Every power transformer above 72.5 kV gets an IVPD test before it ships. The test takes more than an hour, costs the manufacturer significant test bay time, and is the final dielectric test in the sequence. Why?
Because partial discharge is what kills transformer insulation slowly, and most other tests can’t see it.
A lightning impulse test stresses the insulation hard for microseconds and either holds or doesn’t. An applied voltage test stresses it at moderate AC for a minute and either holds or doesn’t. Both miss something important: a transformer can pass both of those tests with a manufacturing defect that will destroy the insulation over months or years of service. The IVPD test is designed to catch that defect before the transformer leaves the factory.
This article covers what’s actually happening during an IVPD test. The physics of partial discharge, what each acceptance criterion is protecting against, how the test detects different defect types, and a worked example showing how the numbers come together for a real transformer.
For the procedure itself — the 14-step sequence, where it fits in the IEC 60076-3 test plan, which transformers it applies to — read the clause-by-clause guide to IEC 60076-3. This article is about what the test is doing under the surface.
Table of Contents
What Partial Discharge Actually Is
The IEC definition: a localized electrical discharge that only partially bridges the insulation between conductors. The key word is partially.
In a transformer, the main insulation is solid (cellulose paper, pressboard) and liquid (mineral oil or ester). The solid materials have dielectric strengths in the range of 100 kV/mm or higher. The oil has roughly 20–30 kV/mm when clean. These are huge numbers.
But partial discharges happen at electrical stresses far below these breakdown values. Why?
Because of defects in the insulation that don’t extend all the way across. A small void inside a sheet of pressboard. A bubble in the oil. A region where the oil hasn’t fully impregnated the paper. A sharp metal edge that concentrates the field. A floating piece of metal between two insulating surfaces.
In each case, you’ve got a small region — usually filled with gas or air at low density — embedded in or adjacent to high-strength solid or liquid insulation. The whole assembly sees the operating voltage. The voltage divides across the layers according to their capacitance. And the small gas-filled region, having much lower permittivity than the solid around it, sees a disproportionately high share of the field.
When the field across that gas-filled region exceeds Paschen breakdown for the gas at that pressure, the gas ionizes. A tiny current flows. The voltage across the void collapses. The discharge extinguishes. Then the field rebuilds and the cycle repeats — often hundreds or thousands of times per second on AC.
The bulk insulation has not failed. The transformer is still operating. But each discharge releases energy into the surrounding insulation, breaking chemical bonds, eroding material, and producing gases that degrade the oil. Given enough time — and “enough time” might be a year, might be ten — the discharges erode a path large enough to cause complete breakdown.
This is why partial discharge matters. It’s the slow-failure mode that distinguishes a good transformer from one that will fail prematurely in service.
The Three Types of PD That Matter in Power Transformers
Not all partial discharge is the same. The IEC 60076-3 test is designed around what experience has shown to be the dangerous types for oil-filled power transformers.
Internal (void) discharge. Gas-filled voids inside solid insulation. These come from incomplete impregnation during manufacturing (the oil never fully penetrated the paper), from thermal cycling (paper shrinks differently from copper, creating gaps), from mechanical stress (delamination of pressboard layers), or from gas evolution during service (decomposition products forming bubbles). Void discharges are the most destructive type because they happen inside the insulation where they can’t be seen and slowly carve a path to ground.
Surface discharge / tracking. Discharges along the surface of insulation rather than inside it. Caused by contamination (moisture, particulates, products of oil degradation) creating partially conductive paths. These show up at oil-paper interfaces, around bushings, on the surface of pressboard barriers. Repeated surface discharges leave carbonized tracks — once tracking starts, it accelerates.
Corona / floating-particle discharge. Discharges from a metal point or a floating piece of metal in an inhomogeneous field. A burr left on a winding conductor. A loose particle in the oil. A piece of swarf in a bushing connection. These produce high apparent charge magnitudes because the metal acts as a small floating electrode that charges and discharges repeatedly.
The IVPD test is sensitive to all three. The acceptance criteria are calibrated against decades of service experience showing what PD levels and patterns correlate with failures in oil-filled transformers.
What the Test Is Actually Measuring
When a discharge happens inside the transformer, charge moves locally — at the void or the surface or the floating particle. That local charge movement induces a charge movement at the transformer terminals. Not the same magnitude as the actual discharge — what gets measured at the terminals is the “apparent charge,” typically much smaller than the real charge at the discharge site, because the coupling is capacitive.
IEC 60270 defines apparent charge in pC (picocoulombs, 10⁻¹² coulombs) and specifies how it’s measured. A coupling capacitor on a bushing test tap or in parallel with the test specimen captures the high-frequency current pulse caused by the discharge. The pulse is processed through a measuring impedance to give a charge value.
The numbers are tiny by everyday standards. 100 pC is 10⁻¹⁰ coulombs. A coulomb is roughly 6 × 10¹⁸ electrons; 100 pC is about 6 × 10⁸ electrons. The current pulses last nanoseconds to microseconds. PD measurement is, in its details, sensitive enough that environmental noise — radio transmitters, power electronics nearby, even fluorescent lights — can drown out the signal. This is why the test acceptance criterion includes a background PD check at the start and end (background must not exceed 50 pC for the test to be valid).
The apparent charge value isn’t the discharge energy or even the discharge charge directly. It’s a calibrated number that correlates with how big a discharge is happening. Operators interpret it based on experience with their transformer designs and against the IEC 60076-3 limits.
A Worked Example: 245 kV Power Transformer
To make the numbers concrete, let’s walk through the IVPD test for a typical large power transformer.
Transformer parameters:
- Three-phase, 200 MVA
- HV winding: Ur = 230 kV (rated voltage), Um = 245 kV (highest equipment voltage)
- LV winding: Ur = 22 kV, Um = 24 kV
- Star-connected HV with non-uniform insulation (graded toward neutral)
- Routine IVPD test required (Um > 72.5 kV)
The IVPD test for the HV winding uses the formulas in IEC 60076-3:2013 clause 7.3.2 (since Um is in the 72.5–170 kV range — wait, 245 kV is in the > 170 kV range, so clause 7.3.3 applies). Let me recompute.
For Um > 170 kV, the routine IVPD test uses:
- Enhancement voltage: (1.8 × Ur)/√3
- One-hour PD measurement voltage: (1.58 × Ur)/√3
With Ur = 230 kV:
- Enhancement voltage = (1.8 × 230)/√3 = 414/1.732 = 239 kV phase-to-earth
- One-hour PD measurement voltage = (1.58 × 230)/√3 = 363.4/1.732 = 210 kV phase-to-earth
- 1.2×Ur/√3 measurement point = (1.2 × 230)/√3 = 276/1.732 = 159 kV phase-to-earth
- 0.4×Ur/√3 background measurement point = (0.4 × 230)/√3 = 92/1.732 = 53 kV phase-to-earth
For reference, the normal operating voltage of the HV winding phase-to-earth is Ur/√3 = 230/1.732 = 133 kV. So:
- Background measurement is at 0.4× normal (53 kV)
- 1.2× measurement is just above normal operating (159 kV vs 133 kV nominal)
- One-hour PD voltage is 1.58× normal operating (210 kV)
- Enhancement voltage is 1.8× normal operating (239 kV)
What this means physically:
At 53 kV (background). Well below operating voltage. No PD should be present from the transformer itself. Anything measured here is environmental noise or measurement-system PD. The 50 pC ceiling at this point is the test validity check — if background is higher, you don’t have a reliable instrument and the whole test is invalid.
At 159 kV (just above operating). This is the post-test check point. After the transformer has been stressed at 1.8× operating, we drop back to 1.2× operating to see if any PD activity was triggered that persists at near-normal conditions. The 100 pC ceiling at this point is the most physically meaningful number in the test — it’s asking “can this transformer operate without significant PD at slightly above its design operating voltage?”
At 210 kV (one-hour PD voltage, 1.58× operating). This is the stress level for the one-hour test. We want to confirm that the transformer can sit at this elevated voltage for an hour without PD growing. The 250 pC ceiling at this point allows for some PD activity (no real-world transformer is completely PD-free at 1.58× operating) but bounds it.
At 239 kV (enhancement, 1.8× operating). This is the highest voltage. The transformer sees it for 60 seconds (Um ≤ 800 kV duration). This isn’t a measurement point — it’s a stress point. The purpose is to provoke any latent defect into discharging so the subsequent one-hour hold can detect ongoing activity.
The test sequence steps the voltage up through all these levels, holds at the enhancement voltage, drops to the one-hour measurement voltage and holds there for the full hour, then steps back down through the measurement points.
The Four Acceptance Criteria, Unpacked
IEC 60076-3:2013 clause 11.3.5 gives five pass criteria (a-e). Criterion (a) is “no voltage collapse” — obvious, that’s a fail by breakdown. The remaining four are the substantive PD criteria.
Criterion b: No PD level during the one hour exceeds 250 pC
This is the ceiling. During the entire one-hour hold at 1.58× operating voltage, no single measurement can be above 250 pC.
What’s the physics? Empirical correlation. Studies of failed transformers and successful service experience converged on this number. PD activity at the 1.58×Ur/√3 stress level above 250 pC, sustained, correlates with failure mode patterns that show up in service.
A single brief burst above 250 pC may be disregarded (clause 11.3.4 explicitly allows this). The criterion is about sustained activity at the highest level.
If a transformer shows, say, 180 pC steady throughout the hour, it passes this criterion. If it shows 220 pC at minute 1, 240 pC at minute 30, 260 pC at minute 45 — that’s a fail unless the 260 pC is a momentary burst.
Criterion c: No rising trend, and no sudden sustained increase in the last 20 minutes
This is the time-evolution criterion. PD levels should be stable or decreasing across the hour, not increasing.
The physics here is critical. PD activity that starts low and grows over time is the signature of a developing defect. A void that’s slowly enlarging. A surface track that’s progressing. A floating particle that’s eroding a position from which it discharges. These are the failure modes the test is most worried about.
A stable PD pattern, even at moderately elevated levels, suggests a defect that has reached a stable state and isn’t propagating fast. That’s still a defect, but it might be tolerable. A rising pattern is a defect that’s actively eating through the insulation.
The “last 20 minutes” qualifier addresses the worst case: PD that stays stable for 40 minutes and suddenly jumps in the final 20. That’s worse than gradual rise across the full hour — it means something just changed and is likely to keep changing.
Criterion d: PD levels during the hour don’t increase by more than 50 pC
This is a quantitative version of criterion c. Even if the trend looks flat by eye, the difference between the highest and lowest measurement during the hour can’t exceed 50 pC.
If a transformer shows 100 pC for 30 minutes, then 145 pC for 30 minutes — passes by 5 pC. Same transformer shows 100 pC for 30 minutes, then 160 pC — fails by 10 pC.
The 50 pC tolerance acknowledges that real PD measurements have natural variation due to measurement noise, environmental conditions, and statistical fluctuations in the discharge process itself. The criterion bounds that variation tightly enough to catch real growth while allowing for measurement reality.
Criterion e: PD at 1.2×Ur/√3 after the hour doesn’t exceed 100 pC
This is the post-stress recovery check. After being stressed at 1.8× operating for 60 seconds and held at 1.58× operating for an hour, we drop back to 1.2× operating (just above normal) and check what PD looks like.
The physics: a transformer that’s been through a stress regime should “settle” when the stress reduces. PD that persists at near-operating voltage after the high-stress regime indicates a defect that’s been damaged or activated by the test and is now discharging at operating conditions too.
A transformer that passes (a), (b), (c), and (d) but fails (e) is telling you something specific: the high-voltage stress damaged something. That’s worse, in some ways, than failing the high-voltage criteria — it means the transformer’s PD-free operating range has shrunk during the test itself.
The hidden criterion: background PD validity
Not numbered in the criteria but mentioned in 11.3.5: background PD at start and end must not exceed 50 pC (100 pC for shunt reactors). This is the validity gate. If the background fails, the test isn’t valid — you can’t determine pass/fail at all. The test must be repeated with better noise control.
What Different Defects Look Like on a PD Measurement
The acceptance criteria are pass/fail. But during the test, the operator is watching the PD level and pattern, and the pattern often tells you what kind of defect you’re dealing with — even before the criteria fail.
A void discharge typically shows up as PD activity that scales with applied voltage in a characteristic way. Below the discharge inception voltage of the void, PD is essentially zero. Above inception, PD jumps to a substantial level and grows slowly with further voltage increase. The phase-resolved pattern shows discharges clustered around the voltage zero crossings on both half-cycles. Pulse magnitudes are moderate and relatively repeatable — the void breaks down at roughly the same voltage every cycle.
If a transformer shows a clear inception voltage and a stable phase-resolved pattern with moderate-magnitude pulses, it has a void discharge. The question is whether the void is large enough to drive PD above the 250/100/50 pC limits.
A surface discharge shows up differently. Inception is usually lower than for voids. The PD level grows more steeply with voltage. The phase-resolved pattern shows discharges spread more broadly across the cycle, often with asymmetry between the half-cycles. Pulse magnitudes are highly variable — you see both small and large pulses mixed.
Surface PD often shows a rising trend over time as the surface degrades — exactly what criterion (c) is set up to catch. A surface defect that triggers at the enhancement voltage may keep growing throughout the hour even at the lower 1.58× hold voltage.
A floating-particle discharge has a distinctive signature. Very high pulse magnitudes (the metal acts as a small electrode), with the particle’s position shifting under the field forces. Pulse magnitudes are erratic. Phase pattern is irregular. The discharges may be intermittent — bursts separated by quiet periods.
The standard’s note about disregarding “occasional bursts of high partial discharge level” exists partly because of this pattern. A floating particle that occasionally produces a 300 pC burst between long quiet periods may not be a failure — it’s a single particle that needs to be removed, not insulation that’s failing.
Corona at a sharp edge or burr looks like a void discharge but with very stable, repetitive pulses at fixed positions in the cycle. Inception is sharply defined. Magnitude is steady. The phase pattern is highly regular.
This is the easiest defect to identify and usually the easiest to fix — find the sharp edge or burr and smooth it.
What Failure Means and What Happens Next
The standard explicitly states: “as long as no breakdown occurs, and unless very high partial discharges are sustained for a long time, the test is regarded as non-destructive.”
This is important. Failing the IVPD test doesn’t mean the transformer is rejected. It means the test triggered an investigation. Annex A of IEC 60076-3:2013 walks through the investigation steps.
The investigation tries to answer several questions in order:
Is the PD real or environmental? Phase correlation, oscillographic monitoring, filter changes on the supply leads. If the indication isn’t actually correlated with the test cycle, it’s noise and the test can be re-run with better filtering.
Is the source internal or external to the transformer? Provisional electrostatic shielding outside the transformer, observation of external corona on test connections, sharp edges in the test setup. External sources can sometimes be eliminated.
If internal, where is it? Acoustic emission detection. UHF electromagnetic sensors. Correlation of PD readings between terminal pairs. These narrow down the location.
What’s the physical nature of the defect? Phase-resolved pattern analysis identifies void vs surface vs corona vs floating particle. Voltage variation studies — how does PD respond to small voltage changes — discriminate further.
Is it processing-related? Sometimes PD comes from incomplete drying or oil impregnation. Re-processing the transformer (additional vacuum, additional drying time) followed by a repeat test may resolve it without disassembly. The standard explicitly mentions this option.
Is the defect actually harmful? Limited variation with voltage, a floating-particle pattern with no time-development — these may be acceptable for service. The decision becomes an agreement between purchaser and manufacturer, often involving extended hold times or modified acceptance limits for the specific case.
Only if all of these investigations fail or confirm a serious defect does the transformer get untanked for inspection. The standard treats untanking as a “last resort” — a noteworthy phrasing that emphasizes the destructive cost of opening up a fully assembled and tested transformer.
Why the One-Hour Hold
Most people new to IVPD ask why the test is so long. Why one hour at the measurement voltage? Why not 10 minutes, or 5?
Two reasons.
First, time evolution. The most dangerous defects are the ones that grow. A void that’s stable for 5 minutes may start to grow at 20 minutes as local heating from the discharge softens nearby material. A surface defect may show stable PD until enough degradation products accumulate to lower the local breakdown strength further. Catching these growth patterns requires time.
Second, thermal equilibrium. Power transformers are massive. The temperature distribution under test conditions takes time to stabilize. PD characteristics are temperature-sensitive — gas in voids expands or contracts with temperature, oil viscosity changes, paper conducts slightly differently. An hour gives the transformer time to reach a stable thermal state during the test, so the PD measurement reflects what the transformer will actually do in steady service rather than what it does during a transient.
The 5-minute hold at the measurement voltage before the enhancement (step e in the procedure) is the warm-up. The 60 seconds at enhancement voltage is the stress trigger. The one hour at measurement voltage is the meaningful observation period.
Larger transformers (Um > 800 kV) get 300 seconds at enhancement voltage instead of 60. Bigger transformer, longer thermal time constants, more stress time needed to provoke defects.
What the Test Doesn’t Catch
IVPD is sensitive to several defect types but blind to others. Worth knowing what’s not in scope:
Defects below the PD inception voltage. If a void is small enough that it doesn’t discharge until well above 1.8× operating voltage, the IVPD test won’t see it. The transformer ships. The defect may grow in service over years until it begins to discharge at operating voltage.
Defects with no associated PD. Some failure modes — winding deformation from short circuits, gradual paper degradation from thermal aging, contamination of the oil — don’t produce PD until they’re far advanced. IVPD won’t catch these.
Defects masked by the dominant PD source. If there are two defects in a transformer and one is much louder than the other, the louder one dominates the measurement. The quieter defect may not be visible until the louder one is fixed.
Test-bay PD that’s actually structural. A loose connection in the test setup, corona on a test lead, PD in a coupling capacitor — these can mimic transformer PD. The standard’s investigation procedure tries to isolate these, but if they’re not isolated, the transformer can be wrongly failed.
This is why IVPD is one test in a suite, not a sole acceptance test. It’s paired with dissolved gas analysis (before and after testing, to catch any decomposition products produced during the test), with the LI/SI tests (which catch different defect types), with the AV/IVW tests (which catch insulation-to-earth weaknesses the IVPD’s induced voltage approach doesn’t stress fully).
The Big Picture
The IVPD test exists because power transformers are expensive, hard to repair in the field, and tend to fail in ways that build up slowly. The faster tests catch gross defects. The slow test catches subtle ones.
The acceptance criteria — 250 pC ceiling, 100 pC at 1.2×Ur/√3, no rising trend, 50 pC max increase — represent decades of accumulated service experience encoded as numbers. They’re not arbitrary. They’re the level at which transformer manufacturers and utilities, working together, have found the boundary between “this transformer will probably last its design life” and “this transformer will probably fail early.”
That boundary isn’t perfect. Some transformers pass IVPD and fail in service. Some fail IVPD, get investigated, get released, and serve their full life without incident. The criteria are conservative — biased toward false alarms rather than missed failures — because the cost of a missed failure is enormous.
If you’re witnessing an IVPD test, watching the PD trace on a Lemke or Omicron screen, you’re seeing the working face of this risk management. The numbers tell you whether the transformer has the kind of defect that history says will fail prematurely. The pattern tells you what kind of defect it is. The acceptance criteria tell you whether it’s bad enough to reject. And the investigation procedure in Annex A tells you what to do if the criteria fail.
It’s not a perfect test. But for catching the slow-failure defects that other dielectric tests miss, it’s the best test the industry has.