amp circuits since the 1980s. Drop in two 10 µF caps and a 0.1 µF bypass, and you get roughly −4.5 V at up to 10 mA — perfect for biasing the negative rail of a single-ended op-amp without adding a second supply. Modern parts like the LM2776 push this to 200 mA with switching frequencies above 1 MHz, shrinking the caps to 1 µF ceramics.

Output impedance is the catch. A charge pump behaves like an ideal source with a series resistance set by switching frequency and flying-cap value:

• R_out ≈ 1 / (f_sw × C_fly)
• At f_sw = 10 kHz with C_fly = 10 µF: R_out ≈ 10 Ω
• Pull 10 mA, lose 100 mV. Pull 100 mA, lose a full volt.

This is why charge pumps shine for low-current rails: LCD bias, EEPROM programming voltage, op-amp negative supply, MOSFET gate drive above the rail (bootstrap circuits in half-bridges are essentially single-stage charge pumps). They're terrible for powering anything hungry.

Design tips:

• Use ceramics (X7R or better) for flying caps — ESR matters, and electrolytics waste charge each cycle.
• Higher f_sw = smaller caps for the same R_out, but more switching loss.
• Output ripple ≈ I_load / (f_sw × C_out) — size C2 for your noise budget.
• Efficiency is fundamentally limited: a doubler driving a load below 2×Vin dissipates the difference. Best efficiency happens when the load voltage is close to the unloaded output.

For ±15 V rails or hundreds of milliamps, use a switcher. For a clean −5 V at 5 mA on a tight board, a charge pump wins on simplicity every time.