PCB units and PPM converter
Mils, millimetres, ounces of copper, microns and parts per million — the conversions that fab drawings mix on the same sheet.
Length
Copper thickness
Drift over temperature
The conversions
Length
1 mil = 0.0254 mm = 25.4 µm 1 mm = 39.37 milA mil is one thousandth of an inch, not a millimetre — the name causes real confusion in mixed-language teams. Useful anchors worth memorising: 4 mil is about 0.1 mm and is a common fine-line minimum; 8 mil is roughly 0.2 mm and is standard capability; 100 mil is 2.54 mm, the classic header pitch.
Copper weight
1 oz/ft² = 34.79 µm = 1.37 milCopper foil is specified by the weight that would cover one square foot. Half-ounce is 17.4 µm, one ounce 34.8 µm, two ounces 69.6 µm. Outer layers gain plating on top of the base foil, so a 1 oz outer commonly finishes near 2 oz — which matters for trace width, for impedance, and for etch capability, since a heavier finish needs wider spacing to etch cleanly.
Parts per million
Δ = nominal · (ppm / 10⁶) · ΔTOne ppm is 0.0001%. The reason to work in ppm rather than percent is that coefficients are small and temperature ranges are large, and the product is what matters. The comparison that decides a design is drift against initial tolerance: a 0.1% resistor with 100 ppm/°C over an industrial range drifts six times its own tolerance, so paying for tighter tolerance without tighter tempco achieves nothing.
Practical notes
- Etch compensation. Fabricators adjust artwork so the etched result matches your intent. Heavier copper needs more compensation and holds tolerance less well.
- Impedance is thickness-sensitive. A stack-up quoted in ounces has real tolerance in microns, and controlled impedance depends on the micron figure.
- Tempco tracking beats tempco. In a divider, what matters is whether the two resistors drift together. A matched network with 100 ppm absolute but 5 ppm tracking is far better than two discrete 25 ppm parts.
- Self-heating is drift too. A resistor dissipating power runs hotter than ambient, so it drifts even in a temperature-controlled room.
Design guidance, not a manufacturing instruction. Verify every result against your fabricator's stack-up and the applicable standard before release. Closed-form models are approximations; the fab's field solver and process window are the authority.
Questions
What people ask about this
Why does copper weight get quoted in ounces?
It is a survival from the days when foil was specified by the weight of copper spread over one square foot. One ounce per square foot works out to about 34.79 µm of thickness, which is why the number is so awkward. Nothing in the process actually weighs anything any more — but the unit persists, and every stack-up drawing still uses it.
Is finished copper the same as the base foil?
No, and the difference bites people. Outer layers get plated during through-hole plating, so a board specified as 1 oz outer typically finishes at 1.5 to 2 oz. Inner layers are not plated and finish at their base weight. If you sized a trace assuming 1 oz on an outer layer, you have margin you did not plan; if you assumed 2 oz on an inner layer, you do not have the copper you think.
What does PPM/°C actually cost me?
Multiply the coefficient by the temperature excursion. A 100 ppm/°C resistor over a 60 °C swing drifts 0.6% — which is six times worse than its 0.1% initial tolerance. For anything precision, tempco usually dominates the error budget over temperature, and buying tighter initial tolerance without also buying tighter tempco is money wasted.
Why do fab drawings still use mils?
Because the industry grew up around imperial component pitches — 100 mil headers, 50 mil pads — and the tooling and vocabulary followed. Most fabs now accept either, but drawings routinely mix them, and a 0.2 mm trace next to an 8 mil clearance on the same sheet is normal. Converting in your head is how errors enter a stack-up.
Stack-up that has to be right first time?
We specify boards the way fabricators want to receive them, and we talk to the fabricator before the design is finished, not after.