PCB Design
What is a decoupling capacitor?
A decoupling capacitor is a small capacitor placed right beside a chip's power pins. When the chip suddenly draws more current, the capacitor supplies it at once, while the current through the board's traces is still catching up. Without one, the chip's supply rail sags every time it switches.
A trace is an inductor
Besides its resistance, a copper trace has inductance: about a nanohenry for every millimetre. A long run to a chip, with the supply's leads, can reach a microhenry. An inductor's current can't jump. When a chip switches on and asks for 2 A, the trace delivers it over about half a microsecond, not at once.
Meanwhile the chip's rail collapses: the chip pulls, the trace can't give, and the voltage at the chip drops toward zero and climbs back as the current builds. That's a brief brownout, every time the chip wakes.
A reservoir beside the chip
A capacitor at the chip hands over the current the trace can't. It loses voltage while it does:
2 A for 0.1 µs out of 10 µF is only 0.02 V. Out of 100 nF it would be 2 V, and out of 1 nF the charge is gone almost at once. Size the capacitor for the current, how long it lasts, and the dip the chip can take.
Live experiment
Switch the chip on, watch the rail
A 5 V supply feeds a chip through 1 µH of trace. Pick a capacitor and where it goes, beside the chip or at the supply's end, then switch the chip on and watch the rail at the chip over 12 µs.
“Hold the rail up” is lesson 3 of 6 in Copper Cay, PCB Design World 1. Free with an account. Lessons unlock in order, so you’ll do the 2 before it first.
Position matters
The capacitor must sit on the chip's side of the trace's inductance. Put the same capacitor at the supply's end, across the connector, and its current has to come through the same inductance: the chip sees the same sag. That's why decoupling capacitors crowd the pins, millimetres away, and why every chip gets its own.
What it doesn't fix
The capacitor is for the instant. Once the current has settled, the copper's resistance sets where the rail rests. In the lesson's circuit, the supply's 0.05 Ω and the trace's 0.1 Ω take 0.28 V of the chip's 1.9 A, so the rail rests at 4.72 V, whatever the capacitor. And a capacitor and a trace's inductance can slosh charge back and forth after a sudden step: with 10 µF, the rail dips to about 4.37 V a few microseconds later before it settles. Real boards calm that with a bigger capacitor nearby as well.
Look at almost any circuit board and you'll find a 100 nF capacitor beside every chip's power pins. A motherboard has hundreds.
Free lesson
Now build it yourself
Each lesson comes in three short parts: see the idea, try it on a live board, then build it yourself, with quick checks as you go.
“Hold the rail up” is lesson 3 of 6 in Copper Cay, PCB Design World 1. Free with an account. Lessons unlock in order, so you’ll do the 2 before it first.
Or start from the very first lesson, no account neededWork it out
Practice questions
Three quick questions from the lessons. Pick an answer to check it.
Show the answer
Answer: Beside the chip, on its side of the trace's inductance. The inductance is between the supply and the chip. A capacitor at the supply end is behind it and can't help the chip. Beside the chip, it can.
- Why not “Anywhere on the rail”? Where it sits matters. Only a capacitor on the chip's side of the trace's inductance can deliver current the instant the chip needs it.
- Why not “At the supply”? At the supply end, the trace's inductance sits between the capacitor and the chip, so the capacitor's current is held back just like the supply's.
- Why not “In series with the trace”? In series, all the chip's current would have to pass through the capacitor, which blocks steady DC. It goes across the supply, from the power pin to ground.
Show the answer
Answer: Any trace between them is an inductor too, and slows the capacitor's help. A long trace from the capacitor has the same problem as the long trace from the supply. The shorter the path, the faster the help arrives.
- Why not “It looks neater”? Looks have nothing to do with it. The trace between the capacitor and the pin is an inductor, and a longer one slows the current the chip needs.
- Why not “It charges faster”? The capacitor's job here is to hand charge out, not take it in. A short path matters because it puts less inductance in the way of that current.
- Why not “It doesn't matter”? It matters a lot: each millimetre of trace adds about a nanohenry, which holds back the capacitor's current just when the chip needs it.
Show the answer
Answer: Within a few millimetres of the power pin, with short fat traces. Every millimetre of trace between the capacitor and the pin is a nanohenry in the way. Decoupling capacitors crowd the pins.
- Why not “Within 10 cm”? 10 cm of trace is about 100 nH, enough to hold the capacitor's current back from the chip. It needs to be millimetres away.
- Why not “On the same board”? Being on the same board isn't enough. Every millimetre of trace between the capacitor and the pin adds about a nanohenry of inductance.
- Why not “Near the connector”? The connector is where the supply comes in, the far end of the trace. A capacitor there helps the supply, not the chip.
“Hold the rail up” is lesson 3 of 6 in Copper Cay, PCB Design World 1. Free with an account. Lessons unlock in order, so you’ll do the 2 before it first.
Questions people ask
What does a decoupling capacitor do?
It's a small store of charge right beside a chip's power pins. When the chip suddenly draws more current, the capacitor supplies it for the first moments, before the current through the board's traces can catch up, so the chip's supply voltage doesn't dip.
Why is a decoupling capacitor usually 100 nF?
100 nF is a common, cheap value that covers the fast, short current pulses of most logic chips. Boards often add a larger capacitor, several microfarads, nearby for longer demands. The right size depends on how much current is drawn, for how long, and how much dip the chip can take: ΔV = I × t ÷ C.
How close should a decoupling capacitor be to the chip?
As close as you can, millimetres from the power pin, with short connections to the pin and to ground. Every millimetre of trace adds about a nanohenry of inductance between the capacitor and the chip, which slows the help down.
Related topics
- How does a capacitor work?Two plates and a gap: how a capacitor stores charge and energy, and why it blocks steady DC.
- How does an inductor work?A coil that fights changes in its current, the high-voltage spike it makes, and how to tame it.
- How does a Zener diode work?A diode built to conduct backwards at a set voltage, then hold it: the simplest regulator.