A lightning strike on a transmission line lasts microseconds. A power transformer is designed to run for forty years. The lightning impulse test is the bridge between those two facts — it proves, in the factory, that the transformer can survive the microsecond event so it can deliver the four decades of service.
The test applies a standardized 1.2/50 microsecond impulse — a voltage that rises to peak in just over a microsecond and decays to half value in fifty microseconds — at a level well above the transformer’s operating voltage. Then it does it again, several times, and compares the records. If the records match, the transformer passes. If they don’t, something inside changed during the test, and the transformer has a problem.
That sounds simple. The physics underneath it is not. A lightning impulse doesn’t distribute through a transformer winding the way operating voltage does. It concentrates stress in places operating voltage never touches. It excites oscillations that can drive internal voltages to nearly twice the applied value. And it does all this in a time so short that the magnetic core — the thing that makes a transformer a transformer — barely participates.
This article covers what’s actually happening during a lightning impulse test. The physics of how the impulse distributes through the winding, why transformers need this specific test, how the test detects faults, and a worked example tracing the numbers for a real transformer.
For the procedural reference — where LI fits in the IEC 60076-3 test sequence, which transformers need it as routine versus type versus special, the exact test step counts — see the clause-by-clause guide to IEC 60076-3. This article is about the why.
Table of Contents
Why Lightning Is a Special Problem
Operating voltage and lightning voltage stress a transformer winding in completely different ways. Understanding that difference is the whole foundation of why this test exists.
At 50 or 60 Hz, voltage distributes through a winding uniformly. Each turn carries roughly the same share of the total voltage — volts per turn is constant. This is the inductive distribution, set by the winding’s inductance, and it’s what the transformer is designed around. The insulation between turns, between layers, and from winding to ground is dimensioned for this even distribution.
A lightning impulse behaves nothing like this. The rise time is so fast — about 1.2 microseconds to peak — that the winding’s inductance can’t respond. Inductors oppose rapid current change; in the first instants of a fast impulse, the inductive paths are effectively open circuits. The voltage doesn’t see the inductance at all.
What it sees instead is the winding’s capacitance. Every turn has capacitance to the next turn (series capacitance) and capacitance to ground and the core (shunt capacitance). In the first instant of a lightning impulse, voltage distributes according to this capacitance network, not the inductance.
And the capacitive distribution is highly non-uniform. The first few turns at the line end — the turns closest to where the impulse enters — take a hugely disproportionate share of the voltage. Instead of a smooth voltage gradient down the winding, you get a steep drop concentrated in the entry turns.
This is the central problem. The insulation between the first few turns, dimensioned for the even operating-voltage distribution, suddenly sees a stress many times higher than its design value. If the transformer wasn’t designed and built to handle this, the inter-turn insulation at the line end fails.
The Distribution Coefficient
The degree of non-uniformity has a name and a formula. The initial distribution coefficient, usually written α (alpha):
α = √(Cg / Cs)
where Cg is the total shunt (ground) capacitance of the winding and Cs is the total series capacitance.
A small α means a nearly uniform distribution — the impulse stress spreads evenly. A large α means a steep, concentrated distribution — the entry turns take most of the stress.
For a plain continuous-disc winding, the series capacitance Cs is small (adjacent turns are connected end to end, with little capacitive coupling between them), so α is large — often in the range of 5 to 15. That’s a steeply non-uniform distribution that concentrates severe stress at the line end.
Transformer designers fight this with winding construction. The most common technique is interleaving — physically arranging the turns so that turns far apart electrically are placed close together physically, dramatically increasing the series capacitance Cs. Higher Cs means lower α, which means a more uniform impulse distribution and lower stress on the entry turns.
Interleaved windings are more work to build (more complex assembly, often manual welding), but they make the transformer survivable under lightning. Some designs use electrostatic shields — floating conductors near the line end that improve the capacitive coupling — to achieve the same goal.
The lightning impulse test is, in part, a verification that this design work was done correctly. A transformer with insufficient series capacitance, or with a manufacturing error in its interleaving or shielding, will show up under the impulse stress that operating voltage never reveals.
The Oscillation Problem
The non-uniform initial distribution is only the first half of the story. The second half is what happens next.
The capacitive distribution that exists in the first microsecond is not the distribution the winding “wants” to settle into. As the impulse continues — over the following microseconds — the inductance starts to participate. The winding tries to redistribute the voltage from its initial capacitive pattern toward the final inductive pattern.
But it doesn’t move smoothly from one to the other. The winding is a network of inductances and capacitances — it’s an oscillator. When the voltage tries to shift from the initial to the final distribution, it overshoots, swings back, overshoots again. Energy bounces between the capacitances and inductances until losses damp it out.
During these oscillations, the voltage at points along the winding can swing well beyond both the initial and final values. In severe cases, the voltage to ground at internal points in the winding can approach twice the applied impulse voltage. A point in the middle of the winding, which would never see more than a fraction of the operating voltage, can momentarily see close to twice the full impulse level during these oscillations.
This is why the test matters for the whole winding, not just the entry turns. The entry turns face the initial-distribution stress. The interior of the winding faces the oscillation stress. A transformer has to survive both.
The designer’s job is to control these oscillations — through winding geometry, through the placement of shields, through the matching of the initial and final distributions so the redistribution swing is minimized. The test’s job is to prove the designer got it right.
Why the Core Barely Matters
One counterintuitive point that trips people up: during a lightning impulse, the magnetic core is almost irrelevant.
The core is what makes a transformer transform — it carries the flux that couples primary to secondary. At 50 Hz, the core is everything. But the core responds to flux, and flux builds through current flowing in the winding, and current can’t change instantaneously through the winding’s inductance. In the microsecond timescale of a lightning impulse, the core’s magnetic behavior simply doesn’t have time to develop.
This is why analysis of lightning impulse distribution treats the transformer as a network of capacitances and inductances — an electrostatic and electromagnetic problem — and largely ignores the core’s nonlinear magnetic properties. The high-frequency oscillations during the first microseconds are dominated by the winding’s own L and C, with the core contributing little.
It’s a useful mental model: at 50 Hz, think flux and core. At lightning speed, think capacitor network and traveling waves. They’re almost different machines.
What the Standard Impulse Looks Like
IEC 60076-3 specifies the standard lightning impulse waveshape per IEC 60060-1: a front time of 1.2 microseconds and a time to half value of 50 microseconds, written as 1.2/50 µs. The tolerances are ±30% on front time and ±20% on time to half value.
The front time is the steep rise — defined through the 30% and 90% points of the rising edge, extrapolated. The 1.2 microsecond figure represents the fast rise of a real lightning surge after it has traveled down a transmission line and reached the transformer terminal. The time to half value, 50 microseconds, represents the decay of the surge.
Why these specific numbers? They’re a standardized representation of real lightning surges as they appear at transformer terminals. Actual lightning is more variable, but decades of field measurement converged on 1.2/50 as a representative, repeatable test waveform that captures the essential stress.
For liquid-immersed transformers, the impulse is normally applied with negative polarity. The reason is practical: negative polarity reduces the risk of flashover external to the transformer (on the bushings and test connections), which would interrupt the test without telling you anything about the transformer’s internal insulation. Positive polarity can be specified, but negative is the default.
How the Test Detects a Fault
Here’s the part that makes the test work, and it’s elegant.
The test doesn’t measure insulation strength directly. It can’t — you can’t put a probe inside the winding. Instead, it relies on a comparison principle: apply the impulse at a reduced level, record the response, then apply it at full level and compare. If the transformer’s internal insulation is intact, the response scales linearly — the full-level records look identical to the reduced-level records, just bigger. If something breaks down internally during the full-level impulse, the response changes shape, and the comparison reveals it.
The standard test sequence for a transformer without internal non-linear elements:
- One reduced-amplitude reference impulse (50–75% of full test voltage)
- Three impulses at full test voltage
The reference impulse is the “fingerprint.” At 50–75% of test voltage, it’s high enough to capture the transformer’s true response but (presumably) low enough not to cause any breakdown. The three full-voltage impulses are then compared against this fingerprint and against each other.
What’s recorded? Two things at minimum:
The applied voltage. Measured per IEC 60060-2. This is the impulse itself.
At least one additional channel — almost always the neutral current. A low-value resistor (a current-measuring shunt) is placed between the winding neutral and earth. The impulse current flowing through the winding to earth is recorded as a voltage across this shunt.
The neutral current oscillogram is where faults reveal themselves. The current waveform is a sensitive signature of what’s happening inside the winding. It contains a high-frequency component (from the capacitive network and the oscillations), a lower-frequency disturbance, and the effects of wave reflections from the grounded end of the winding.
When the transformer is healthy, the neutral current waveform at reduced voltage and at full voltage are identical in shape — scaled by the voltage ratio. When a fault occurs — an arc between turns, a breakdown from turn to ground — the waveform changes. The classic signatures:
- Inter-turn or turn-to-ground arcing produces a train of high-frequency pulses, similar to those on the front of the impulse, appearing where they shouldn’t be. The waveshape changes.
- A local partial discharge (no full breakdown) produces high-frequency oscillations superimposed on the trace without a change in the overall waveshape.
- A major insulation failure can cause partial or complete collapse of the applied voltage wave itself — the impulse generator’s energy dumps into the fault.
The voltage oscillogram alone is a weak detector — it only catches faults severe enough to collapse the voltage, missing faults on less than about 5% of the winding. The neutral current method is far more sensitive, which is why it’s the standard fault-detection channel.
The acceptance criterion in IEC 60076-3 is, in essence: no significant differences between the records at the reduced reference level and the full test level, beyond what can be explained by the small voltage variation between them. Significant differences mean something changed inside — a fault. The transformer fails.
The Chopped Wave: A Harder Test
The full-wave lightning impulse (LI) tests the winding against a complete surge. The chopped-wave impulse (LIC) tests it against something nastier.
In service, a lightning surge entering a substation often gets “chopped” — a protective gap flashes over, or air insulation breaks down, suddenly clamping the voltage to near zero. This produces an extremely fast voltage collapse on the tail of the impulse, with a rate of change far steeper than even the impulse front.
That fast collapse is brutal on inter-turn insulation. The sudden change excites high-frequency oscillations and concentrates stress, much like the impulse front but in some ways worse. So the chopped-wave test deliberately reproduces it.
The chopped impulse is a full impulse that’s truncated on the tail, typically at 2 to 6 microseconds, by a chopping gap. The peak voltage of the chopped impulse is higher than the full wave — in IEC 60076-3, the LIC level is 10% above the LI level. The transformer has to survive a higher peak and the violent chop.
A representative LIC sequence for a transformer without non-linear elements:
- One reduced full-wave reference impulse
- Two full-wave impulses at 100%
- One reduced chopped-wave impulse
- Two full-level chopped-wave impulses
- Two more full-wave impulses at 100%
The full waves bracket the chopped waves so that any damage caused by the chopped impulses shows up in a comparison of the full-wave records taken before and after. If the final full-wave records differ from the initial ones, the chopped impulses damaged something.
The chopped wave is recorded on a faster time base because the action happens in a shorter window — the chop and its immediate aftermath need fine time resolution to evaluate.
For transformers at the highest voltages (Um above 170 kV), the chopped-wave test is a routine test, and the standard’s logic absorbs the plain LI test into the LIC sequence — you do the full-and-chopped sequence rather than a separate full-wave-only test. At lower voltages, LIC is a special test, done only when specified.
The Neutral Impulse Test (LIN)
The neutral terminal of a transformer normally sees very little voltage — it’s usually at or near ground. So why test it with an impulse?
Because of how the neutral is earthed in service. If the neutral isn’t directly grounded — if it’s earthed through an impedance, or via an arc-suppression device, or left to float under some conditions — then surges traveling down the winding can produce significant voltage at the neutral. The neutral insulation, which is typically graded down to a lower level than the line end, needs to be verified.
The LIN test applies impulses directly to the neutral terminal with the other terminals earthed. One difference from the line-terminal test: the front time can be much longer — up to 13 microseconds, versus 1.2 for the line terminal. This reflects how surges actually reach the neutral in service: they travel the length of the winding first, which slows and stretches the wavefront. Testing the neutral with a 1.2 µs front would be unrealistically severe.
LIN is always a special test in IEC 60076-3 — never routine. It’s done only when the purchaser specifies it, and only makes sense when the neutral isn’t directly earthed in service.
A Worked Example: 145 kV Power Transformer
To ground the numbers, take a typical sub-transmission transformer.
Transformer parameters:
- Three-phase, 90 MVA
- HV winding: Ur = 132 kV, Um = 145 kV
- Star connected, neutral brought out and directly earthed in service
- LI is a routine test (Um in the 72.5–170 kV range)
From IEC 60076-3:2013 Table 2, the standard rated lightning impulse withstand voltage for Um = 145 kV is, depending on the row chosen, 550 kV or 650 kV. Take the 650 kV level for this example (common for transmission-connected units).
So:
- LI test level = 650 kV peak
- LIC test level = 650 × 1.1 = 715 kV peak (if chopped-wave specified; it’s a special test at this Um)
- Reference impulse level = 50–75% of 650 = 325 to 488 kV
Compare to the winding’s normal duty:
- Operating voltage phase-to-earth = Ur/√3 = 132/√3 = 76 kV RMS, or about 108 kV peak
So the lightning impulse test stresses the line terminal to 650 kV — about six times the peak operating voltage to earth. That factor reflects the reality that a lightning surge at the terminal vastly exceeds operating voltage; the transformer must survive it.
Now consider what happens inside. With a steep capacitive distribution (say α = 8 for a moderately interleaved winding), the first portion of the winding — perhaps the first 10% of turns at the line end — might absorb 40% or more of that 650 kV in the initial microsecond. That’s 260 kV concentrated across a small fraction of the winding that, under operating conditions, would carry only about 11 kV (10% of 108 kV peak). The inter-turn insulation at the line end has to withstand more than 20 times its operating stress for that microsecond.
And during the subsequent oscillations, an internal point in the winding might briefly see voltages approaching 1.8 to 2 times the applied — over a megavolt of momentary internal stress on a transformer whose nameplate says 132 kV.
These numbers are why the test exists, and why winding design (interleaving, shielding, graded insulation) is such a central part of transformer engineering. The lightning impulse test is the proof that the design and the manufacturing actually deliver the withstand the nameplate promises.
The test itself: apply one reference impulse at, say, 70% (455 kV), record the voltage and neutral current. Apply three impulses at 650 kV, record each. Overlay the neutral current traces. If all four traces (scaled for voltage) are identical in shape, the transformer passes. If the full-voltage traces show high-frequency pulse trains or waveshape changes that the reference doesn’t, there’s an internal fault, and the transformer fails.
What Failure Looks Like and What Comes Next
Unlike the IVPD test — which the standard treats as essentially non-destructive and which triggers investigation rather than rejection — a lightning impulse failure is usually a real breakdown. Something arced or broke down inside the winding. That’s a more serious event.
If the transformer fails the LI test, IEC 60076-3 (clause 15) requires the complete test sequence to be repeated at full level after repair. Parts clearly not involved in the failure may be excluded from retest at the purchaser’s discretion, but the burden is on establishing that they weren’t affected — internal transients from the breakdown can damage insulation far from the actual fault site.
If the failure was in a bushing (not the transformer itself) and the purchaser is satisfied the transformer wasn’t affected, the bushing can be replaced and testing continued. If the failure was external — a flashover on a test connection rather than inside the transformer — the particular impulse is simply repeated; an external flashover tells you nothing about the transformer and isn’t a failure of the unit.
Locating the fault after a failed impulse test is its own discipline. The neutral current trace gives a first indication of severity and rough character. Comparison between terminals can localize it. Sometimes the failure is audible — a crack from inside the tank — or visible as bubbles in the oil. But pinpointing an internal impulse fault often means untanking the transformer, which is why a failed impulse test is an expensive, serious event in a factory.
What the Test Doesn’t Catch
The lightning impulse test is specific to fast-transient withstand. It’s blind to several other failure modes:
Slow insulation degradation. Thermal aging of paper, gradual moisture ingress, oil contamination — none of these show up in a microsecond impulse. The IVPD test and dissolved gas analysis cover those.
Power-frequency insulation-to-earth weakness. That’s the applied-voltage and induced-voltage tests’ job. A transformer can pass LI and fail AV, or vice versa — they stress different insulation in different ways.
Defects that don’t arc at test level. If an inter-turn insulation weakness is just strong enough to survive the test impulse but degraded relative to design, the impulse test passes. The weakness might progress in service.
Mechanical problems unrelated to insulation. Loose clamping, winding looseness, core problems — the impulse test isn’t designed for these, though severe cases sometimes show up indirectly.
This is why LI is one of several dielectric tests, not a standalone acceptance. It pairs with the chopped wave (faster transients), the switching impulse (slow transients), the applied and induced voltage tests (power-frequency stress), and the PD test (long-term degradation risk). Each catches what the others miss.
The Big Picture
A lightning impulse test compresses a transformer’s worst nightmare into a few microseconds and asks: can you survive this? It does so by exploiting the physics that makes lightning dangerous — the fast wavefront that distributes capacitively, concentrating stress where operating voltage never reaches, and exciting oscillations that drive internal voltages toward twice the applied value.
The test detects faults not by measuring insulation directly but by comparison: a healthy winding responds to a small impulse and a large impulse with the same waveform shape, while a faulted winding doesn’t. The neutral current oscillogram is the sensitive instrument that reveals the difference.
The standardized 1.2/50 waveform, the reduced-then-full comparison method, the negative polarity default, the chopped-wave escalation, the neutral-terminal variant — every detail in IEC 60076-3 and IEC 60060-1 traces back to making this comparison reliable and the stress representative of what real lightning does.
When you witness a lightning impulse test, watching the generator fire and the traces appear on the recorder, you’re watching a transformer prove it can take the hit it was built to survive. Six times operating voltage at the terminal, twenty times operating stress on the entry turns, a momentary internal megavolt — all in fifty microseconds, repeated and compared. If the traces match, the transformer is ready for forty years of the real thing.