Trace fusing current
What it takes to open a trace, how long it survives a fault, and why you should not use one as a fuse.
Fusing current at this duration
—A
—
The physics
Onderdonk — the adiabatic case
I = A · √( log₁₀((T_melt − T_a)/(234 + T_a) + 1) / (33 · t) )
Where A is cross-section in circular mils and t is
duration in seconds. The assumption is that the fault is short enough that no
meaningful heat escapes into the laminate — everything goes into raising copper from
ambient to its melting point at 1083 °C. Since energy is what matters, the current
scales as 1/√t: ten times the duration means about a third of the
current will do the same damage.
Preece — the steady-state case
I = 80.4 · d^1.5 (d in mm, wire in free air)An empirical result from the 1880s, still quoted because it is close enough for round wire. It has no time term because it describes equilibrium — the current at which heat generated equals heat lost. For a PCB trace bonded to laminate it is pessimistic, since the board conducts heat away far better than air does, but it marks the other end of the regime.
The failure that is worse than fusing
Copper melts at 1083 °C. FR-4 laminate begins to decompose around 300 °C and chars well below the melting point of the trace above it. Long before a trace opens, the laminate underneath it is being destroyed — delaminating, outgassing, and in the worst case carbonising into a conductive path that maintains the fault after the copper has gone. This is why a trace makes a poor fuse: the interesting failure happens to the board, not to the conductor, and it is not repeatable.
What to design to instead
Size traces for ampacity at an acceptable temperature rise, not for fusing. A properly sized trace typically fuses at five to ten times its rated current, which is exactly the margin you want. The trace width calculator gives that number. Use this tool to answer a different question: whether an existing trace survives a known fault before upstream protection clears it.
Where this stops being accurate
- Onderdonk assumes a perfectly adiabatic pulse. Above roughly a second, real traces shed heat into the board and survive more than this predicts.
- No laminate damage model. The board can be destroyed at currents well below the fusing figure.
- Assumes uniform width and thickness. Real traces fail at their narrowest point — a neck at a via, a nick from routing, or an etch variation.
- Vias, plane transitions and solder joints in the path can fail long before the trace does.
- Copper foil tolerance is typically ±10%, and it applies directly to this answer.
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
Is a PCB trace a legitimate fuse?
Almost never, and the reason is repeatability. A real fuse has a specified breaking capacity, a known I²t and a controlled arc. A trace opens somewhere along its length, at an unpredictable current, possibly with an arc that carbonises the laminate and creates a conductive path — which is worse than not opening at all. Some low-power products do use trace fusing deliberately, but it needs testing to destruction across the tolerance range, not a calculation. Treat this tool as telling you when a trace will fail, not as a way to design a protection device.
Why does the answer depend so heavily on time?
Because fusing is about energy, not current. For pulses shorter than a few seconds the trace is adiabatic — it has no time to shed heat, so all the energy goes into raising its own temperature, and what matters is I²t. For longer durations it reaches thermal equilibrium with the board and the answer becomes an ampacity problem instead. The curve below crosses between those two regimes.
What is the difference between Onderdonk and Preece?
Onderdonk models the adiabatic case — a short pulse where the conductor heats without losing energy to its surroundings — and takes duration as an input. Preece is an older empirical formula for the steady-state fusing current of a wire in free air, with no time term. They answer different questions, and where they disagree tells you which regime you are in.
How much margin should a trace have over its normal current?
Fusing is the wrong limit to design against. A trace should be sized for its ampacity at an acceptable temperature rise, which is typically five to ten times below its fusing current. If your normal operating current is anywhere close to the fusing current, the trace is running far too hot long before it opens, and the laminate is degrading. Use the trace width calculator for the number you should actually build to.
Fault-current path that has to behave predictably?
Protection coordination, arc energy and PCB survivability are design decisions, not afterthoughts. We size the whole chain.