High-Voltage Testing of Apparatus

Every chapter so far has circled one question: will this insulation hold, or will it break down? A designer answers it on paper with a margin — the apparatus is built to withstand stresses well beyond its normal working voltage, because in service it will meet lightning surges, switching transients and temporary overvoltages far larger than its rated value. But a margin claimed on paper is worthless until it is demonstrated. High-voltage testing is the demonstration: a controlled overvoltage is applied to the finished apparatus, and its survival proves the margin is real.

This is where Parts 4 and 5 come together. The test voltage is produced by the generators of Part 4 — a transformer cascade or resonant set for AC, a rectifier for DC, a Marx generator for impulses — and its magnitude and shape are read by the dividers, sphere gaps and recorders of Part 5. Testing is the application those two parts were building toward, and it looks ahead to Part 7, where the chosen test levels are justified by insulation coordination — the deliberate matching of equipment strength to the overvoltages the network can deliver.

Not every test serves the same end, and the standards sort them by purpose. A type test (or design test) is performed once, on a representative sample, to prove that the design meets its specification; it is elaborate, expensive, sometimes destructive, and need not be repeated on every unit. A routine test is performed on every manufactured unit to catch defects in materials and workmanship; it must be non-destructive and quick, so a sound unit passes unharmed. Between them sits the sample (acceptance) test, run on a few units drawn from a batch, and after installation comes the commissioning or field test, a gentler check that the apparatus survived transport and erection and is safe to energise.

Cutting across those classes is a deeper distinction in what the test asks of the insulation. A withstand test applies a specified voltage for a specified time and asks only one thing: did it survive? Pass means no breakdown; the apparatus is undamaged and goes into service. A disruptive-discharge (flashover) test instead raises the voltage until the insulation actually breaks down, measuring its real strength — which necessarily damages, or at least disturbs, the object. The choice between them depends on whether the insulation can recover.

That recovery is the dividing line. Self-restoring insulation — the air around a gap, the clean surface of an outdoor insulator — regains its full strength after a flashover, so it can be tested to breakdown again and again, and its strength is described statistically. Non-self-restoring insulation — the oil-paper inside a transformer, the dielectric of a cable — is permanently destroyed by a single breakdown, so it may only ever be given a withstand test, never deliberately broken down. This single property decides which test a given piece of apparatus can be subjected to, and it is why internal insulation is always proven by survival, never by destruction.

The most universal stress is the power-frequency (AC) withstand test: the apparatus is subjected to a specified voltage, well above its rated value, at 50 or 60 Hz for a stated time — classically one minute. It exercises the bulk insulation at the frequency it lives with in service, and is a routine test on almost everything. The voltage comes from a testing transformer or, for large capacitive loads such as cables, a resonant set (Chapter 12), and its peak is read with a peak voltmeter or sphere gap (Chapter 16).

For outdoor apparatus the test is done in two conditions. The dry test checks the insulation in clean, still air; the wet test sprays the insulator with water of specified conductivity and rate, simulating rain, because a wet surface flashes over at a markedly lower voltage than a dry one. An outdoor insulator must pass both. Together with the standardised stresses, the panel of four test waves below is the toolkit of high-voltage testing — and three of the four are simply the waveforms generated back in Part 4.

Some apparatus is stressed with direct voltage. Cables, capacitors and HVDC equipment are naturally tested with DC, both because that is how they are stressed in service and because a DC source can be made compact for the large capacitances involved — an AC test of a long cable would demand an enormous charging current, whereas DC charges it once and holds. The DC withstand test applies a specified level, typically for several minutes, and is often combined with a leakage-current measurement: a steadily rising leakage as the voltage is held is an early sign of a deteriorating dielectric, making the DC test a diagnostic as well as a pass/fail gate.

The caution is that DC does not stress insulation the same way AC does. The field distribution under DC is governed by conductivity rather than permittivity, so the stress falls differently across a composite dielectric, and a DC test is not a substitute for an AC one — it is a different question asked of the same object. This is why the standards prescribe DC tests specifically where the service stress is DC, and prescribe AC tests elsewhere.

The most demanding stress is the lightning impulse, the standard 1.2/50 µs wave of Chapter 13, applied by a Marx generator and recorded by the impulse divider and digitiser of Chapter 17. The level an apparatus must withstand is its Basic Insulation Level (BIL) — the cornerstone figure of insulation coordination. The impulse test comes in two important forms. The full-wave test applies the complete 1.2/50 µs wave. The chopped-wave test truncates the tail with a sphere or rod gap that flashes over a microsecond or two after the crest — a trigatron-fired gap, exactly the triggering of Chapter 14 — reproducing the steep collapse a nearby flashover would inflict and testing the insulation against that sudden cut.

Because impulse breakdown of self-restoring insulation is statistical, the test uses a counting criterion rather than a single shot. A common form applies a fixed number of impulses at the rated level — for example fifteen impulses of each polarity — and the apparatus passes if no more than a set number (often two) cause a disruptive discharge across self-restoring parts, with no breakdown permitted in non-self-restoring internal insulation. The repeated application is the only honest way to characterise a strength that scatters from shot to shot.

At extra-high and ultra-high voltages a different transient becomes the design-limiting stress: the switching impulse, the long 250/2500 µs wave produced when circuit breakers operate. For systems below about 300 kV the lightning impulse governs, but above it the slow switching surge takes over, because long air gaps are weaker against the slow front of a switching impulse than against the fast lightning wave — a counter-intuitive fact that makes switching-impulse testing essential for EHV/UHV apparatus. The wave is generated by the same Marx generator with its shaping resistors retuned for the much longer times, and the test, like the lightning-impulse test, is evaluated statistically through the 50% flashover voltage.

Outdoor insulation faces an enemy beyond voltage: a contaminated, wetted surface. Salt spray near coasts, industrial dust and desert sand settle on insulators and, when dampened by dew or drizzle, form a conducting film that can flash the insulator over at a fraction of its clean strength. The artificial-pollution test reproduces this in the laboratory — typically a salt-fog or a solid-layer method that coats the insulator with a defined contamination and then energises it — and is the test that decides how long an insulator string must be for a given polluted environment.

Because flashover of self-restoring insulation scatters, its strength is not a single number but a probability distribution. The central figure is the 50% flashover voltage \(U_{50}\), the level at which half the applications cause flashover, usually taken as normally distributed with a standard deviation \(\sigma\). From these, the statistical withstand voltage — the level with only a small (conventionally 10%) chance of flashover — is

To find \(U_{50}\) in the laboratory without endless trials, the up-and-down method is used: an impulse is applied, and the next is set one step lower if it flashed over or one step higher if it withstood, so the test voltage hunts around the 50% level. Averaging the applied levels over many shots gives \(U_{50}\) efficiently. Two further points complete the picture. First, the external-insulation test voltages must be corrected for air conditions exactly as the sphere gap was in Chapter 16 — an atmospheric correction factor \(K_t\) (air density times a humidity term) scales the standard level to the test-room conditions. Second, all of this statistics applies only to self-restoring insulation; the oil-paper inside a transformer gets a withstand test and nothing more.

For each item, first identify what it tests — the purpose of a test class, the withstand/disruptive distinction, one of the four stresses, or the flashover statistics — then apply it. Difficulty rises down the list.