The Grammar of Testing: Taxonomy and Cross-Cutting Tests

Learn each test once. Then only the parameters change.

This is the dictionary module. Before entering any specific equipment, you master the taxonomy of testing — routine, type, special, receipt, commissioning and maintenance — and you run, on the bench, each of the cross-cutting tests: the ones that appear, with small variations of level and criterion, in practically every high voltage equipment family. The pedagogical promise is explicit. Each test is learned once, in depth of procedure, acceptance criterion and typical trap. In the equipment modules that follow, those same tests are simply parameterized by object.

The grammar analogy is not decorative. Just as a finite number of grammatical rules generates infinite sentences, a small number of fundamental tests generates, by combination and parameterization, every test plan of every equipment family. Insulation resistance measured on a 15 kV bushing and on a 500 MVA hydro generator is the same test. What changes is the megohmmeter voltage, the acceptance limit and the failure mode you are looking for. An engineer who understands that stops memorizing dozens of isolated procedures and starts reasoning with a matrix: test by family by parameter.

What you will be able to do

• Classify any test correctly under the normative taxonomy of routine, type and special, and under the application context of receipt, commissioning and maintenance, citing the formal definition of each category in the applicable product standard.

• Perform, with complete procedure and adequate records, the ten cross-cutting routine tests: visual, dimensional and documentary inspection; insulation resistance with polarization index and dielectric absorption ratio; power-frequency applied voltage; full and chopped lightning impulse; winding and contact resistance; capacitance and tan delta; partial discharge; temperature rise; tightness; and sequence verification.

• Justify each acceptance criterion numerically — voltage levels by class, times, waveform tolerances, limits in pC, µΩ and kelvin — from the applicable standard and its current edition.

• Build the optimal sequence of a test plan, explaining why non-destructive tests precede dielectric ones and why the partial discharge measurement closes the sequence after the impulse.

• Complete and defend the test-versus-family summary table, recognizing that the physics is the same and only level, criterion and failure mode change.

Six words that organize everything

Every product standard organizes its verification requirements into three categories with remarkably constant wording from family to family. A routine test is performed on every unit manufactured, individually, and its logical function is to prove the absence of a manufacturing defect: a shorted turn, a void in the resin, a connection someone forgot to torque. It does not prove the design is good; it proves this specific unit was built to a design already known to be good. A type test is performed once on a unit representative of a design, or again when the design changes significantly, and it proves the design itself: withstanding the full lightning impulse of the class, temperature rise inside the thermal limits, short-circuit capability. It is expensive and often degrading, which is why it is not repeated per unit. A special test is performed by agreement between purchaser and manufacturer to answer a specific contractual question the other two do not cover: noise level, zero-sequence impedance, a reference frequency response, PD under DC.

To those three factory concepts add three life-cycle contexts. The receipt test, run by the customer on arrival, compares what arrived against what was tested at the factory, usually repeating a subset of routine tests with comparison criteria rather than absolute acceptance. Commissioning, run in the field before first energization, establishes the baseline — the electrical fingerprint of the equipment as finally assembled. And maintenance tests, periodic or on demand, only make sense against that baseline and the historical trend. The lesson to carry for the rest of a career: absolute value approves in the factory, trend diagnoses in the field.

Insulation resistance: what one gigaohm proves and what it does not

The megohmmeter applies stabilized DC and measures the resulting current, made of three parts with different dynamics: capacitive current decaying in seconds, absorption or polarization current decaying in minutes, and conduction or surface leakage current that stays constant. The reading is therefore a function of time, and the diagnostic indices are born from that time dependence. Test voltage follows equipment class: 500 to 1,000 V below 1 kV, 1,000 to 2,500 V from 1 to 2.5 kV, 2,500 to 5,000 V from 2.5 to 5 kV, 5,000 V from 5 to 12 kV, and 5,000 to 15,000 V above 12 kV.

The dimensionless indices remove the effect of object size: DAR = R(60 s)/R(30 s) and PI = R(10 min)/R(1 min). Classic interpretation: DAR below 1.0 indicates compromised insulation, 1.0 to 1.25 questionable, 1.25 to 1.6 acceptable, above 1.6 excellent. IEEE Std 43-2013 recommends a minimum PI of 2.0 for class B, F and H insulation, and PI below 1.0 warrants immediate investigation. Two interpretation traps. First, insulation resistance roughly doubles for every 10 °C drop in winding temperature, so comparing measurements without correcting to a 40 °C reference is comparing different objects. Second, in very dry modern insulation with R(1 min) above about 5 GΩ, the absorption current is so small that PI loses statistical meaning, and IEEE Std 43 itself waives the minimum PI criterion in that case. So what does 1 GΩ prove? That there is no clear conduction path: no gross moisture, no carbonized tracking, no contact to ground. What it does not prove: impulse withstand, absence of partial discharge in cavities, integrity under power-frequency stress. High insulation resistance is a necessary condition, never a sufficient one, which is why it opens the test sequence rather than closing it.

Applied voltage and impulse: the two dielectric verdicts

The applied voltage test, also called separate-source withstand or AC hipot, verifies the main insulation between a complete winding and ground by applying a 45 to 65 Hz sinusoid for 60 seconds from an independent source, with the winding under test short-circuited and the others earthed. Two level languages coexist. In IEC terms the level is indexed to the highest voltage for equipment Um: IEC 60076-3:2013 prescribes 20 kV for Um = 7.2 kV, 28 kV for 12 kV, 38 kV for 17.5 kV, 50 kV for 24 kV and 70 kV for 36 kV. In traditional ANSI class terms the same equipment families are described by their voltage class, and the applied levels come from the corresponding tables. Always ask which table governs the contract. The failure criterion is breakdown — a sudden voltage collapse detected by the source overcurrent relay or by the divider. Audible surface scintillation without collapse demands investigation but is not by itself a failure. The post-breakdown procedure depends on the insulation type: self-restoring insulation such as air, gas and external surfaces may be re-energized and repeated, while in non-self-restoring insulation such as oil-paper and solid polymers a breakdown is permanent damage and you investigate rather than re-apply to confirm.

The impulse withstand test uses the 1.2/50 µs wave with T1 at ±30% (0.84 to 1.56 µs), T2 at ±20% (40 to 60 µs) and ±3% on crest, with IEC 60060-1:2025 extending the positive T1 tolerance to 2.4 µs for equipment above Um = 800 kV. The chopped wave reproduces an external flashover cutting the tail between 2 and 6 µs after the virtual origin, typically at 110% of the full BIL crest, producing the steep derivative that stresses the first turns of a winding. Sequences differ by insulation nature, and the statistics of Module 10 explain why. For non-self-restoring insulation the classic sequence is one reduced impulse at 50 to 75% of BIL followed by three full impulses per terminal, with diagnosis by oscillogram comparison — any divergence in shape between reduced and full waves reveals an incipient internal failure without having to drive it to collapse. For self-restoring insulation the 15/2 procedure applies: fifteen impulses per polarity, accepted with at most two breakdowns. External insulation also requires atmospheric correction, where Kt = k1·k2 combines air density and humidity terms; without it, the same insulator appears to change its BIL between a cold dry day and a hot humid one.

Resistance: the same method across six orders of magnitude

Two measurements share one principle, DC voltage drop, and have opposite purposes. On windings, use a stabilized-current milli or micro-ohmmeter with current typically limited to 10% of rated winding current so the measurement does not heat the object, and wait for the reading to settle, because an L/R time constant on a large winding can exceed minutes. The value only means something corrected to the reference temperature, 75 °C or 85 °C depending on the product standard, through R2 = R1·(k + T2)/(k + T1) with k = 234.5 for copper and 225 for aluminium, per IEEE Std C57.12.90. The result feeds three decisions: load loss calculation, the reference for temperature rise, and phase-to-phase comparison, where an imbalance above roughly 2 to 3% calls for investigation of connections and tap changer.

On circuit breaker and disconnector contacts the objective inverts: you are hunting microhms, not ohms. IEC 62271-1 requires DC measurement at a minimum of 50 A, and laboratory and field practice uses 100 A or more, because low currents do not break through oxide films and give optimistic readings. The routine criterion is relative, not absolute: measured resistance must not exceed 1.2 × Ru, where Ru is the resistance recorded before the temperature rise test during type testing. Typical magnitudes for high voltage breakers run to tens of microhms per pole, commonly 30 to 80 µΩ from 145 to 550 kV, with the exact value set by the design. The classic trap is clamping badly and including the terminal resistance: the measurement is four-wire Kelvin, with the potential points inside the current points.

Capacitance, tan delta and partial discharge

The capacitance bridge measures the pair (C, tan δ) typically at 10 kV and power frequency. Routine testing compares C against the nameplate or design value, since a capacitance deviation reveals geometric deformation or a short between capacitive layers, and compares tan δ against the product standard limit, since elevated tan δ reveals moisture, contamination or dielectric ageing. Three connection modes do the work. UST measures the insulation between two ungrounded terminals, excluding everything that goes to earth. GST measures everything between the energized terminal and earth. Guarded GST measures the path to earth excluding the branch diverted to the guard. Combining the three isolates individual insulation sections in a multi-terminal object, which is how a condenser bushing's C1 main insulation, measured in UST through the test tap, is separated from C2 between tap and flange, measured in guarded GST. Orders of magnitude worth memorizing for new units at 20 °C: oil-impregnated paper bushings at 0.2 to 0.4% against a 0.5% limit, resin-impregnated bushings at or below 0.35% against 0.85%, oil-immersed windings at 0.2 to 0.5%, and new XLPE cable below 0.1%, where above 1% indicates severe degradation. In service, IEEE Std C57.19.100 guidance is to monitor a bushing above twice nameplate and remove it above three times.

Partial discharge measurement is the routine test most sensitive to localized defects: a sub-millimeter cavity in a kilometer of cable shows up in picocoulombs. IEC 60270, now in its 2025 edition retitled around charge-based measurement and formally extended to AC up to 500 Hz and to DC, defines the classic circuit: a discharge-free source, a blocking impedance or filter that stops source noise reaching the measuring circuit, the test object Ca, a discharge-free coupling capacitor Ck in parallel, and the measuring impedance in the Ck branch feeding the integrating instrument that displays apparent charge in pC. Because apparent charge depends on the Ca/Ck ratio and on the stray capacitance of the setup, the system is calibrated at every setup by injecting a known charge pulse at the object terminals and adjusting the instrument scale factor. The voltage profile typically rises to a pre-stress level above the measurement voltage, holds, drops back, and only then records the charge, the aim being to ignite defects that show inception hysteresis. Typical factory acceptance limits: 100 pC for power transformers in the induced voltage test with PD, 10 pC for instrument transformers and 5 pC in specific classes and conditions, 10 pC for dry-type transformers, 5 to 10 pC for extruded cables in routine, and 5 pC for gas-insulated switchgear.

Temperature rise, tightness, and the sequence that protects the evidence

Temperature rise is a type test by nature but cross-cutting to every family. Three loading methods: real load, rare in a factory because it needs full-power supply; equivalent short circuit, the standard for transformers, where the secondary is shorted and current is injected to produce total losses, no-load plus load, while top-oil and then winding rise are measured; and back-to-back opposition, where two identical units push power through each other and the supply covers only the losses. IEC 60076-2:2011 limits for oil-immersed ONAN and ONAF units are 60 K top oil, 65 K mean winding by the resistance-variation method, and 78 K hot spot. The resistance-variation measurement is the practical application of the previous section: measure cold R1 at known θ1, then after shutdown measure R2 against time and extrapolate to the instant of switch-off to compensate the cooling during the manoeuvre, and obtain θ2 = (R2/R1)·(234.5 + θ1) − 234.5 for copper. The classic trap is recording the rise without recording the losses actually injected and the reference ambient, because a rise is a delta and a delta is only comparable against documented conditions. Tightness has three logics: insulating oil verified by overpressure of roughly 0.3 to 0.5 bar above the static column held for 12 to 24 h with no visible leak or pressure drop; gas systems where IEC 62271-1 sets an admissible leakage rate of 0.5% per year of gas mass for closed pressure systems, verified by an integral method or by localized sniffing and complemented by dew point measurement, since moisture is the precursor of corrosive arc by-products; and enclosures rated by the IP code of IEC 60529.

The order of tests is not bureaucracy, it is preservation of both the object and the evidence. Non-destructive first, dielectric last, partial discharge after the impulse. Start with inspection, which catches non-conformities for free. Then the reference measurements — insulation resistance with PI and DAR, winding resistance, capacitance and tan delta — which establish the fingerprint of the object before any stress. Only then the dielectric tests, applied, induced and impulse, from least to most severe. And closing, PD at the measurement voltage plus a repeat of the reference measurements: if the impulse opened a crack or carbonized a path, the PD reading and the before-and-after comparison will say so. That is the verdict that the object survived intact, not merely that it did not explode. The classic sequencing error is running the dielectric tests before the reference measurements. If a breakdown occurs, the evidence of the initial state is gone forever, there is no longer any way to know whether the defect pre-existed or was created by the test, and the commercial discussion between manufacturer and customer loses its technical arbiter. Sequence is therefore also a contractual protection clause.

Syllabus

• Routine, type and special tests: formal definitions and what each one actually proves

• Receipt, commissioning and maintenance: absolute value approves in the factory, trend diagnoses in the field

• Where each equipment family formally defines its taxonomy

• Visual, dimensional and documentary inspection as the cheapest and most neglected test

• Insulation resistance: capacitive, absorption and conduction currents; test voltage by equipment class

• DAR and PI: interpretation ranges, temperature correction, and when PI loses meaning

• Power-frequency applied voltage: 60 s at 45 to 65 Hz, level tables, failure criteria

• Self-restoring versus non-self-restoring insulation and what to do after a breakdown

• Lightning impulse withstand: full and chopped waves, tolerances, reduced-impulse comparison

• The 15/2 statistical procedure and atmospheric correction factors

• Winding resistance: current limit, L/R settling, temperature correction with k = 234.5 or 225

• Contact resistance: why 50 A minimum, the 1.2 × Ru criterion, and four-wire connection

• Capacitance and tan delta: UST, GST and guarded GST modes; typical values by object

• Partial discharge as a routine criterion: the IEC 60270 circuit, calibration and voltage profile

• Typical PD acceptance limits by family, from 100 pC down to 5 pC

• Temperature rise: real load, equivalent short circuit and back-to-back methods; resistance-variation measurement

• Tightness: oil overpressure, gas leakage rate and dew point, and the IP code

• The optimal test sequence and why it is also a contractual protection clause

• Cross-cutting type tests: atmospheric correction, wet and artificial pollution tests, RIV and visual corona

Laboratory work

A deliberately repetitive rotating bench. You run the same four tests — insulation resistance with PI and DAR, applied voltage, resistance, and routine partial discharge — on three different objects: a condenser bushing, a terminated XLPE cable sample, and a medium voltage current transformer. Per station: ten minutes of inspection and documentary recording of the object; insulation resistance at 5 kV with R(30 s), R(60 s) and R(10 min) recorded, DAR and PI computed and temperature corrected; applied voltage at the object's class level for 60 s, recording the ramp procedure of roughly 1 kV/s from below one third of the level and the ramp down; a resistance measurement appropriate to the object, meaning the CT winding, the cable conductor, or, on the bushing, capacitance and tan delta in UST instead; and calibration of the PD circuit with a 50 pC injection followed by measurement with the pre-stress profile. Each pair then fills the same standardized results sheet for all three objects and presents a five-minute comparison: where the criteria coincided, where they diverged, and why. The point is to feel in your hands that the procedure is one and only the parameterization changes. Companion exercise: take the test-versus-family table with the level and criterion cells blank and fill it in from the product standards on the shelf.

Key takeaway

Filling that table by hand produces a discovery no lecture replaces — the patterns. Applied voltage levels climb in standardized steps of 20, 28, 38, 50 and 70 kV. Partial discharge limits fall an order of magnitude from bulk insulation at 100 pC to compact critical insulation at 5 pC in gas-insulated switchgear and instrument transformers. Resistance changes six orders of magnitude, from ohms in a winding to microhms in a contact, while the four-wire method stays the same. Read the table by rows and you learn the test; read it by columns and you learn the equipment. And keep the two taxonomy sentences: routine proves the absence of a manufacturing defect, type proves the design, special answers one specific question the purchaser asked. Whoever sees the pattern never memorizes a table again — they derive it.

Module 11 of the Atlas Energy Academy · The Lab as a System · about 8 hours · Video series (10 × 12–15 min) plus bench practice · Prerequisites: Module 10 — Metrology, Calibration and Test Statistics