Why Inrush Current Trips Things That Should Hold

Why this matters

A breaker that trips on start and holds all day is not a breaker with a problem. It is a breaker being handed a current profile it was never selected for, and the usual field response - fit a bigger one - removes protection from a conductor that still has the same ampacity it had yesterday. The customer gets a fix that lasts until the conductor insulation gives up.

Inrush is not a current. It is a current and a duration, and every protective device is a curve on those two axes. Until you have both numbers, you are guessing which side of the curve you are on. This article is built around a worksheet with both numbers filled in, because the discipline of writing the fields down is most of the diagnosis.

What inrush actually is, by load family

Different loads produce inrush for entirely different reasons, and the reason determines the fix.

Across-the-line induction motors draw roughly 5 to 8 times full-load current at the instant of start, because a stationary rotor presents almost no back-EMF and the winding looks close to a short circuit until the shaft turns. That range covers standard induction motors started directly on the line, and specifically excludes electronically commutated motors and inverter-fed motors, whose drives limit starting current electronically and produce nothing like this profile. Nameplate locked-rotor amps is the number to start from. Duration is set by how fast the load accelerates, so it is a mechanical question as much as an electrical one.

Transformers draw magnetizing inrush that has nothing to do with the load on the secondary. It comes from core flux, and its magnitude depends on where on the voltage waveform the contacts happen to close and on the residual flux left from the last de-energization. It can reach a large multiple of rated current for a few cycles, and it is different every time the same transformer is switched.

Capacitive loads - switch-mode power supplies, LED drivers, electronic ballasts, drive input sections - charge an empty capacitor through nearly zero resistance. The peak is very high and very short, a fraction of a cycle, and it adds when many small units are switched together.

Three different mechanisms, three different-shaped events, one meter setting. This is why an inrush-capable clamp that records peak and lets you see the decay is worth carrying.

The worksheet, filled in

A rooftop condensing unit whose 20 A breaker trips on start, not every time, mostly on hot afternoons.

Capturing this requires an energized reading on a running circuit, so work under 29 CFR 1910.333(a)(1), which permits energized work only where de-energizing would introduce additional or increased hazards or is infeasible due to equipment design or operational limitations, with diagnostic testing as the qualifying case, using a clamp and leads rated CAT III or above at the voltage present. Do not force starts by pushing a contactor armature in by hand; you are placing your fingers in an arcing path and in a mechanism that closes hard. And limit how many starts you command: every start at locked-rotor current dumps a large amount of heat into the rotor and windings in a second or two, motor manufacturers publish a limit on starts per hour, and compressors carry anti-short-cycle requirements you can damage the machine by defeating.

Reading the filled worksheet

The peak is not the story. 94 A against a 20 A breaker is 4.7 times rating, and the magnetic instantaneous element of a standard thermal-magnetic breaker picks up much higher than that, commonly around ten times rating though the exact value is a manufacturer and breaker-type characteristic you read off their published curve rather than assume. So the peak alone was never going to trip it, and indeed it does not trip on a cool morning at essentially the same peak.

The story is the duration. The thermal element responds to accumulated heating, which tracks current squared multiplied by time. Same current, but 1.9 seconds instead of 0.4 seconds, is 1.9 / 0.4 = 4.75 times the heating energy delivered into that element on a single start. That is the entire difference between the start that holds and the start that trips.

So the diagnostic question stops being "is the breaker too small" and becomes "why is this motor taking almost five times as long to accelerate." The worksheet already contains the leading candidate: 198 V at the terminals during start. Starting torque on an induction motor varies with the square of applied voltage, so at 198 V on a 208 V nominal system, roughly 95 percent of nominal, available starting torque is about 0.95 squared = 90.5 percent of what it would be at nominal, and the motor takes correspondingly longer to reach speed. The other candidates are mechanical or refrigerant-side: a system that has not equalized before restart, a failing start component, a dragging bearing, or a load that has genuinely changed.

Why the same start trips in July and holds in January

Thermal-magnetic breakers are calibrated at a reference ambient, commonly 40 C for molded-case devices, and their thermal element responds sooner as the ambient around it rises, because the element starts closer to its trip temperature. Manufacturers publish an ambient-correction table for exactly this, and the correction is the only value worth using: rules of thumb here are worth less than the two-column table in the catalogue.

The worksheet says 52 C inside a closed rooftop enclosure. That is above the calibration ambient and it moves the whole curve, so the same 94 A for the same 0.4 s that passed in January sits closer to the trip boundary in July. Combine that with an acceleration time that is also longer in July because head pressure is higher, and the intermittency stops being mysterious. Two independent seasonal effects push the same event across the same boundary.

This is also why "I tested it three times and it was fine" proves nothing when the test was at nine in the morning. Test at the condition the complaint describes, or note in the ticket that you did not.

The fix hierarchy, and the one fix that is not a fix

Work down this list. The order is by how much of the actual problem each option removes.

1. Shorten the event. Correct whatever is extending acceleration: supply voltage at the load during start, refrigerant-side pressure not equalizing before restart, start components, mechanical drag. This is the only option that leaves the equipment healthier as well as the breaker happier, because a motor that accelerates slowly is heating its own rotor on every start.

2. Match the device curve to the load, without raising the protection level beyond what is permitted. NEC Article 430 handles motor circuits with two separate protections doing two separate jobs: branch-circuit short-circuit and ground-fault protection, sized as a percentage of full-load current with the allowed percentage differing by device type precisely because their curves differ, and running overload protection sized separately and much closer to full-load current. A dual-element, time-delay fuse tolerates a starting profile that a non-time-delay fuse of the same rating will not, which is why the permitted percentages are not the same. Selecting a device whose curve fits the measured profile is engineering; selecting a bigger number is not.

3. Reduce the peak, and understand the trade. A soft starter, a wye-delta or part-winding starter, or a drive reduces the current peak - and in doing so extends the acceleration time, because it is reducing starting torque to reduce current. You are trading one axis of the curve for the other. That is often the right trade, but it is a trade, and a soft starter fitted to a load that was already accelerating slowly can make the thermal picture worse rather than better.

4. Cool the enclosure. Ventilating or shading a panel that runs 12 C above the breaker's calibration ambient restores margin you are otherwise buying with hardware.

The one that is not a fix: fitting a larger breaker to stop the trips. The conductor's ampacity did not change, the motor's overload protection requirement did not change, and the thing that was warning you about a slow start is now silent. If the panel schedule shows a device larger than the conductor and the motor circuit rules permit, that is a finding to write up, not a precedent to follow.

The transformer case: why it is one start in ten

Magnetizing inrush deserves its own note because it produces the most confusing complaint in this whole area: a transformer or a transformer-fed control circuit that trips its primary protection occasionally, with no pattern, and passes every test you run afterwards.

Nothing is wrong with the transformer. The magnitude of magnetizing inrush depends on the instantaneous voltage at the moment the contacts close and on the residual magnetism left in the core from the last time it was switched off. Close at the worst combination and you get a large first-cycle current; close at the best and you get almost none. The contacts close wherever they close, so the same transformer produces a different inrush every time and occasionally produces one big enough to reach the magnetic element.

The response is to select primary protection with a curve that tolerates the worst-case first cycles rather than to chase a component fault, and to stop replacing transformers that test perfectly. The same reasoning applies to a lighting circuit that started tripping on switch-on after an electronic retrofit: running current went down substantially and the switch-on peak went up, because dozens of empty capacitors now charge simultaneously at whatever point on the wave the switch closes.

Acceptance criteria before you leave

Write down what "fixed" means before you test it, so the test can fail:

• Re-capture the inrush profile at the condition that produced the complaint, not at a convenient one. For this unit that means the hot afternoon, with the enclosure closed and the panel at its normal running temperature.
• The pass condition is a measured time above 2x FLA back in the range the healthy start showed. On this job the repair brought the trip-event profile from a 92 A peak with 1.9 s above 32 A down to a 93 A peak with 0.5 s above 32 A: peak essentially unchanged, duration cut by 1 minus (0.5 / 1.9) = 73.7 percent of the heating energy per start.
• Then require consecutive starts without a trip at that condition, counted and written on the ticket. One successful start proves nothing against a fault that was already intermittent.

References

• NEC Article 430 for motor branch-circuit short-circuit and ground-fault protection and separately sized running overload protection
• Manufacturer time-current curves and published ambient-correction data for molded-case circuit breakers, and manufacturer limits on motor starts per hour
• 29 CFR 1910.333(a)(1) for the energized-work justification governing live inrush capture
• See related: How a Motor Draws What It Draws; What a Transformer Actually Does; Single-Phase and Three-Phase in Practice