VLF & Cable Testing

Partial Discharge Testing: What It Finds, What It Misses, and What It Cannot

VA-TEK Engineering VA-TEK Engineering 10 min read

A newly installed 11 kV feeder passed its VLF withstand test, was energised, and failed inside a week. The post-mortem found a void in the first joint, a few millimetres across, sitting quietly at a stress the withstand test never asked it to survive. The withstand test was not wrong. It was asked the wrong question.

Partial discharge testing asks a different one: not “does this insulation hold voltage” but “is there a place inside it that is already breaking down”. It is the only common field test that answers with a location.

The short answer

  • What it finds: localised, non-bridging breakdowns — voids, delaminations, semicon edge defects, contaminated or badly made joints and terminations. It points at a place in the cable.
  • What it misses: distributed ageing. A cable full of water trees that are not yet discharging will read clean.
  • Not destructive. PD measurement is a passive observation of discharges that occur at a given test voltage. It injects no additional stress beyond the voltage the insulation is already under.
  • Test high enough or don’t bother. Most discharge inception voltages sit above 1.5 U₀; CIGRE TB 728 recommends measuring at or above 1.7 U₀. Below inception you will measure nothing on a defective asset.
  • Numbers do not travel. A pC figure from one cable is not comparable with a pC figure from another, or from the same cable under a different excitation. Read inception/extinction values and trends, not absolute charge.

1. What partial discharge is

Partial discharge is a breakdown that does not complete. Inside a void in solid insulation, or along an interface between materials with different permittivity, the local field can exceed the breakdown strength of the gas or surface while the bulk insulation, which is much thicker and much stronger, still holds. The result is a small, fast, repeated charge transfer — a discharge that partially bridges the insulation.

Each event transfers a tiny amount of charge measured in picocoulombs, releases a little heat, and degrades the surrounding material. In organic insulation the process is cumulative: the discharge erodes the void wall, the wall carbonises, the field concentrates further, and the discharge grows. This is why PD is treated as a diagnosed failure mode rather than a curiosity — once it starts, it tends to accelerate.

The mechanisms that produce it are, in rough order of how often they show up in field work: voids and contamination in extruded insulation; poor bonding or contamination at cable joints; wrongly installed stress control at terminations; delamination at the semicon/insulation interface; metallic protrusions and floating components in GIS and switchgear.

2. The four ways to measure it

PD is not one instrument. The measurement method follows the asset and the energisation available:

Method Used on Practical note
Conventional PD measurement (IEC 60270) Factory and laboratory testing; off-line field tests at power frequency The reference method. Direct coupling, calibrated in pC, well-defined bandwidth. Needs a power-frequency source of adequate capacity.
VLF PD Shielded MV and HV cable in the field Makes PD measurement possible on cable lengths that no 50 Hz source can charge. Interpretation differs from power-frequency PD — see section 5.
On-line PD monitoring In-service cable, switchgear, transformers, rotating machines No outage required; sensors are permanently installed. Separates discharge from noise statistically. NFPA 70B-2023 lists PD among on-line electrical tests.
UHF and acoustic GIS and switchgear specifically UHF is immune to the acoustic noise floor of a substation; acoustic triangulates a location. Complementary rather than interchangeable with electrical methods.

Choosing between them is a question of what you need the answer to do. A type test needs the reference method and calibration. A commissioning decision on a long cable needs VLF PD because there is no other practical way to stress it. A monitoring programme on an ageing switchgear fleet needs on-line. This article goes into the VLF case specifically.

3. Is partial discharge testing destructive?

No — and this question is asked often enough to be worth answering precisely.

The measurement itself is passive: you observe the charge that flows on the coupling capacitor or HFCT when a discharge occurs. Nothing is injected into the insulation beyond the applied test voltage.

The applied voltage is a different matter, and it is where the honest answer lives. To find a defect you have to reach its inception voltage, which is normally above the operating stress. On cable, PD is measured during or immediately after a VLF withstand, so the insulation experiences the withstand stress whether or not you attach a PD detector. In that sense PD adds no risk beyond what the withstand was going to apply. It does not add risk of its own.

The exception worth flagging: repeated energisation of a cable that already has an active discharge can accelerate degradation. That is an argument for measuring once, properly, at the right level — not for measuring at a safe low level and hoping.

4. Why the test level is 1.7 U₀ and not something comfortable

Discharge inception — the voltage at which a void begins to discharge at all — usually sits above operating stress and often above 1.5 U₀. If you test at a voltage below inception, the defect does not discharge, your detector reads the noise floor, and you conclude the cable is clean. The test is not wrong; it was never running.

That is the reasoning behind CIGRE TB 728’s recommendation to measure at or above 1.7 U₀. It is not a safety margin, it is an inception threshold. The corollary is that a “no PD detected” result is only meaningful if you state the voltage it was measured at — a clean result at 1.0 U₀ means very little.

On cable this normally comes for free, because the withstand level sits well above 1.7 U₀ and the PD measurement is taken during the hold and the rundown. The rundown matters: as voltage decreases, defects extinguish, and the extinction voltage is a more repeatable quantity than the inception voltage.

5. The comparison trap

The most common mistake in field PD work is comparing pC numbers across assets. Three reasons they do not travel:

  • Apparent charge depends on the coupling path. What you measure is the charge appearing at your terminals, not the charge in the void. Cable length, sensor position and coupling capacitance all change the number for an identical defect.
  • VLF PD is not power-frequency PD. At 0.1 Hz the discharge physics changes: the time between voltage reversals is long enough for charge to redistribute, so discharge inception and the observed pattern differ from the 50 Hz case. Amplitudes and phase-resolved patterns are not directly transferable between the two excitations.
  • Residual charge from the previous test affects the reading. Ordering matters. A measurement taken immediately after a withstand can read differently from one taken on a rested cable.

What is comparable is the behaviour of one asset over time, measured the same way, at the same voltage, with the same coupling. Inception voltage, extinction voltage and the trend of either across successive tests are the quantities to record. Absolute charge in pC is a diagnostic detail, not a score.

6. Partial discharge testing of HT cable

For medium-voltage distribution cable — the 11 kV and 33 kV classes that carry most of the world’s urban distribution — the practical route to a PD measurement is a VLF source with external coupling. The set applies the stress; an HFCT or coupling capacitor with a PD detector does the measurement. The two are bought separately because the coupling and detection chain has to be calibrated and is often shared across a fleet.

Two constraints shape the choice. First, the waveform must be sine. Cosine-rectangular VLF is a perfectly good withstand waveform and completely useless for PD diagnosis, because the fast transitions swamp the discharge signal. Second, the source has to reach the required test level: PD measurement at 1.7 U₀ on a 33 kV cable needs a set with headroom above that, which is where the higher voltage ratings in a family pay for themselves.

Where the cable is long, current becomes the limiting factor rather than voltage: a 0.1 Hz source can only charge so much capacitance. The VLF test set selector works that arithmetic from cable capacitance and run length, and the test voltage guide covers the levels for each cable class.

7. Where PD fits against the other two tests

Partial discharge, tan delta and VLF withstand answer three different questions and have three different blind spots. Withstand gives a pass or a fail today and sees distributed ageing not at all. Tan delta averages the whole insulation volume and misses a single bad joint. PD locates localised defects and says little about bulk condition.

Run together on one energisation, they produce a proof test and a condition record from the same site visit — which is what IEEE 400.2-2024 formalised as a monitored withstand, and what this comparison of the three tests sets out step by step. The withstand-versus-diagnostic distinction explains when a diagnostic measurement earns its cost and when it does not.

Frequently asked questions

What is a normal partial discharge level? There is no universal threshold. Interpretation depends on the asset type, the measurement method, the coupling and the voltage at which the reading was taken. For cable and GIS, an absence of detectable discharge above the relevant inception level is the useful result; a specific pC limit is only meaningful within a defined method and asset class.

Can partial discharge be measured on-line? Yes, and for in-service assets it is often the only option. Permanently installed sensors on cable joints, switchgear and transformers, combined with noise separation, allow PD to be trended without an outage. NFPA 70B-2023 lists PD among recognised on-line electrical tests.

Does a clean PD test mean the cable is good? It means no discharge was detected at the level and method used. That rules out active localised defects above the detection threshold. It does not rule out distributed ageing such as water trees that are not yet discharging — which is why a PD result is read alongside tan delta rather than instead of it.

Why can’t I use a cosine-rectangular VLF set for PD? The fast polarity transitions of a CR waveform produce broadband transients that mask the discharge signal and make phase-resolved analysis meaningless. PD diagnosis requires a sinusoidal excitation.

What about GB/T 3048.12-2025? China’s PD test standard took effect on 1 May 2026 and replaces the 2007 edition, but its scope is factory and type testing of cables — it explicitly does not cover post-installation PD measurement of cable systems. Field PD on installed cable still follows the IEEE and CIGRE guidance.

VA-TEK builds VLF withstand sets and integrated VLF-plus-tan-delta systems for 5–46 kV shielded cable; PD coupling and detection are added to the source. For a test-voltage or coupling recommendation against a specific cable schedule, request a quote or contact the engineering team. For the wider method, start with the VLF cable testing guide.

Sources: IEC 60270; IEEE Std 400.2-2024; CIGRE TB 728 (2018); NFPA 70B-2023; GB/T 3048.12-2025 (published 2025-10-05, effective 2026-05-01). Test levels and interpretation must be confirmed against the current official standard and the operator’s own procedure before a live job.

Stay Updated with VA-TEK Insights

Get technical updates on VLF testing, relay protection and transformer diagnostics — new products, application notes and test guides. No spam, unsubscribe anytime.

VA-TEK Technical Updates
💬