Trace Width Selection: Current, Rise and Impedance

Trace width is one of the first decisions made in a layout and one of the least examined afterwards. Too narrow and the trace runs hot, drops voltage and may fail over time; too wide and routing space is wasted that dense areas cannot spare. The correct width comes from three requirements rather than from a single rule of thumb.

What Determines a Trace Width

The width has to satisfy the current the trace will carry, the voltage drop the circuit can tolerate and, where the signal is fast, the impedance the design requires. These three constraints are independent, and the widest of the three answers is the one that governs.

A trace that meets only the current requirement may be far too narrow for a controlled impedance interface, and a trace sized purely for impedance may be far too thin for a supply rail. Checking all three before routing avoids a redesign later.

Current Capacity and Temperature Rise

Copper has resistance, and current through it produces heat. The permitted temperature rise is a design choice: a ten degree rise is conservative for a signal trace, while a supply rail may be allowed more if the board has thermal relief and the ambient is controlled.

The relationship is not linear. Doubling the width roughly halves the resistance but reduces the temperature rise by less, because a wider trace also has less surface area per unit of copper to dissipate heat from. The current calculation accounts for this and should be run rather than estimated.

<img src="https://www.gopcba.com/wp-content/uploads/2026/09/231-1.jpg" alt="Copper thickness cross section used for trace width calculation” />

Copper Thickness

Copper weight is expressed in ounces per square foot, and one ounce corresponds to roughly thirty five microns of thickness. Doubling the weight halves the resistance of a trace of the same width, which is often cheaper than doubling the width.

Heavier copper has its own consequences. The minimum trace and space that a fabricator can hold becomes larger as the copper thickens, because the etching process has more material to remove, and the panel cost rises with the weight requested.

Outer and Inner Layers

An outer layer trace dissipates heat into the air, while an inner layer trace is surrounded by laminate that conducts heat away slowly. Internal traces therefore run hotter for the same current and must be widened accordingly, or the copper weight increased.

The reference plane also affects the behaviour. A trace running directly over a plane benefits from the copper beneath it, whereas one that crosses a split has no such benefit and can behave quite differently from the calculation.

Impedance Controlled Traces

Where the signal must present a defined impedance, the width is determined by the dielectric thickness, the dielectric constant and the copper thickness. There is no freedom to choose: the geometry that achieves fifty ohms in a given stackup is a specific width.

That width should be calculated from the stackup that will actually be built, not from a typical value. A change of prepreg thickness or dielectric constant moves the number, and the fabricator should be given latitude to adjust the width within a stated band to hold the target, as the trace planning rules describe.

Voltage Drop Over Distance

On a long run, resistance matters even when the temperature rise is acceptable. The voltage arriving at the load is reduced by the drop along the trace, and that loss may be the limiting factor on a supply that feeds a device at the far end of the board.

The calculation is straightforward: resistance per unit length, multiplied by the current and the length. Where the drop is significant, widening the trace or moving the regulator closer to the load is usually more effective than accepting the loss.

Power and Ground Distribution

Supply and return paths carry the same current and should be treated as a pair. A narrow ground return is just as much a problem as a narrow supply trace, and the power plane allocation should reflect the current each rail has to carry.

Where a plane is used instead of a trace, the question becomes one of plane integrity rather than width, but the same current and temperature limits apply. Distributing a high current rail across a narrow neck in a plane creates exactly the local heating that a trace would.

Manufacturing Minimums

Every fabricator has a minimum trace width and spacing, expressed in the fabrication notes along with the copper weight. Designing at the minimum is possible, but it reduces yield and increases cost, and there is rarely a reason to do it on a prototype.

The pad design interacts with the trace width as well, since the trace has to meet the pad without creating a neck that concentrates current. A short taper from a wide trace into a narrow pad is preferable to an abrupt step.

A Practical Selection Method

Start with the current requirement and calculate the width for the acceptable temperature rise at the copper weight in use. Then check the voltage drop over the length, and finally confirm that any impedance controlled net matches its calculated geometry.

Round up to a standard value, note the result in the layout constraints so the router applies it consistently, and review the list once more before release. Widths recorded as constraints are far more likely to be honoured than widths that exist only in the designer’s memory.

Fault Conditions and Fusing

Normal operation is not the only case to consider. A fault can push a large current through a trace for a short period, and the question is whether the trace survives long enough for a fuse or a protection device to act.

Very narrow traces can act as fuses themselves, which is sometimes deliberate and more often accidental. Where a trace is expected to survive a fault, the width should be chosen from the fault current and its duration rather than from the steady state value.

Wide Traces in Dense Areas

A wide trace consumes routing channels, and in a dense area it can make the difference between completing the layout and not. Two narrower traces in parallel are sometimes used, though they should be treated as a single electrical path with a shared thermal calculation.

Routing a supply as a short wide spine with narrower branches is usually the practical answer. The trunk carries the combined current, and each branch is sized for the load it actually feeds rather than for the total.

Recording Widths as Constraints

Widths belong in the rule set rather than in the designer’s head. Assigning nets to classes by current and impedance lets the routing tool apply the correct width automatically, and it makes the intent visible to anyone reviewing the file.

The class definitions also document the design. A net class named for its current rating tells the next engineer why the trace is that wide, which is information that a bare layout cannot convey once the original author has moved on.

FAQ

How wide should a power trace be? Wide enough to carry the current at the permitted temperature rise, which depends on the copper weight and whether the trace is internal or external. It should be calculated rather than estimated.

Does a wider trace always run cooler? It runs cooler for the same current, but the improvement per unit of width diminishes, because a wider trace also has proportionally less surface area to dissipate heat from.

Can I use the fabricator’s minimum width everywhere? It is buildable but unwise. Designing at the limit reduces yield, raises cost and leaves no margin for the process variation that occurs in normal production.