Transformer inrush
A Socratic walk-through of transformer inrush — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why can switching on an idle transformer trip a breaker that its full running load never comes close to troubling?
A transformer sits with nothing connected to its secondary, doing no work at all. Close the breaker and, for a handful of cycles, it draws a current far larger than it ever draws at full rated load — large enough that the breaker may trip. Then it settles down to drawing almost nothing.
The strange part is not the size of the current but that the worst moment for a transformer is the moment it is asked to do the least. And stranger still: energise the same transformer twice, in identical conditions, and you may get two quite different surges. Something is being remembered.
Reasoning it through
REASONING #Start with what actually opposes current in an unloaded transformer. It is not resistance — the winding's resistance is small by design. What holds the current down is that a rising current builds magnetic flux in the iron core, and rising flux generates a back-voltage in the winding that very nearly cancels the applied voltage. Almost no net voltage means almost no current. A transformer running normally draws a tiny magnetising current for exactly this reason.
So the question becomes: under what conditions does the flux fail to hold that current down?
Turn Faraday's law around. The applied voltage sets the rate of change of flux, so the flux at any moment is the running integral of the voltage since you started, plus whatever flux was already there. Write the supply as a sine wave and integrate it and you get a cosine, offset — and it is the offset that matters. Suppose you close the contacts exactly at a voltage zero crossing. The voltage then spends a full half-cycle pushing in one direction before it reverses, and the flux climbs the whole way. The integral of a half sine gives twice the amplitude the steady-state cycle ever reaches. The core is asked to carry roughly double its normal peak flux.
Now close instead at the voltage peak. From there the voltage reverses after only a quarter cycle, and the flux traces exactly the excursion it would have traced anyway. No offset at all. The instant of closing, an interval no operator controls, decides the entire outcome.
And the memory? When a transformer is switched off, the flux does not return to zero — the iron keeps some of it, the remanent flux, exactly as a magnet keeps its magnetisation. The integral starts from that value, not from nothing. Close at a zero crossing with the remanence pointing the same way as the first half-cycle and you add the two insults together.
Why is doubling the flux catastrophic rather than merely awkward? Because a transformer core is deliberately worked close to its saturation knee. Core steel is one of the most expensive things in the machine, and flux density is precisely what you buy with mass — so nobody specifies a design flux at half of saturation and pays for twice the iron. Doubling a flux that already sits well within a factor of two of the knee drives the core hard into saturation with certainty.
And a saturated core is, magnetically, close to no core at all. Above the knee the iron's incremental permeability falls toward that of air, so the magnetising inductance — the thing generating the back-voltage — largely disappears. The applied voltage is now opposed by little but the winding's resistance and leakage inductance: the impedances that limit a fault, not the ones that limit a load. The current follows accordingly.
The surge dies away because the winding's resistance dissipates the offset that produced it, so the flux excursion drifts back toward symmetry over cycles to seconds. Large transformers, being efficient, have low resistance and therefore decay slowly — efficiency and a long inrush are the same property seen twice.
The analogy
THE ANALOGY #Think of a swing pushed by someone with a fixed rhythm, where you choose the moment you let go of the seat. Let go at the top of the pusher's return stroke and the swing takes up the motion it would have had anyway. Let go just as a full push begins, and that push runs its whole length from a standing start, carrying the swing far higher than it ever normally goes — higher again if the seat was already displaced when you took hold of it. The pusher does nothing unusual; the starting instant decides everything.
a swing pushed too far merely goes higher, whereas a core pushed past saturation does not go further into anything — it stops responding altogether, and it is that collapse of response, not the excess flux itself, that lets the current run.
Clarifying the model
THE MODEL #The account rests on one load-bearing claim: the current is set by how far the flux integral is driven past the saturation knee, and therefore by the point on the voltage wave at closing together with the residual flux already in the iron. Notice what it does not involve: the load, since the secondary can be open, or any fault in the transformer.
That claim is sharply falsifiable. It predicts that closing consistently at the voltage peak, into a core whose remanence has been managed, should largely remove the surge — and point-on-wave switching equipment does exactly this, and works. The refuting observation would be an inrush indifferent to closing instant and to residual flux — and if that were found, the flux-integral account would have to be abandoned entirely, not patched.
The most useful refinement is about what the protection does with all this. A relay cannot simply be desensitised, or it would miss a real internal fault. It discriminates instead: a saturating core draws a grossly distorted, one-sided current rich in second harmonic, whereas fault current is far closer to sinusoidal. Second-harmonic restraint is the standard answer, and it is worth naming what that represents. The physics could have been fixed in the core, by working the iron further below saturation. It was fixed in the relay instead, because logic is cheap and silicon steel is not: the transformer will saturate on nearly every energisation for the rest of its life, by design, and everything downstream is specified to tolerate it.
A picture of it
THE PICTURE #How to readRead top to bottom as time over the first few cycles. The breaker acts once, at the top, and the unlucky detail is the instant it chooses. Follow the arrows down the winding and core lifelines for the causal chain: voltage becomes flux by integration, flux passes the knee, the core's opposition vanishes, and only then does current appear. The last exchanges are the deliberate engineering answer — the relay does not prevent any of it, it recognises it and stands down.
What became clearer
WHAT CLEARED #An idle transformer is dangerous to switch on precisely because it is idle: with no load in the picture, the only thing restraining current is the core's ability to generate back-voltage, and that ability is destroyed by exactly the flux excursion a badly timed closing produces. Flux is the integral of applied voltage from wherever the iron was left, so closing at a voltage zero into a core that remembers its last magnetisation asks for roughly twice the normal peak flux and gets saturation. The design does not prevent this. It works the iron near the knee because iron is expensive, accepts saturation on every energisation, and pushes the problem into a relay that can tell a distorted magnetising current from a fault.
Where to go next
ONWARD #- Why geomagnetically induced currents saturate a core through the same mechanism, but continuously.
Key terms
TERMS #| Term | What it means |
|---|---|
| Remanent flux | the magnetisation an iron core retains after being de-energised, from which the next energisation starts. |
| Saturation | the condition in which further magnetising current produces almost no further flux, so the core's inductance collapses. |
| Second-harmonic restraint | protection logic that recognises inrush by the distortion of its waveform and declines to trip. |
| Point-on-wave switching | controlled closing timed to the supply waveform, chosen to avoid a flux offset. |
Every term the collection defines is gathered in the glossary.