Partial Discharge Measurement and Analysis

A partial discharge (PD) is a localised electrical discharge that bridges only part of the insulation between two conductors. It is not a full breakdown — the main insulation still holds and the apparatus keeps working — but a small spark has occurred at a weak spot, and it will occur again and again, every cycle, as long as the voltage is applied. Three forms dominate. An internal discharge happens inside a gas-filled void or cavity trapped in solid insulation. A surface discharge creeps along an interface where the tangential field is high, such as the edge of an electrode on an insulator. And corona is the discharge in gas at a sharp point or thin wire, met already in the breakdown chapters.

What makes PD so important is that it is both a symptom and a cause. It is a symptom because a measurable discharge betrays a defect — a void left in casting, a delamination, a sharp burr — that should not be there. It is a cause because each tiny spark bombards the cavity wall, eroding the material, growing conductive channels called electrical trees, and over months or years carving a path that finally becomes a full breakdown. Detecting PD therefore predicts failure long before it happens, which is why a PD test is among the most valued routine and diagnostic tests of all.

Why should a small gas void discharge while the solid around it, far stronger, stays intact? The answer is a piece of field theory from the very first chapters. At the boundary between the solid and the gas, the normal component of the electric flux density \(D = \varepsilon E\) is continuous. The gas has permittivity \(\varepsilon_0\) and the solid \(\varepsilon_0\varepsilon_r\); for a flat void lying across the field, continuity of \(D\) forces the field inside the gas to be \(\varepsilon_r\) times larger than in the surrounding solid:

So the void suffers a double disadvantage. It experiences a field several times stronger than the bulk (a typical \(\varepsilon_r\) of 3–5), and the gas inside it breaks down at a far lower strength than the solid that surrounds it — air gives way at about \(3~\mathrm{kV/mm}\) while good solid insulation withstands tens of kilovolts per millimetre. The combination guarantees that, as the applied voltage rises, the gas in the cavity is the first thing to break down, sparking over while every other part of the insulation comfortably holds. That first spark is the partial discharge.

To turn that picture into something measurable, the cavity is represented by a three-capacitor equivalent circuit — the classic abc model. The cavity itself is a capacitor \(C_c\); the slice of healthy dielectric directly in series with it (above and below the void) is \(C_b\); and all the remaining sound insulation in parallel is \(C_a\). Before any discharge, the applied voltage divides capacitively, placing across the cavity the fraction

The cavity \(C_c\) behaves like a miniature spark gap. As the AC voltage climbs, \(V_c\) climbs with it until it reaches the gas breakdown value — the cavity's own little inception level. At that instant the void sparks over, its voltage collapses almost to zero, and the discharge is done. The terminal voltage at which this first occurs is the partial discharge inception voltage, and below it the insulation is silent. Each time the applied wave drives \(V_c\) back up to inception, the cavity fires again, producing the characteristic train of pulses, several per half-cycle, clustered on the rising flanks of the voltage.

When the cavity fires, a real charge moves inside the void — but that charge is buried in the insulation and cannot be reached. What the external circuit can sense is the sudden small change it forces at the terminals: the collapse of the cavity voltage by \(\Delta V_c\) pushes a pulse of charge through \(C_b\) into the measuring circuit. This terminal-measurable quantity is the apparent charge \(q\):

It is called "apparent" because it is not the true charge that flowed in the void — that real charge, involving \(C_c\) as well, is always larger and is inaccessible. The apparent charge is simply the charge that, if injected at the terminals, would produce the same terminal effect as the discharge. That is exactly the quantity we can calibrate and read, so PD severity is universally quoted as an apparent charge in picocoulombs: a few pC in good insulation, hundreds or thousands of pC in deteriorating apparatus. A rising apparent charge over time is a clear warning that a defect is growing.

A PD pulse is a tiny, very fast event — a few picocoulombs in a few nanoseconds — riding on top of a large power-frequency voltage. Pulling it out is the job of the straight detection circuit standardised in IEC 60270. The test object \(C_x\) is energised, and beside it a coupling capacitor \(C_k\) is connected in parallel. When the cavity fires, the fast pulse needs a low-impedance path at high frequency, and \(C_k\) provides it: the pulse circulates between \(C_x\) and \(C_k\), passing through a measuring impedance \(Z_m\) placed in series with one of them. Across \(Z_m\) the pulse develops a small voltage that is amplified, filtered and displayed, while the power-frequency voltage — being low-frequency — is blocked from the detector.

Two refinements matter in practice. The detector may be narrow-band (tuned to a chosen frequency for good noise rejection) or wide-band (preserving the pulse shape for analysis). And because the signals are so small, external interference — radio, switching noise, corona on the leads — easily swamps a real PD; a balanced (bridge) circuit that compares the test object against a discharge-free reference can cancel common interference, letting the genuine internal discharge stand out.

The detector produces a voltage pulse, but we want an answer in picocoulombs, and the relationship between them depends on the stray capacitances of the whole set-up. The cure is direct calibration: before the test, a known charge is injected at the terminals of the test object and the detector's response is noted, fixing the scale. The known charge is produced by stepping a small, accurately known voltage \(V_0\) across a small, accurately known capacitor \(C_0\):

A calibrator that switches, say, \(5~\mathrm{V}\) across a \(1~\mathrm{pF}\) capacitor injects a clean \(5~\mathrm{pC}\) pulse; the detector deflection it produces then defines "\(5~\mathrm{pC}\)" on the screen, and every later reading is scaled against it. Calibration must be done on the actual circuit, with the actual test object connected, because it is the capacitances of that arrangement that set the response — a calibration is not transferable from one object to another. With it done, the instrument reports apparent charge directly, and the diagnostician can compare today's reading against the standard's permitted level and against the object's own history.

The magnitude of the apparent charge tells you how much discharge there is; the pattern tells you what kind. Because PD pulses fire at preferred parts of the AC cycle, plotting each pulse's magnitude against the phase angle of the voltage builds a picture — the phase-resolved PD (PRPD) pattern — whose shape is a fingerprint of the defect. Internal voids tend to discharge symmetrically on the rising flanks of both half-cycles, giving twin rabbit-ear clusters; surface discharges skew differently between the half-cycles; and corona at a point appears strongly in one polarity, betraying its asymmetry.

Two voltages complete the diagnostic. The PD inception voltage (PDIV) is the level at which discharges first appear as the voltage is raised; the PD extinction voltage (PDEV) is the level at which they cease as it is lowered. The extinction voltage is always the lower of the two, because once the cavity has been ionised it re-discharges more easily — residual charge and free electrons lower the bar — so the voltage must drop below inception before the discharges stop. A healthy margin between the working voltage and the PDIV is a primary design goal, and a PDIV that falls over time is a clear sign of progressing damage.

The electrical method just described is the reference, but a discharge announces itself in several ways at once, and each has its own detector. A PD makes a faint sound — an ultrasonic click — that acoustic sensors can pick up and even locate by timing. It emits light, so an optical method using ultraviolet-sensitive cameras can see surface discharge and corona in the open. It produces chemical by-products, so in an oil-filled transformer dissolved-gas analysis (DGA) reads the gases the discharge cooks out of the oil, distinguishing PD from overheating and arcing. And it radiates radio-frequency energy, so the UHF method, sensing emissions in the hundreds of megahertz, has become the standard for gas-insulated switchgear where the metal enclosure both shields external noise and guides the UHF signal to internal antennas.

Taken together these methods make PD the foundation of modern insulation diagnostics. A withstand test (Chapter 18) gives a single pass/fail at a moment in time; a PD measurement reveals the slow internal health of the insulation and how it changes — which is why it sits at the heart of the condition-monitoring strategies of Chapter 21, alongside the dielectric-loss measurement we turn to next.

For each item, first identify what it tests — the nature of PD, the field-enhancement reason, the abc model, the apparent charge, the detection/calibration circuit, or the patterns — then apply it. Difficulty rises down the list.