The Difference Between an Overload and a Short Circuit
Why this matters
Both events end with a tripped device and a customer saying the breaker keeps going. From there the two jobs share nothing: one sends you to the load and the other sends you into the walls. Techs who cannot tell them apart on arrival do the same thing to both, which is reset it, watch, and guess. The evidence that separates them is available before you take a single reading, and it is mostly free: how long it ran before it tripped, whether the trip was instant, whether anything is marked, and whether the timing depends on how much is running.
Do not reset it again
29 CFR 1910.334(b)(2) states plainly that after a circuit has been de-energized by a protective device, it may not be manually re-energized until it has been determined that the equipment and circuit can be safely energized. On a suspected short that is not a formality. Each reset closes an unfaulted device onto a fault, and the fault does not care that you are holding the handle.
If a device operated with an audible bang, or you find scorching, pitting, or carbon marking, treat the equipment as damaged: do the arc-flash risk assessment under NFPA 70E-2021, 130.5 before operating anything, keep the enclosure door closed where the equipment supports operating it that way, and stand out of the line of the door.
An arcing fault vaporizes metal and burns insulation, and the residue is an inhalation hazard, not a contact one. Ventilate the space, work from outside the plume, and use respiratory protection rather than gloves where fume or a sharp burnt smell persists. Do not clear an enclosure with compressed air. On older switchgear, be aware that some legacy hardware carries cadmium plating, and heating or abrading it releases cadmium oxide fume, which is controlled under 29 CFR 1910.1027 and requires respiratory protection.
Everything that follows in the walls is done de-energized, locked and tagged under 29 CFR 1910.333(b)(2), or 29 CFR 1926.417 on construction work, and proved dead with the live-dead-live sequence at NFPA 70E-2021, 120.5. The live readings described below sit inside the troubleshooting exception at 29 CFR 1910.333(a)(1), with instruments and leads rated for the system under 29 CFR 1910.334(c)(2).
The distinction is the path, not the number
An overload is too much current on the path it was supposed to take. The circuit is intact, the load is real, and every conductor is doing its job - there is just more of it than the design allowed. Magnitudes are modest, typically a fraction above rating up to a few times rating, and the event is sustained.
A short circuit is current on a path it was never supposed to take, between circuit conductors, at an impedance far lower than any load. Magnitudes are limited only by the source and the impedance of the loop, so hundreds or thousands of amps, and the event is over in a fraction of a cycle.
A fault to a grounded or groundable part is a third category with its own behavior and its own protective devices, covered in a sibling article on what a ground fault actually is.
That path difference is why the two leave completely different evidence, and the evidence is what you read on arrival.
The gate
Classify the event on four observations before any instrument comes out, and require all four to agree:
- Time to trip. Sustained running before the trip means a thermal element operated, which means overload. Instant means a magnetic element operated, which means a low-impedance fault.
- Dependence on how much is running. An overload's timing tracks the load: more running means a sooner trip. A short does not care what is running.
- Physical evidence at a point. Localized carbon, pitting, a blown-open conductor, a mark on an enclosure: a short. Nothing visible anywhere: overload.
- Measured load current. Above rating with the circuit running: overload. Normal or unmeasurable because it will not stay closed: short.
Where the four disagree, stop and read the exceptions section below rather than forcing the classification, because the disagreement is itself the finding.
Callback one: trips after half an hour
A 20 A branch circuit serving fixed electric heating. The customer reports it trips after roughly 20 to 40 minutes every time, resets normally, and runs fine again for another half hour. No smell, no marks anywhere, breaker warm to the touch on the outside of the enclosure.
Run the gate. Time to trip is long and consistent, so a thermal element. Timing tracks running time, and the customer confirms it trips sooner on the coldest days when everything runs longest. Nothing is marked anywhere in the circuit. Three of four say overload before any meter comes out.
The fourth observation confirms it. Clamped with the circuit running: 23.4 A on a 20 A circuit, which is 117 percent of rating. That number also explains the delay rather than just supporting the verdict. At 1.17 times rating, a thermal element's published curve sits in the tens of minutes, which is exactly the 20 to 40 minute window the customer described, and 23.4 A is nowhere near any magnetic element's pickup, so an instant trip was never physically possible on this circuit.
The cause was a second heating load added to the circuit by a previous trade. The repair is a circuit change, not a device change. Nothing in the wiring was faulty, and nothing in it needed replacing.
What would flip this. If the measured running current had come in at or below 20 A and the circuit still tripped on a delay, the load is not the problem and the suspect list becomes the device's own thermal element, the ambient temperature at the enclosure, or a resistive connection at the breaker's lug making heat that the thermal element reads as load. Those are covered in sibling articles on how a protective device decides to open and why a loose connection gets hot.
Callback two: trips the instant it closes
A 20 A branch circuit on a lighting and receptacle load. The customer heard a bang the first time. The breaker resets and re-trips immediately, with nothing switched on, and there is faint carbon marking visible at the edge of a junction box cover.
Run the same gate. Time to trip is instant, so a magnetic element. The trip does not depend on what is running - it happens with every load switched off, which by itself eliminates overload entirely, because an overload with no load connected is a contradiction. There is localized physical evidence at one point. Load current cannot be measured, because the circuit will not stay energized.
Four of four say short circuit, and the fourth one is the tell that closes it: an overload needs a load, and this circuit trips with nothing connected.
De-energized, locked and tagged, with the circuit conductors disconnected from the panel, a resistance measurement between the ungrounded conductor and the grounded conductor read essentially zero rather than the open reading a healthy circuit gives. That is a conductor-to-conductor path, which is the definition. Opening the marked box found the cable pinched under the cover screw, with the screw driven through both conductors' insulation.
Estimate what actually flowed. On a 120 V circuit with a loop impedance of roughly a quarter of an ohm from source to fault and back, current is 120 divided by 0.25, about 480 A. Treat that as an order-of-magnitude figure: real available fault current is set by the supply transformer and the impedance of everything between it and the fault, and the authoritative number comes from the utility and an engineering study, not from a field estimate.
Why one leaves a crater and the other leaves nothing
This is where the two events stop being "small" and "big" and start making sense.
Compare total heating energy, which is current squared times time. The overload in callback one: 23.4 squared is 547.6, times 1,800 seconds of running before the trip, giving about 985,600 in units of amps squared times seconds. The short in callback two: 480 squared is 230,400, times an illustrative 0.01 seconds, which is under one cycle at 60 Hz, giving about 2,300 in the same units.
The overload delivered more than 400 times the total heating energy of the short. And it left no mark anywhere, while the short burned a visible crater into a cable.
The reason is not magnitude, it is concentration. The overload's energy was spread along every foot of the circuit's conductors, developed over half an hour, with the surroundings carrying heat away the entire time, so the conductors reached a modest steady temperature and stopped there. The short's energy went into a contact patch smaller than a pinhead in under a hundredth of a second, faster than anything nearby could conduct it away, plus arc energy at the fault point that is separate from the conductor heating altogether. Same physics, two completely different rates.
Carry that as the mental model: damage follows power density and duration, not total energy. It also explains why a slow-developing overload is genuinely more dangerous to the customer than it looks, because it produces no evidence at all until insulation has been cooking for months.
Where the clean binary breaks
Three cases sit between the two, and all three are the ones that hurt people.
The high-impedance arcing fault. An unintended path with enough impedance that current lands between overload and short levels - too low to reach a magnetic element's pickup, and possibly below the thermal element's curve for a long time. It is an unintended path, so it is not an overload, and it does not behave like a short. It is the classic ignition source, and it is why arc-fault detection exists as a separate function that responds to the arcing signature in the waveform rather than to a current level.
The legitimate transient. A motor start draws several times full-load current for seconds, which is short-circuit territory by magnitude alone. It is neither fault. The distinguishing property is duration, and that is why the two devices on a motor circuit are sized so differently. A sibling article on locked-rotor conditions carries that reasoning.
The intermittent short. A conductor damaged but not yet through, arcing only when thermal expansion or vibration closes the gap. It produces instant trips with no consistent pattern, which fails observation 2 of the gate. When observations 1 and 3 say short and observation 2 says nothing consistent, believe 1 and 3: an intermittent short is still a short, and repeated resets while you look for the pattern are exactly what 29 CFR 1910.334(b)(2) is written against.
How to verify you got this right
For an overload verdict, the confirmation is a measured current, not a fixed device. Clamp the circuit under its full realistic load after the repair and confirm it now runs below rating with margin, and record the load conditions alongside the reading so next winter's tech can compare like with like. A circuit measured at 19.8 A on a 20 A device is not fixed, it is one added appliance from being back.
For a short verdict, the confirmation is that the fault path is gone and the mechanism that made it is gone. A resistance check between conductors with the circuit isolated should read open, not merely higher than before. Then answer the second question: what damaged the cable, and is the same condition present elsewhere in the run? A cover screw driven through a cable was driven by somebody's habit, and that habit was applied at every box on the same day.
References
- 29 CFR 1910.334(b)(2) - prohibition on manually re-energizing a circuit after a protective device has opened until it is determined safe; 29 CFR 1910.334(c)(2) for instrument ratings
- 29 CFR 1910.333(a)(1) and (b)(2) - the troubleshooting exception and safe work practices for electrical work; 29 CFR 1926.417 for the construction counterpart
- 29 CFR 1910.1027 - cadmium exposure control, where legacy plated hardware is heated or abraded
- NFPA 70E-2021, 120.5 (establishing an electrically safe work condition) and 130.5 (arc flash risk assessment)
- NFPA 70 (National Electrical Code), Article 240 for overcurrent protection - use the edition adopted in your jurisdiction
- See related: How a Protective Device Decides to Open; What a Ground Fault Actually Is; What a Locked Rotor Condition Actually Is