A Buried PCB Trace Must Be Two and a Half Times Wider

Three times the current needs four and a half times the width

The relationship is not proportional. Going from one amp to three takes the required width from 0.300 to 1.367 millimetres, a factor of 4.55, because the heat generated grows with the square of the current while the cooling grows only with the width. Doubling the copper thickness is the cheaper lever where it is available: two-ounce copper halves the required width exactly, since the cross-section is what carries the current and it can be bought in thickness as easily as in width. On a crowded board that trade is often the difference between routing a power rail and redesigning the layout.

The temperature rise is a budget, not a limit

Allowing twenty degrees of rise instead of ten cuts the required width by a third, to 0.197 millimetres, which looks like free space until you ask where those degrees come from. The rise adds to whatever the board is already at: a trace running twenty degrees hot inside an enclosure that already sits at sixty is at eighty, and every component near it is warmer for the privilege. Ten degrees is the common default because it leaves room for the rest of the thermal budget, and choosing more is a decision about the whole enclosure rather than about one track.

Frequently asked questions

What does one ounce of copper mean?

It is a thickness expressed as a weight per area: one ounce of copper spread evenly over a square foot, which works out at about 35 micrometres. The unit survives from an era of imperial measurement and stuck because fabricators quote it, so a board specified in millimetres everywhere else still has its copper described in ounces. Two ounces is twice as thick, half an ounce is half. Since the current-carrying ability depends on cross-section, knowing the thickness is what lets a width be turned into an area, and quoting a width without the copper weight leaves the specification incomplete.

Is the width the only thing that matters?

No — length decides the voltage lost along the way, and a trace wide enough not to overheat can still be too thin for the drop it introduces. A long thin run to a sensitive analogue supply may need widening for the voltage rather than for the temperature, and the two constraints do not give the same answer. Vias are the other common surprise: a single via carries far less current than the trace it connects, so a wide power track crossing between layers usually needs several of them side by side rather than one.