Every insulation system leaks a little. Apply AC voltage across ground-wall insulation and the current splits into two parts: a large capacitive current that leads the voltage by 90°, and a small resistive current in phase with it that turns into heat. The dissipation factor — tan δ — is the ratio of those two. It is the tangent of the loss angle, and it tells you how lossy the insulation is.
For a stator winding, that single number carries a lot. A well-cured, well-impregnated coil has low, stable losses. A coil with the wrong resin, an incomplete cure, or voids left in the insulation loses more — and loses progressively more as the voltage climbs and those voids start to discharge. IEC 60034-27-3 is the standard that turns that behavior into a measurement you can use to accept a new coil or trend an aging one.
This is the machine-winding version of the test. If you came here for the transformer side, that is a different procedure on different geometry — see transformer power factor / tan delta testing. This page covers stator bars, coils, and complete windings.
Table of Contents
What it applies to, and what it can’t do
The method is for form-wound stator bars and coils with a conductive slot coating, rated 6 kV and above, whether uninstalled, installed in a core, or built into a complete winding — new or aged. It covers vacuum-pressure-impregnated (VPI) and resin-rich fully-loaded insulation. It does not apply to non-impregnated bars, coils, or windings.
Two limits on what the number means are worth stating up front:
- It does not localize. Unlike an off-line partial discharge measurement, tan δ gives you the average dielectric loss of the whole insulation volume between the conductor and the slot coating. It tells you the insulation is lossy; it does not tell you where. For localization you need partial discharge testing.
- It does not predict life. Trending tan δ over time flags a change in condition, but it cannot forecast time to failure. It is a quality and condition indicator, not a countdown.
One more note on terminology: tan δ and power factor (cos φ) are physically different quantities, but for any modern winding they are numerically almost identical, because tan δ stays well below 100 × 10⁻³. People use the terms interchangeably; the standard uses tan δ.
The three numbers that matter
The test does not produce one value. It produces a curve of tan δ measured in steps as the voltage rises from 0.2 UN to rated voltage UN, and from that curve you pull three characteristic values. Each one reveals something different.
tan δ at 0.2 UN — the baseline. Measured at low voltage, below where any voids start discharging, this is the intrinsic dielectric loss of the solid insulation. It reflects the cure state and the resin system. A higher-than-normal baseline points to something structural — the wrong resin composition, an inadequate cure, or surface contamination and moisture. It is your read on manufacturing quality.
Delta tan delta (Δtan δ) — the step increase. As you raise the voltage in 0.2 UN steps, you record how much tan δ climbs at each step. The largest step increase, expressed per 0.2 UN, indicates void content and delamination. Solid-insulation losses barely move with voltage. Voids do the opposite — as voltage rises, more of them begin to discharge, and each discharging void adds loss. A steep Δtan δ is the signature of gas-filled voids inside the insulation.
Tip-up — the headline void indicator. Defined as the difference between tan δ at 0.6 UN and at 0.2 UN, the tip-up is the classic single-figure measure of how much the insulation “tips up” with voltage. A low tip-up means a dense, well-impregnated insulation with few voids. A high tip-up means the insulation is ionizing internally well below rated voltage — the mark of poor impregnation or delamination.
Read together, these three separate two very different problems. A high baseline with a low tip-up points at the solid material — cure, resin, contamination. A normal baseline with a high tip-up points at voids and delamination. That distinction is the whole diagnostic value of the test, and it maps directly onto the stator winding failure mechanisms you are trying to catch.
One trap: negative tip-up
Sometimes tan δ starts high and falls as voltage rises. That is almost never the insulation — it is a contact-resistance artifact. Too high a contact resistance at the measuring electrode reads as inflated loss at low voltage; as voltage climbs, more contact points ignite, the resistance drops, and the reading falls. The same thing can happen with an inner conductor shield whose contact to the strands improves under voltage. A negative tip-up is a signal to check your electrode contact, not to pass or fail the coil.
How the measurement is made
The voltage is raised step by step, typically in 0.2 UN intervals from 0.2 UN to rated voltage. At each step the voltage is held until the reading stabilizes — unstable readings usually mean a measurement problem, not the insulation. Tests are run at ambient temperature in air, because those conditions give the most reproducible picture of the insulation’s condition. Winding temperature, ambient temperature, and relative humidity all get recorded, because the tan δ-versus-voltage curve shifts with temperature.
Three instrument types do the job, all referenced to the high-voltage measuring requirements of IEC 60060:
- Schering bridge. The classic high-voltage bridge, balanced against a low-loss standard capacitor. It is sensitive to stray capacitance to earth, so it needs screened cabling and a Wagner earth circuit.
- Transformer ratio-arm bridge. Uses a current comparator around a magnetic core; it rejects stray-capacitance effects without needing the Wagner earth.
- Digital phase-shift systems. Computer-controlled, measuring the phase difference between the current through the test object and through a standard capacitor. With fiber-optic isolation, these can measure permanently earthed machines in the field.
The reference standard capacitor itself has to be near-lossless — tan δ below 0.01 × 10⁻³ across the test range — or it corrupts the measurement.
Guarding: essential for bars and coils
On an individual bar or coil, surface currents leaking off the end of the conductive slot coating will inflate the reading if you let them. So the standard requires guard ring electrodes to intercept those currents and keep them out of the measurement.
The slot coating is contacted along its full length — metal plates, wrapped foil, or spiralled wire — to simulate the slot and avoid a false high reading from poor electrode contact. A grounded guard ring is then placed near each end of the slot coating, either with a small temporary insulation gap (max 4 mm) or on top of the stress-control coating. After testing, any cut gap is repainted to restore the coating.
The catch: guarding is not practicable on a complete winding. That is the single biggest reason bar/coil results and complete-winding results are not directly comparable — the complete-winding number is influenced by the nonlinear stress-grading coatings, the end windings, and the surface condition, none of which a guard ring removes.
The limits — and why they’re debated
For new single bars and coils, tested guarded at room temperature up to a rated voltage of 21 kV, the standard gives maximum values:
| Characteristic value | Maximum |
|---|---|
| Initial tan δ at 0.2 UN (tan δ0.2) | 20 × 10⁻³ |
| Delta tan delta per 0.2 UN step (Δtan δ / 0.2 UN) | 5 × 10⁻³ |
| Tip-up (tan δ0.6 − tan δ0.2) | 5 × 10⁻³ |
Above 21 kV, limits are set by agreement. And the accuracy of your instrument has to be folded in — the standard says the spread of the computed characteristic values, including measurement uncertainty, must fall below the limit, not just the nominal reading.
Treat these numbers as guidelines, not hard verdicts. They are among the more contested figures in machine-insulation practice. Field studies across large populations of coils and bars — and work within CIGRE’s rotating-machines committee — have argued the limits are too restrictive for some insulation systems, and that windings exceeding them can still be reliable in service. The limits also only apply to the specific voltage steps defined; use different steps and they no longer hold. In practice, manufacturer-specific acceptance criteria and trend data on your own fleet often carry more weight than the table alone.
Bars and coils versus complete windings
The two are almost different tests wearing the same name.
Bars and coils are a manufacturing quality-control tool. You test a statistically meaningful sample from a production run, guarded, to prove the insulation is well cured, well impregnated, and consistent across the batch. The capacitance at 0.2 UN comes along for free and doubles as an impregnation check.
Complete windings are a condition-assessment tool. You cannot guard them, so the result folds in the stress-grading system and surface condition, but a new winding’s tan δ becomes the reference point for trending as it ages. Measure each phase separately and all phases together; with the other phases tied to the frame, you get the most information about both the slot portion and the end windings. A drift over years of service flags a change in the main insulation, the slot coating, the stress-control coating, or the end-winding support.
Where tan δ sits among winding tests
Dissipation factor is one of a handful of stator-winding insulation tests, and it answers a question the others don’t.
- Insulation resistance and polarization index run at DC and find moisture and contamination across the whole insulation. They say nothing about internal voids.
- Partial discharge testing detects and can localize individual discharge sites. tan δ measures the aggregate loss those discharges produce, without localizing them — the two are complementary, not redundant.
- Dissipation factor sits between them: an AC, whole-volume measure of dielectric loss that is especially good at void content, impregnation quality, and cure.
Run as a set on a new machine, they establish the baseline. Trended over a machine’s life, they build the condition picture that drives overhaul decisions.
FAQ
What is a dissipation factor test?
An AC measurement of the dielectric losses in stator winding insulation. It measures tan δ — the ratio of resistive to capacitive current — over a range of voltages, and derives characteristic values that indicate insulation quality and condition.
What is a good tan δ value for a stator winding?
IEC 60034-27-3 sets maximums for new guarded single bars and coils up to 21 kV: 20 × 10⁻³ for the baseline tan δ at 0.2 UN, and 5 × 10⁻³ each for the delta tan delta per step and the tip-up. These are debated as conservative, so manufacturer criteria and trend data often apply alongside them.
What does tan delta tip-up tell you?
The tip-up — tan δ at 0.6 UN minus tan δ at 0.2 UN — indicates void content and impregnation quality. A high tip-up means the insulation contains voids that begin discharging below rated voltage, typical of poor impregnation or delamination.
What is the difference between the baseline and the tip-up?
The baseline (tan δ at 0.2 UN) reflects the solid insulation — cure state, resin, contamination. The tip-up reflects voids and delamination that discharge as voltage rises. A high baseline with low tip-up points at the material; a normal baseline with high tip-up points at voids.
Why can’t a complete winding be guarded like a bar or coil?
Guard ring electrodes intercept surface currents at the ends of a bar’s slot coating, but a complete winding’s geometry doesn’t allow it. As a result, complete-winding readings are influenced by the stress-grading coatings, end windings, and surface condition, and are not directly comparable to guarded bar or coil results.
Can a dissipation factor test predict when insulation will fail?
No. Trending tan δ flags a change in insulation condition, but the standard is explicit that it cannot predict time to failure. It is a quality and condition indicator, used alongside other diagnostics.
Is dissipation factor the same as power factor?
They are physically different — tan δ uses the loss angle, power factor uses the phase angle — but numerically almost identical for modern windings, where tan δ stays well below 100 × 10⁻³. The terms are often used interchangeably.