Tan Delta and Capacitance Testing

A volumetric, global quantity — blind to the localised defect by construction.

By the time you reach this module you have measured dissipation factor dozens of times — on the bushing of a power transformer, on instrument transformers, on stator bars — always as a routine number compared against a table limit. This module turns that number into a diagnostic tool. It starts from the physics of dielectric loss, crosses the metrology of the bridge and of the measurement modes, and arrives at what actually separates the test engineer from the instrument operator: baseline, temperature correction, critical trend reading, and honest recognition of the cases where tan delta lies — the good value hiding a localised defect, and the bad value that is only surface contamination.

The short answer of this module is that tan delta is a volumetric and global quantity. It integrates all the losses in the dielectric volume under test and divides them by the reactive energy stored. That characteristic is its strength, giving sensitivity to distributed degradation such as moisture, thermal ageing and oil contamination, and its structural weakness, giving relative blindness to localised defects. A discharging cavity or a wet joint can be diluted into the healthy capacitance of kilometres of cable or tonnes of oil-paper. Everything that follows develops the practical consequences of that duality.

What you will be able to do

• Explain the physics of tan delta — what raises dielectric losses (ionic conduction, interfacial polarisation, moisture, discharges in cavities) and what does not (larger capacitance, higher voltage, bigger object, since tan delta is dimensionless).

• Derive and apply tan δ = I_R/I_C = P/(ω·C·U²), convert between dissipation factor and power factor, and justify why the two are numerically indistinguishable below about 10%.

• Configure measurements on bushings (C1 and C2), current transformers, voltage transformers, power transformers (CH, CL, CHL), cables and rotating machines, choosing correctly between UST, GST and GST-guard and predicting which currents enter the result.

• Build baselines and trend curves, apply individual temperature correction, and explain why the generic fixed-factor tables were abandoned by the standards.

• Interpret tip-up in rotating machines per IEC 60034-27-3 and tan delta at 0.1 Hz in cables per IEEE 400.2-2024 with their current numerical criteria.

• Recognise field interference — electrostatic coupling, surface leakage, power frequency noise — apply rejection techniques, and produce an auditable report recording configuration, temperature, humidity and uncertainty.

The physics, and the metrological challenge hidden inside it

An ideal insulation under sinusoidal voltage would be a pure capacitor: current leading by exactly 90 degrees, average active power zero. Real insulation conducts a little and polarises with delay. In the parallel model the dielectric is a capacitance in parallel with an equivalent loss resistance. The capacitive current is I_C = ω·C_p·U in quadrature, the resistive current is I_R = U/R_p in phase, and the angle δ between the total current and the purely capacitive axis defines tan δ = I_R/I_C = 1/(ω·R_p·C_p) = P/(ω·C·U²). That last form is the operational one: the bridge measures dissipated active power, capacitance and voltage, and tan delta is the ratio of loss power to reactive power.

Two related conventions cause endless confusion in reports. Dissipation factor is tan δ; insulation power factor is sin δ. They relate by PF = DF/√(1+DF²). North American practice publishes limits in percent power factor and European practice in tan delta, and the numerical difference only becomes perceptible at high values — at 1.0% DF the PF is 0.99995%, and at 10% DF the PF is 9.95%. The editorial rule is simple: never move a value from a percent power factor column into a tan delta column without declaring the conversion, even though below 10% the difference is smaller than the uncertainty of any field instrument.

Modes are current topologies, not instrument buttons

The three measurement modes are three topologies for the return of current. In grounded specimen test mode, the measuring circuit sits in the earth path and measures every current returning to earth. In GST-guard, selected paths are diverted to the guard terminal and excluded from the result. In ungrounded specimen test mode, the insulation between two ungrounded terminals is measured, automatically excluding everything that flows to the tank or the structure. The operational question before every measurement is always the same: which current do I want to measure, and where do the ones I do not want return?

Two errors around this matrix deserve naming. The first is accidental swapping of the C1 and C2 arrangements on a bushing: the test tap typically withstands 500 V to 1 kV, and applying 10 kV to it can puncture it. The second is measuring C1 in GST instead of UST. In GST the measuring circuit picks up every current returning to earth, including surface leakage across the bushing porcelain and stray paths to the flange, in addition to the intended C1. On a humid day, surface leakage grows and adds watts to the measurement, inflating the apparent tan delta and producing a false positive on a healthy bushing.

Temperature correction is the number one trap

Raising temperature increases ionic mobility, reduces oil resistivity, increases paper conductivity and accelerates interfacial polarisation. Because conductivity follows an Arrhenius law and the conductive share of tan delta is proportional to σ/ω, a few degrees Celsius can move the reading significantly — and the more moist or aged the insulation, the more it moves. This is the mechanism behind the classic field puzzle: a reading that rose 30% between morning and afternoon.

The modern solution exploits the frequency-temperature equivalence of dielectric response: the tan delta against frequency curve of an oil-paper system shifts horizontally with temperature without changing shape. The instrument runs a short frequency sweep at the actual object temperature, identifies the individual shape of the response, and shifts the curve by an Arrhenius model to calculate the equivalent value at 20 °C. Reading those curves is itself diagnostic. New, dry insulation gives a low curve with a small slope. Wet or contaminated insulation gives high values with a strong thermal slope and marked growth at low frequency. Aged insulation gives a curve shifted upward with a possible change of shape caused by polar oxidation products. The thermal slope is an indicator of condition in its own right.

The verdict procedure for a suspicious jump therefore has three steps. First the physics: real insulation degradation does not evolve in hours, so an increase of 30% between 9 am and 3 pm has negligible probability of being ageing and very high probability of being temperature or surface condition. Second the verification: record the object temperature at both measurements, not the air temperature; check humidity and surface condition; repeat at a shifted frequency. Third the verdict: correct both readings to 20 °C by the individual method and compare against the baseline. If, corrected, they coincide, the asset is stable and the increase was a thermal artefact. If they diverge, there is an event to investigate — starting with capacitance.

Baseline, trend, and the eight cases where the number lies

The order of authority for judging a reading is: nameplate or factory report value, first commissioning measurement, the asset's own history, other phases of the same asset, sister units, statistical population, and only then the generic standard limit. There is no universal annual growth rate — no standard establishes a rule such as five percent per year — and any rate is only meaningful if the campaigns share configuration, voltage, frequency, mode, corrected temperature and a traceable instrument. Change within historical repeatability is stability; between two and three standard deviations, repeat and review conditions; above three standard deviations, or monotonic growth across three campaigns, investigate even without crossing an absolute limit.

Eight cases vaccinate against the two symmetric failures of judgement. Tan delta good, equipment bad: a defective joint in a long XLPE cable, where healthy kilometres dilute the loss; a bushing that is statistically anomalous against its model population while still under 0.5%; a shorted capacitive foil where C1 steps before tan delta reacts; and a small cavity with severe PD whose discharge energy is a negligible share of global loss. Tan delta bad, equipment good: a wet or contaminated porcelain, where surface leakage enters the GST result and returns to normal after cleaning, drying and guarding; electrostatic interference giving a high, unstable or negative reading that disappears when frequency is shifted or grounding improved; an inadequate generic thermal correction inflating the corrected value; and a circuit included by mistake — an arrester, an auxiliary VT, a connected tertiary, an unguarded surface. On that last point, one myth needs killing: a negative tan delta is not better-than-perfect insulation. In a passive capacitive system it indicates interference, a wrong guard or ground configuration, coupling between windings, or inductive elements in the measured circuit.

Syllabus

• The parallel model and the loss angle: tan δ = I_R/I_C = 1/(ω·R_p·C_p) = P/(ω·C·U²)

• A worked example at 10 kV that shows the metrological challenge: resolving 19 µA in phase inside 3.77 mA in quadrature

• Complex permittivity, conduction, and why moisture shows up disproportionately at low frequency

• Polarisation mechanisms: electronic, ionic, dipolar, interfacial, conduction, ionisation

• Dissipation factor against power factor and the redaction rule for reporting either

• Typical values by insulation system, as training material and never as acceptance criteria

• Instrumentation: the Schering bridge, the compressed-gas standard capacitor, and the modern digital bridge

• Why 10 kV became the traditional test voltage and when not to use it

• The three measurement modes as current topologies, not instrument buttons

• The mode-by-equipment matrix and the destructive C1/C2 connection error

• Acceptance criteria for liquid-immersed transformers and for oil-impregnated-paper bushings

• Temperature correction: why the fixed-factor tables were removed and how individual correction works

• Tip-up in rotating machines: definitions and current limits

• Tan delta at 0.1 Hz in medium voltage cable: mean, tip-up and stability criteria

• Interference families and rejection techniques, and the myth of negative tan delta

• Baseline, comparison hierarchy, capacitance as the most objective alarm, and the auditable report

• Eight cases where the number lies

Laboratory work

Bench 1 — condenser bushing: measure C1 in UST at 10 kV and C2 in GST-guard at 500 V, compare against nameplate, and simulate a foil short didactically with an auxiliary capacitor in parallel to observe the C1 step with tan delta almost unchanged. Bench 2 — medium voltage cable sample with an induced thermal defect: measure tan delta at increasing voltage before and after controlled heating of the sample, plot the tan delta against temperature curve, and calculate the error a fixed correction factor would have introduced. Bench 3 — sensitivity analysis: measure the same sample in the morning in a cold environment and in the afternoon when warmed, reproducing the field puzzle, and require the student to decide, with the individual curve in hand, whether the variation is degradation or temperature. All results recorded in an auditable report that includes complete asset identification, instrument and calibration certificate, mode and connection diagram for every measurement, voltage, frequency and number of readings, object temperature (not air temperature) and relative humidity, thermal correction method applied, capacitance and tan delta with uncertainty, explicit comparison against nameplate, history and other phases, and any interference anomaly with the action taken.

Key takeaway

Analyse capacitance first and tan delta second. Correct temperature individually or do not correct at all. And never let tan delta condemn or acquit alone — it belongs in the report next to capacitance, partial discharge, oil analysis, insulation resistance and, where available, frequency response.

Module 20 of the Atlas Energy Academy · Diagnostic Methods · about 6 hours · Video series (8 × 10–15 min) plus practice · Prerequisites: Part IV completed