Decoupling Capacitors — Why Every IC Gets One and What Happens If You Skip It

Every reference schematic you've ever looked at has them: small capacitors, usually 100nF, placed as close as possible to the power pins of every integrated circuit. You've probably placed them yourself, copying the pattern because it's clearly the right thing to do.

But understanding why they're needed — what problem they actually solve — makes you a much better PCB designer and helps you debug the class of failures they prevent.

The problem: ICs draw current in spikes

Digital ICs — microcontrollers, logic gates, communication chips — don't draw current smoothly. Every time an output switches from low to high or high to low, a brief spike of current flows as the output driver charges or discharges the capacitive load on the pin. This happens in nanoseconds.

Your power supply and the copper traces that connect it to the IC can't respond that fast. The trace has inductance — it resists rapid current changes. By the time the power supply responds to the demand spike, the IC has already experienced a brief voltage dip on its power rail.

For slow logic this doesn't matter much. For fast logic, communication peripherals (SPI, I2C running at MHz frequencies), and anything with multiple outputs switching simultaneously, these voltage dips cause glitches. The IC gets confused about the state of its power rail and behaves unpredictably.

What the capacitor does

A capacitor stores charge. A 100nF capacitor placed right next to the power pins acts as a local energy reservoir. When the IC needs a burst of current faster than the power supply can deliver it, the capacitor supplies the current instead — and recharges from the power supply afterwards, slowly, in a way the supply can handle.

The capacitor decouples the IC's instantaneous demand from the supply's response time. That's the origin of the name.

Why placement matters so much

The whole mechanism depends on the capacitor being able to deliver current faster than the power supply. That means the path from the capacitor to the IC's power pin must be as short and low-inductance as possible.

If your decoupling capacitor is 2cm away from the IC, connected via a long thin trace, the inductance of that trace reintroduces the problem you were trying to solve. The capacitor can't respond fast enough because the trace is in the way.

Rule: the decoupling capacitor should be within 1–2mm of the power pin it's decoupling, and the connection to the power pin should be as direct as possible — ideally via a via directly under the capacitor to the power plane, with a short stub to the IC pin.

• 100nF (0.1µF): handles high-frequency switching noise, use one per power pin.
• 10µF: handles lower-frequency droop from larger current draws, one per supply rail.
• 1µF: bridging capacitor, sometimes used between 100nF and 10µF for mid-frequency.
• Placement: as close to the pin as physically possible, on the same side as the IC.

What happens if you skip them

On a breadboard prototype, you often get away without decoupling capacitors. Breadboards have high parasitic capacitance everywhere — the breadboard itself acts as a distributed decoupling network. It's inefficient and noisy, but it usually works well enough to not fail obviously.

On a real PCB, without decoupling, you'll see: sporadic resets of microcontrollers (especially under load), communication errors on SPI/I2C buses, ADC readings that are noisier than expected, outputs that randomly glitch when other parts of the circuit switch.

"I had an ESP32 that would reset randomly about once every 20 minutes. Added 100nF caps to every power pin and it ran for three weeks solid without a reset. That was the whole fix." — maker at a Bangalore electronics meetup

The 'bulk' capacitor is different

The large electrolytic capacitor you often see near the power input of a board (100µF–1000µF) is different from decoupling capacitors. It handles board-level supply droop when large loads switch on — like a motor driver engaging, or an LED array lighting up. It's not a substitute for per-IC decoupling.

Think of it this way: the bulk cap is the warehouse, the decoupling cap is the shelf next to the workbench. Both serve the same general purpose — available local charge — but at very different timescales and distances.