PCB Trace Width Calculator
Minimum trace width for a current and temperature rise per IPC-2221, on inner or outer layers and any copper weight, with the resistance and voltage drop.
Amps, continuous.
°C above ambient. 10 is the usual design figure.
1 oz = 35 µm.
mm, for resistance and drop.
°C inside the enclosure, not the room.
Minimum width
0.3 mm → +10.0 °C · 0.4 mm → +6.2 °CThe two fab widths either side of the minimum, with the rise each one actually gives at this current.
16.3 mil² (10.5 × 10⁻³ mm²)16.0 mΩ16.9 mΩ16.9 mV16.9 mW35 °CSame current, other copper
Minimum width for 1 A at +10 °C across the copper weights a fab offers.
| Copper | Outer layer | Inner layer |
|---|---|---|
| 0.5 oz | 0.601 mm (23.7 mil) | 1.56 mm (61.5 mil) |
| 1 oz | 0.300 mm (11.8 mil) | 0.781 mm (30.8 mil) |
| 2 oz | 0.150 mm (5.9 mil) | 0.391 mm (15.4 mil) |
| 3 oz | 0.100 mm (3.9 mil) | 0.260 mm (10.3 mil) |
Check a width you already have
mm, on the copper and layer above.
1.45 A+4.3 °C (copper at 29 °C)Where the formula comes from
IPC-2221 gives the relationship between a trace's cross-section, the current through it, and how much hotter than its surroundings it runs: I = k · ΔT0.44 · A0.725, with the current in amps, the rise in degrees, the area in square mils, and k of 0.048 for an outer layer or 0.024 for an inner one. It is a curve fitted to charts that the US National Bureau of Standards drew in the 1950s from measurements on the boards of the time, carried into the IPC standards and used ever since.
In 2009 IPC-2152 replaced the charts with new measurements, and for most outer-layer traces on FR-4 the older formula turns out to be conservative — it asks for more copper than the newer data says is needed. That makes it a reasonable first pass, which is how this tool presents it. Its stated range is 0–35 A, rises of 10–100 °C, widths to 400 mil and copper from ½ to 3 oz; outside that the tool still computes a number but flags it as extrapolation.
Temperature rise is the design choice
The formula does not tell you how wide a trace has to be. It tells you how hot a given trace gets, and you choose the rise you are willing to accept. Ten degrees is the conventional answer: the copper stays well below anything that stresses the laminate, nearby parts do not notice, and there is margin for the formula's own approximations. Twenty is common when space is short. Past thirty the question becomes what else on the board is being heated.
The rise is above ambient, and the ambient that matters is the air inside the enclosure next to the board, which in a sealed box in a warm room can be 50 or 60 °C before anything on the board contributes. The tool adds the two and reports the copper temperature, and warns when it approaches the point where standard FR-4 and its solder joints start to be uncomfortable.
Inner layers need two and a half times the width
An outer trace has air on one side and can radiate and convect. An inner trace is wrapped in laminate, which conducts heat poorly, so the same current warms it more. IPC-2221 halves k for inner layers, and because area goes as the 1.38th power of 1/k, the inner width is 21.38 ≈ 2.6 times the outer one — not double, which is the usual guess. A 1 A trace that is 0.3 mm on the outside needs 0.78 mm inside. For anything carrying real current, the outer layers are the cheap place to route it.
Copper weight is the other lever
Doubling the copper thickness halves the width for the same cross-section, which is why power boards use 2 oz. It is not free: heavier copper needs wider spacing, coarser minimum features, and costs more, and it does not help a via. The comparison table shows the same current across every weight a fab offers, so the trade between width and copper is visible before the stack-up is chosen.
One subtlety: outer layers are usually plated during manufacture, so a board ordered as 1 oz often finishes with 1.3–1.4 oz on the outside. That is extra margin in your favour, but it varies by fab and process, so the tool works from the nominal weight you specify.
Resistance and drop are a separate check
Passing the thermal test does not mean the trace is electrically invisible. Copper is about 17 nΩ·m at room temperature and gains 0.4 % per degree, so the tool computes the resistance at the operating temperature, then the drop and the power the trace itself dissipates. A 10 mm run of 0.3 mm 1 oz copper is about 16 mΩ — nothing for a logic signal, but 5 A through 200 mm of it is 1.6 W of heat and 320 mV of drop, which is a problem for a 3.3 V rail and a disaster for a current-sense line. Sense connections should be routed separately as Kelvin pairs so the drop in the power trace does not end up in the measurement.
Round up to what the fab makes
The minimum is a minimum. Fabs make widths in steps, etching takes a little copper off each edge, and their published minimum width is for signals rather than for current. So the tool shows the standard widths just under and just over the minimum with the rise each one actually produces at your current. If the width just under gives 10.1 °C against a 10 °C target, that is a judgement; if it gives 14 °C, take the next one up. The LED resistor and other calculators here set currents; this one sizes the copper that carries them.
Where the formula stops applying
IPC-2221 is for one continuous DC current in an isolated trace on an ordinary board in still air. Several hot traces side by side heat each other. A trace under a component sees the component's heat. A pulse of milliseconds can carry many times the continuous rating because copper has thermal mass, and a trace in a sealed box with no airflow will run hotter than the formula says. Vias are a different problem altogether, with a plated barrel whose cross-section is far smaller than the trace feeding it. Treat the number here as the start of the thermal design, not the end of it.
Frequently asked questions
What temperature rise should I design for?
10 °C is the conventional figure and what most calculators default to; it keeps the copper well clear of anything that matters and leaves margin for the parts of the formula that are approximations. 20 °C is common where board space is tight. Beyond 30 °C you are relying on the board and everything near the trace tolerating the heat, and the rise adds to the ambient inside the enclosure, not the room.
Why does an inner layer need so much more width?
Because it cannot shed heat to the air. IPC-2221 halves the constant for inner layers, and since area scales as the 1.38th power of that, the width comes out about 2.6 times larger — not twice, which is the figure people often assume. Routing a high-current trace on an inner layer is expensive in width; keep them on the outside when you can.
How accurate is IPC-2221?
It is a curve fit to charts drawn in the 1950s, which IPC-2152 replaced in 2009 with measurements on modern boards. For ordinary outer traces on FR-4 the 2221 figure is conservative — the newer data allows more current for the same rise. It is still the number nearly every calculator and datasheet quotes, and a safe first pass. If you need every last amp, use IPC-2152 and its correction factors.
Should I round up to a standard width?
Yes, and the tool shows you what that costs or gains. The minimum is a minimum; the fab makes widths in steps, and its etching tolerance shaves some copper off the edges. The two bracketing widths are shown with the rise each actually produces at your current, so you can see whether the width just under the minimum is close enough or the next one up is worth it.
The trace is thermally fine. Why does the tool show a voltage drop?
Because heat and drop are different problems. A trace that only warms 10 °C at 5 A can still drop 50 mV over its length, which matters for a current-sense line, a reference, or an ADC input, and a long supply run can drop enough to trip a brownout detector. The resistance is calculated at the operating temperature, since copper is about 0.4 % per degree more resistive as it warms.
What about vias, pours and short pulses?
Not covered here. A via carries current through its plated barrel, which is a much smaller cross-section than the trace, so a high-current net usually needs several. A copper pour behaves as a very wide trace but with its own thermal behaviour. And a pulse of a few milliseconds can carry far more than the continuous rating because the copper has thermal mass; IPC-2221 is for continuous current.
The code behind this
The arithmetic on this page is published as the trace module — the same functions you are using here, with the same tests. MIT, no dependencies.
Read the sourceOn npmnpm i electronics-math
Related tools
- I²C Pull-Up Resistor CalculatorThe window between pulling low and rising in time, with the standard values inside it.
- NTC Thermistor CalculatorADC counts to temperature, with the resolution and self-heating the divider decides.
- Ohm's Law CalculatorAny two of V, I, R and P give the other two, and it shows its working.