Tan Delta vs Partial Discharge Testing: Which to Use

Both tan δ and partial discharge testing react to the same underlying defect — voids in stator ground-wall insulation that discharge under voltage. So it is easy to assume they are two ways of doing one job, and to pick whichever your lab already owns. They are not interchangeable. They answer two different questions, and the difference decides which one you need.

The short version: tan δ tells you how much the whole insulation is losing; partial discharge tells you where the loss is coming from. This page is the decision. For each test on its own, see the dissipation factor pillar and partial discharge testing.

The one distinction that matters

Tan δ is a bulk measurement. It reports the average dielectric loss of the entire insulation volume between the conductor and the slot coating, as a single number at each voltage. IEC 60034-27-3 is explicit that the measurement gives no indication of how loss is distributed within the insulation, and — unlike an off-line partial discharge measurement — does not permit localization of weak points.

Partial discharge testing is a site measurement. It detects individual discharge events and, off-line, can locate where along the winding they originate and characterize what kind of defect is producing them.

So the same void population shows up in both, but differently. In tan δ it raises the aggregate loss and the tip-up. In PD it appears as discrete pulses you can count, trend, and place. One gives you a whole-winding health number; the other gives you a map.

What each one catches that the other misses

Tan δ catches uniform, distributed problems that PD can under-report. A winding that is globally poorly cured, poorly impregnated, or thermally aged has elevated loss everywhere. That raises tan δ cleanly. It also catches the solid-material story — cure state, resin, contamination, moisture — through the low-voltage baseline, which is not a discharge phenomenon at all and does not show up as PD.

PD catches concentrated, localized problems that tan δ can average away. A single bad slot, one delaminated coil, or a specific end-winding discharge site is a small fraction of the total insulation volume. Its contribution to the whole-winding average loss may be modest, so tan δ can look acceptable while a real, localized weak point exists. PD finds that site because it is looking at events, not averages.

That asymmetry is the practical case for not treating them as substitutes. A good tan δ result does not prove the absence of a dangerous local defect, and a PD hit does not tell you whether the bulk insulation is sound.

The practical differences

Beyond what they measure, the two differ in how and when you run them.

• Tan δ is inherently off-line and stepwise. You de-energize, apply a controlled AC voltage in steps from a low value to rated, and read the loss curve. It is a bounded, repeatable measurement well suited to acceptance testing of new bars and coils.
• PD can be off-line or on-line. Off-line PD, like tan δ, uses an applied voltage and can localize. On-line PD monitors the machine in service through permanently installed couplers, catching discharge behavior under real operating stress and temperature — something neither tan δ nor off-line PD sees. The trade-off between the two is its own decision, covered in online vs offline partial discharge testing.
• PD produces patterns, tan δ produces a curve. PD data is often read as phase-resolved patterns that fingerprint the discharge mechanism — slot, surface, internal void — as described in PD PRPD patterns. tan δ produces three characteristic numbers and a loss-versus-voltage curve. Different data, different interpretation skills.

When to use which

For a clean decision, match the test to the question:

• Accepting a batch of new coils or bars? Tan δ. It is built for manufacturing quality control — uniform cure, consistent impregnation, statistical sampling against limits. It is bounded, repeatable, and quick per specimen.
• Hunting a suspected local defect? Off-line PD. When you need to know which coil or slot, tan δ’s average will not point you there.
• Watching a machine age in service? Both, trended. On-line PD tracks discharge behavior under real conditions between outages; tan δ during an outage tracks the bulk insulation’s drift from its baseline.
• Reading the solid-material condition — cure, contamination, moisture? Tan δ’s low-voltage baseline. This is not a discharge signal, so PD does not see it.

Why the answer is usually “both”

Neither test predicts time to failure, and neither sees the whole picture alone. The standard framing is that tan δ assesses the dielectric behavior of the insulation as a whole, while PD resolves the individual sites. Run as a pair, they cover each other’s blind spots: tan δ confirms the bulk insulation is sound and uniform, PD confirms no dangerous local site is developing, and both feed a trend that drives overhaul timing.

On a condition-assessment program for a fleet of machines, the sensible pattern is a baseline of both on commissioning, on-line PD for continuous watch, and tan δ plus off-line PD at outages. The two are complementary diagnostics, not competing ones — which is why the machine-insulation standards treat them as parts of the same 60034-27 family rather than alternatives. They also sit alongside DC tests like insulation resistance and PI, which catch moisture and contamination the AC tests interpret differently.

FAQ