Trace width

What it computes

How wide a copper trace has to be so that a given DC current heats it by no more than a chosen amount above ambient. Two data sets are on offer, and they do not agree.

IPC-2221 (Generic Standard on Printed Board Design, Section 6.2, conductor sizing charts) is the one every older calculator uses. The curves were drawn in the 1950s from a small set of measurements; the closed form everyone fits to them is

I = k · ΔT^0.44 · A^0.725          I in A, ΔT in °C, A in mil²
k = 0.048  external layers
k = 0.024  internal layers

A = (I / (k · ΔT^0.44))^(1/0.725)
W_mil = A / (1.378 · oz)            1 oz/ft² = 1.378 mil = 34.79 µm

IPC-2152 (Standard for Determining Current Carrying Capacity in Printed Board Design, 2009) replaced those charts with a much larger measured data set. It is published as charts, not equations. The calculator uses the common curve fit of the universal chart — 1 oz external copper, still air, 70 mil board, no plane nearby — which tracks the chart to within about ±10 %:

I = 0.0454 · ΔT^0.5453 · A^0.6533     same units
A = (I / (0.0454 · ΔT^0.5453))^(1/0.6533)

IPC-2152's most useful finding is that a copper plane within about 1 mm of the trace roughly halves the temperature rise for the same current. The optional plane correction applies exactly that: the predicted rise is halved, which is the same as solving the fit with 2·ΔT. It is a rule of thumb lifted from the standard's correction charts, not a separate fit.

Any of the three variables can be the unknown — width, current or temperature rise — the equation is just rearranged. Given a trace length the tool also reports DC resistance, voltage drop and dissipated power at the operating temperature:

R = ρ · L / (W · T) · (1 + α · (T_ambient + ΔT − 20 °C))
ρ_Cu = 1.68e-8 Ω·m     α = 0.00393 /°C
V = I · R      P = I² · R

Worked example

3 A, 10 °C rise, 1 oz copper, external layer, 100 mm long, 20 °C ambient.

IPC-2221 external
  ΔT^0.44      = 10^0.44        = 2.754
  k · ΔT^0.44  = 0.048 · 2.754  = 0.1322
  I / that     = 3 / 0.1322     = 22.70
  A            = 22.70^(1/0.725) = 22.70^1.379 = 74.2 mil²
  W            = 74.2 / 1.378   = 53.8 mil = 1.37 mm

IPC-2221 internal (k = 0.024)
  A = (3 / (0.024 · 2.754))^1.379 = 45.40^1.379 = 192.9 mil²
  W = 192.9 / 1.378 = 140.0 mil = 3.56 mm

IPC-2152 universal chart, no plane
  ΔT^0.5453        = 10^0.5453        = 3.510
  0.0454 · 3.510   = 0.1594
  I / that         = 3 / 0.1594       = 18.83
  A                = 18.83^(1/0.6533) = 18.83^1.531 = 89.4 mil²
  W                = 89.4 / 1.378     = 64.9 mil = 1.65 mm

Resistance of the 1.65 mm trace at 30 °C
  W · T = 1.6476e-3 m · 34.79e-6 m = 5.732e-8 m²
  R20   = 1.68e-8 · 0.1 / 5.732e-8  = 29.3 mΩ
  R30   = 29.3 · (1 + 0.00393 · 10) = 30.5 mΩ
  V     = 3 · 0.0305 = 91 mV
  P     = 9 · 0.0305 = 0.27 W

So IPC-2221 external says 1.37 mm, IPC-2152 says 1.65 mm, and IPC-2221 internal says 3.56 mm. Read that carefully: at this current the 2152 universal chart is slightlymore conservative than 2221 external, and 2221 internal is more than twice as wide as anything 2152 measured. Put a plane 0.2 mm under the trace and 2152 with the plane correction gives 3 A at 10 °C on 0.92 mm (solve with ΔT = 20 °C: A = 50.1 mil², W = 36.4 mil).

Where it stops being valid

• IPC-2221 range. The charts cover up to 35 A, 100 °C rise, 400 mil width and 0.5–3 oz. Past that the equation still produces a number; the tool flags it as extrapolation.
• IPC-2221 internal is not a measurement of internal traces. It is the external curve halved, a safety factor applied in the 1950s. IPC-2152 measured real internal traces and found they run at most a few percent hotter than external ones in still air, and often cooler, because the laminate conducts heat better than air does. The IPC-2152 "conservative" chart is literally the old 2221 internal curve, kept for people who need to cite it.
• IPC-2152 universal chart assumptions. 1 oz, external, still air, 70 mil thick board, plane far away. Thinner boards, heavier copper and vacuum all need the standard's correction factors, which the fit does not include. The plane correction here is a single factor of two; the standard's charts show it varying with distance and plane copper weight.
• Neither is a thermal model of your board. A trace under a hot regulator, inside an enclosure, or next to a dozen other current-carrying traces will run hotter than any chart. Parallel traces do not get independent heating budgets.
• DC only. Skin depth in copper is about 66 µm / √f(MHz), so above a few MHz current crowds to the surface and resistance climbs. For switching-converter outputs the ripple current matters too.
• Short pulses. The charts assume steady state. For fault or inrush currents lasting under a second use a fusing estimate (Onderdonk) instead; the trace can carry far more briefly.
• When it matters — over about 5 A, or anywhere a 20 °C rise would be a problem — use the actual IPC-2152 charts with their correction factors, or a thermal simulation, and then measure it with a thermocouple on the first article.

Common mistakes

• Using the 2221 internal curve because "it's safe". It costs you double the copper for no measured reason and pushes people toward 2 oz inner layers they do not need.
• Trusting the fab's finished copper weight. "1 oz" outer layers start as ½ oz foil and get plated up; the finished thickness is anywhere from 30 to 45 µm depending on the shop. Inner layers are what they say. Ask for the spec, or design to the thinner figure.
• Forgetting the resistance. A 3 A trace at 10 °C rise is fine thermally, but 91 mV of drop on a 3.3 V rail is 2.8 %. Long return paths through a cut-up ground fill count double.
• Sizing the trace and starving the via. A 1.65 mm 1 oz trace has 0.057 mm² of copper; a single 0.3 mm via with 25 µm plating has 0.026 mm², about half of that. Use the via calculator and add vias.
• Picking a 10 °C rise by habit. The rise adds to ambient. 10 °C on a board already at 70 °C inside a box is a different situation from 10 °C on a bench, and FR-4's Tg is not the limit — the solder joints and electrolytic caps next to the trace are.
• Necking down at pads and through solder mask openings. The narrowest point sets the temperature; a thermal relief spoke is often the real bottleneck.

Further reading

• IPC-2152— the standard itself, with the full chart set and correction factors. Worth buying if you design power boards.
• IPC-2221B— the generic design standard; Section 6.2 still carries the conductor-sizing charts.
• TI SNVA419, Thermal Design By Insight, Not Hindsight— how copper area, vias and airflow actually move board temperature, with measured numbers.
• Saturn PCB Design Toolkit— free Windows tool whose conductor calculator uses its own IPC-2152 curve fits including the plane-distance and board-thickness corrections.