IEC 60270: What a Partial Discharge Reading Actually Measures

When a void inside the insulation of a dielectric film stack breaks down locally, the charge that moves is in the picocoulomb to nanocoulomb range and lasts tens of nanoseconds. You cannot see it. What the instrument sees is the current pulse that this local breakdown pushes through the specimen's own capacitance, and the integral of that current is what gets recorded. IEC 60270 calls it apparent charge for exactly this reason: it is the charge you would have to inject at the terminals to reproduce the same pulse height.

Two consequences follow, and they explain most of the arguments between suppliers and customers.

First, the reading scales with the specimen. A 3 m three-layer busbar may present 1.5–4 nF between conductor and earth, an 80 mm coupon a fraction of that. Inject the same void activity into both and the larger part shows a larger apparent charge at the terminals, because more of the pulse current is available to the measuring impedance. Comparisons across geometries are therefore only meaningful if the test circuit is described.

Second, the reading scales with the circuit. Coupling capacitor value, lead length, grounding, bandwidth and the position of the calibrator all move the number. A 10 pC result from a lab with a compact shielded setup and a 10 pC result from a lab that measured through two metres of unshielded lead are not the same claim, even when both printed "10 pC" on the report.

For the physical background of where discharges start in a busbar, see the partial discharge primer; this page is about the measurement standard that turns that physics into a number.

Before any PD measurement, a calibrator injects a known charge — 10 pC, 50 pC, 100 pC — into the test circuit, and the instrument's scale factor is set from that injection. The detail that decides whether the result is trustworthy is where the charge is injected.

The calibrator must be connected in parallel with the specimen, at the specimen terminals, with the same circuit and the same grounding that will be used for the test. A calibrator clipped to the instrument input, or injected across the coupling capacitor alone, proves only that the instrument can count; it says nothing about the path the real PD pulse takes through the specimen leads and the measuring impedance.

Two practical checks are worth demanding in any report:

• Calibration before and after the test. If the scale factor drifts during a 30-minute run, the second calibration exposes it. Drift usually points to a loose connection or a change in stray capacitance, and the whole run should be repeated rather than averaged.
• Calibrator accuracy and date. A calibrator whose last verification was four years ago, with a 5% tolerance, already spends half the margin of a 10 pC acceptance limit.

The classic error in busbar work is calibrating with the specimen disconnected. The moment the busbar is connected, its capacitance and the lead inductance change the transfer function, and the 50 pC injection becomes an unknown at the terminals. If the setup is rebuilt after calibration, the calibration is void. Our own quality process treats a rebuilt circuit as a new test, not a continuation.

A conventional narrowband PD circuit is simple: specimen, coupling capacitor, measuring impedance, amplifier, detector. IEC 60270 allows both narrowband (typically centred between 100 kHz and 1 MHz) and wideband detection, and the standard's job is to make the reading traceable rather than to prescribe one circuit.

Busbars break the comfortable assumptions in three ways.

Capacitance. A laminated stack with PET or DLF films is a distributed capacitor, and its value can be large compared with the coupling capacitor. If the coupling capacitor is not comfortably larger than the specimen, the pulse divides in a way that depends on geometry, and the calibration performed at one point no longer represents the whole specimen.

Rise time. Conductor-to-conductor spacing in a busbar is small, so pulses are short and their spectrum extends high. A detector set for a 100 kHz narrowband window may under-read the fastest events; a wideband setup sees them but also sees more of the ambient noise. The bandwidth, the gain and the timebase must therefore be in the report, not just the pC value.

Geometry of earth. The busbar's own conductors are part of the ground path once installed in an enclosure. Testing a bare stack on a bench, with an earth return that runs past the specimen, produces readings that mean something different from the same stack bolted to a copper bar assembly in its final orientation. If the application is an inverter DC link, ask for the low-inductance design context too, because the layout that reduces loop inductance also changes where the field concentrates.

Every PD number is a difference between signal and background, and the honest way to report it is to state the background. A common acceptance rule is that the background must be no more than half of the acceptance limit — with a 10 pC limit, background below 5 pC, and preferably below 2 pC. A report that shows 9 pC measured against 8 pC of background is not reporting a discharge; it is reporting an environment.

Discrimination matters as much as sensitivity. Phase-resolved patterns separate the usual suspects:

• Internal voids in the film stack repeat twice per cycle, are broadly stable in phase and grow with voltage in a predictable way.
• Corona from the high-voltage lead or bushing sits at a fixed phase, often asymmetric between half-cycles, and does not move when the specimen is re-oriented but does disappear when the lead is dressed differently.
• Surface activity at cut edges is erratic, humidity-sensitive, and often appears only above a threshold voltage that shifts from day to day.
• Switching and drive noise from nearby EV or inverter test benches is broadband and correlated with the other equipment's operating state, not with the applied voltage.

Non-conventional methods — acoustic, UHF, or optical detection — are useful cross-checks for locating a source, but they do not replace the calibrated pC reading when the acceptance criterion is written in picocoulombs.

IEC 60270 is a measurement standard. It defines the quantity, the calibration procedure and the reporting requirements. It deliberately does not set an acceptance limit, because a limit depends on the product, the voltage class and the consequence of failure.

The limit comes from somewhere else: a product standard, a customer specification, or an internal design rule. In low-voltage switchgear assemblies, verification practice and the division of responsibilities between type tests and routine tests are set out in IEC 61439. Clearances and creepage that determine how much field stress the insulation sees come from IEC 60664-1, which is why a PD discussion without those two numbers is incomplete.

In practice, laminated busbar specifications in industrial and e-mobility work often call for something like 10 pC at a stated multiple of the maximum system voltage, held for a stated time, with the criterion applied at the final step after a voltage ramp. The exact voltage, the hold time and the criterion have to be written down, because three suppliers quoting "10 pC" may be testing at three different voltages. A supplier reporting "PD free" without a voltage and a background level has reported nothing.

There is also a sequencing point that is easy to lose: a PD test performed before the hipot test can be invalidated by the hipot test itself, which may condition or damage a marginal void. Where both are required, the order and the conditioning history belong in the same document. We cover the acceptance side of the voltage tests in the hipot acceptance guide.

A pressed film stack is not a homogeneous dielectric. It is layers of film, adhesive, copper and air that was never fully expelled, and the discharge sites follow that structure.

Voids between film layers. Lamination pressure and temperature decide how much air remains at the interface. A void a few tens of micrometres thick between two PET layers sees the full field across the insulation and is the textbook source of internal discharge. This is a process variable, not a design variable: two suppliers with identical drawings can differ here.

Foil edges and triple points. Where the copper foil ends inside the insulation, the field concentrates at the edge, and the point where copper, film and any residual air meet is a triple point with enhanced stress. Burrs left by slitting move that concentration further.

Cut edges of the finished part. The cut face exposes the layer structure; if the cut is mechanical and the edge is smeared or delaminated, surface tracking and surface discharge become possible, especially at high humidity.

Stud and insert interfaces. A threaded insert pressed into the stack creates a path where the field meets a metal part with a different surface finish and a possible micro-gap. Voids at these interfaces are the reason interface inspection appears in our void and CT inspection guide alongside electrical tests.

Which site dominates depends on the voltage class and the layer count. High-layer-count battery busbars with thin films at low voltage can pass easily; the same construction at 1,000 V DC with a polluted cut edge can fail a limit it would have met as a coupon. That is why a PD result should always be reported together with the part geometry it was measured on.

A single line reading "PD < 10 pC, pass" cannot be audited. Six elements make a report usable, and their absence is a negotiating point rather than a technical detail.

• Circuit diagram with specimen capacitance, coupling capacitance, measuring impedance and the grounding arrangement drawn as built.
• Calibration record: calibrator model, certificate date, injected charge, injection point, and scale factor before and after the run.
• Background level measured with the specimen energised to the same voltage but with the discharge source absent where possible, or measured before the ramp with the circuit closed.
• Voltage programme: steps, ramp rate, hold times at each step, and the voltage at which the criterion was applied.
• Readings and pattern: qmax at each step plus phase-resolved screenshots with gain and timebase noted, so a reviewer can see whether the activity was internal, surface or corona.
• Criterion and statement: which clause or customer requirement the limit comes from, and a clear pass or fail at that voltage.

Four red flags worth raising before a purchase order: calibration mentioned without its injection point; no background figure; a test voltage lower than the application's maximum system voltage; and a report whose PD value is quoted without bandwidth or gain. Each of those weakens the claim, and none of them is fixed by repeating the test at the same lab with the same omissions.

Print this and keep it next to the instrument. It is written for a laminated busbar specimen, and it assumes the goal is a number somebody else can defend.

• 1. Specimen capacitance measured and written down, before and after connection.
• 2. Coupling capacitor at least an order of magnitude above specimen capacitance, or the deviation explained.
• 3. Grounding single-point, with the earth return not routed across the specimen face.
• 4. Background recorded at test voltage, not at zero; target below half the acceptance limit.
• 5. Calibrator injected at the specimen terminals, with the final circuit in place.
• 6. Calibration repeated at the end of the run and the two scale factors compared.
• 7. Bandwidth, gain and timebase recorded in the log.
• 8. Voltage steps and hold times taken from the specification, including the conditioning step if one is required.
• 9. PRPD screenshots saved at every step where activity appears, not only at the final step.
• 10. Criterion quoted from its source, and the pass statement written against the voltage actually applied.

None of this makes a bad part good, and none of it is paperwork for its own sake. It makes a good part provable — which is the only thing a PD measurement can offer a purchasing decision. If the part also has to survive mechanical qualification, the same logic applies to the sample to series path, where evidence has to be tied to the exact configuration that ships.