IEC 60034-27-3 covers two very different test objects under one method: individual bars and coils, and complete stator windings. They use the same tan δ measurement, but they answer different questions, run with different setups, and get judged by different rules. Treating a complete-winding result like a bar result — or comparing the two directly — is a common mistake.
The split comes down to one job each. Bars and coils are a manufacturing quality-control test. Complete windings are a condition-assessment tool. Everything else follows from that. This builds on the dissipation factor pillar.
Two objects, two jobs
Individual bars and coils — proving the build. You test a statistically meaningful sample from a production run to confirm the insulation is well cured, well impregnated, and consistent across the batch. The setup is controlled: guard ring electrodes in place, the slot coating contacted along its full length, each object conditioned before measurement. Because the setup is repeatable, the results can be checked against the Table 1 acceptance limits and compared coil-to-coil.
Complete windings — tracking the machine. Here there is no acceptance table. You measure the whole winding — before assembly, where the winding is addressed on its own, or after assembly, where bushings and other components also influence the reading — to set a baseline on a new machine and then trend it over the machine’s life. A shift in tan δ over years of service flags a change somewhere in the system: the main insulation, the conductive slot coating, the stress-control coating, the end windings, or the slot support.
| Bars & coils | Complete windings | |
|---|---|---|
| Purpose | Manufacturing QC | Condition assessment / trending |
| Guarding | Guard ring electrodes | Not practicable |
| Acceptance limits | Table 1 applies | None — trend instead |
| What’s measured | Individual object, guarded | Whole winding, per phase and all phases |
| Compared against | Limits and other coils | The machine’s own baseline |
Why the two aren’t comparable
The single biggest reason a complete-winding number can’t be read like a bar number: guarding is not practicable on a complete winding. Without a guard ring to strip out the end effects, the result is considerably influenced by the nonlinear resistance of the stress-grading coatings relative to the length of the slot portion — influences a guarded bar test removes. So a complete winding will generally read differently from the individual coils that went into it, and the difference is the setup, not a change in the insulation.
That is also why complete windings have no Table 1 limits. The stress-control coating, surface condition, and end windings all fold into the number, and no single population limit can account for them. Condition is judged by how the winding drifts from its own baseline, not by where it sits against a table. A consistent upward trend is the trigger for a visual inspection and further diagnostics — not a fixed threshold.
Measuring a complete winding in practice
The complete-winding test carries setup requirements the bar test doesn’t, and most of them are about getting a clean, safe, comparable reading.
Measure each phase and all phases. For an individual phase, the other phases are connected to the stator frame. Measuring each phase separately and all together gives the fullest picture of both the slot portions and the end windings.
The grounding configuration is the crux. A complete machine’s core and frame are solidly grounded. Some measuring instruments can work with a grounded test object; some cannot. If the instrument can’t, the core or frame ground has to be disconnected — which is feasible in a factory or repair shop, but not on an installed machine. On an installed machine, the supply and instrument are therefore operated ungrounded, which is a job for personnel with the right training and experience, for safety reasons. (This is the situation the fiber-optic digital systems in how tan δ is measured are built for.)
Isolate the winding from everything else. Connected equipment — cables, bus-ducts, instrument transformers — is disconnected so it doesn’t disturb the reading. The rotor, bearings, slot RTDs, and instrument screening are securely bonded to the frame so they sit at the same potential during the test.
Hydrogen-cooled generators get tested in air. Measure under a normal air atmosphere, not under operating hydrogen pressure. Hydrogen pressure suppresses partial discharge activity in voids, which would hide exactly the delamination trend you’re testing for.
What a complete-winding result can’t do
A complete-winding tan δ is a whole-insulation average, so it does not pinpoint a local defect. If a measurement returns a high power factor or tip-up, you can try to narrow it down by subdividing the winding into individual parallels or coil groups to gauge the extent of the damage — though that subdivision often isn’t practical. The better route to localizing weak points is partial discharge testing, off-line or on-line, which is designed to find individual discharge sites that tan δ averages together.
That limitation is not a weakness of the complete-winding test so much as a reminder of what it is for: it tells you the bulk insulation is drifting, and PD tells you where. Run together, they cover the machine.
FAQ
What’s the difference between tan delta on bars/coils and on complete windings?
Bars and coils are tested guarded, as a manufacturing quality-control check against acceptance limits. Complete windings can’t be guarded, have no limits, and are measured to set a baseline and trend the machine’s condition over time. The two results aren’t directly comparable.
Why can’t complete windings be guarded?
The standard states guarding techniques are not practicable for complete windings. Without a guard ring, the reading is influenced by the nonlinear stress-grading coatings relative to the slot length — effects a guarded bar test removes — so a complete winding reads differently from its individual coils.
Are there acceptance limits for complete-winding tan delta?
No. The stress-control coating, end windings, and surface condition all affect the reading, so no fixed table applies. Complete windings are judged by trending against their own baseline; a consistent upward trend triggers inspection and further diagnostics.
How do you measure tan delta on an installed machine that stays grounded?
If the instrument can’t work with a grounded object and the ground can’t be disconnected — as on an installed machine — the supply and instrument are operated ungrounded, which requires appropriately trained personnel. Digital systems with fiber-optic isolation are designed for this case.
Why test hydrogen-cooled generators in air rather than under hydrogen?
Hydrogen pressure suppresses partial discharge activity in voids. Measuring under normal air lets those discharges show up, so the test can reveal the delamination and ageing trend you’re looking for.
Can complete-winding tan delta find a local fault?
Not directly — it averages the whole insulation. You can subdivide the winding into parallels or coil groups to narrow it down, but that’s often impractical. Partial discharge testing is the tool for localizing individual weak points.