Most people who run a Polarization Index test can tell you the formula: 10-minute IR divided by 1-minute IR. Fewer can tell you why that ratio is meaningful, where it stops being meaningful, or what a result of 1.1 on a modern epoxy-mica winding actually implies — compared to the same reading on an asphaltic machine from 1965.
That gap matters. The PI is one of the most widely used diagnostic tests in motor maintenance, and it is also one of the most frequently misread. This article covers the physics behind the test, what each result range actually tells you, and where the test has hard limits that no amount of correct procedure can fix.
Table of Contents
What the PI Measures — and What It Doesn’t
The PI is not a direct measurement of insulation condition. It is a measurement of how much the total leakage current changes between 1 minute and 10 minutes of applied DC voltage.
When DC voltage is applied to a winding, several currents flow simultaneously. The capacitive current is large at the start and decays to essentially nothing within the first 10 seconds. After that, three currents remain:
Absorption current arises from molecular polarization inside the insulation. Polar molecules in the epoxy, mica, polyester, and asphalt components of the groundwall reorient in the DC electric field. Electrons become trapped at the interfaces between layers of different materials. Both processes require current from the supply while they are happening, and both decay over time as the system approaches equilibrium. In modern epoxy-mica insulation, this current falls to essentially zero within about 10 minutes. In older asphaltic systems, the decay is slower and may extend well beyond 10 minutes.
Surface leakage current flows across the outer surface of the insulation along the end turns. It is driven by semi-conductive contamination — oil, carbon dust, moisture, salt, chemical deposits. This current is constant with time. It does not decay.
Conduction current flows through the bulk of the insulation itself. In modern epoxy-mica insulation, this current is effectively zero as long as the insulation is intact. In older asphaltic systems, and in any insulation that has absorbed moisture into its bulk, this current is measurable and constant with time.
The Polarization Index uses the time behaviour of the total current to separate these components indirectly. If the total current drops significantly between 1 minute and 10 minutes, the absorption current is dominant — meaning the leakage and conduction currents are small relative to it. The insulation is probably clean and dry. If the total current stays approximately flat, the surface leakage or conduction current is dominant — meaning contamination or moisture is present in quantities that matter.
The PI quantifies this:
PI = IR₁₀ ÷ IR₁
A PI near 1.0 means the IR barely changed between 1 and 10 minutes. The current was flat. Leakage or conduction current is dominating. Something is wrong with the surface or bulk condition of the insulation.
A PI above 2.0 means the IR increased substantially. The current was still declining at 10 minutes. Absorption current is dominant. The insulation is behaving like clean, dry, intact insulation should.
Why PI Is Less Temperature-Sensitive Than IR Alone
Raw IR readings are strongly dependent on temperature. A 10°C rise in winding temperature can reduce IR by five to ten times in thermoplastic insulation systems. This makes trending IR across tests taken at different temperatures difficult — the temperature correction factors in IEEE 43 are approximations, and their reliability diminishes as the measurement conditions diverge further from the calibration range.
The PI largely cancels this problem. Both the 1-minute and 10-minute readings are taken at the same winding temperature — the test only lasts 10 minutes, which is not long enough for the winding to change temperature significantly. Whatever temperature correction factor applies to IR₁ also applies to IR₁₀. When you divide one by the other, the correction factor cancels.
This means PI values can be meaningfully compared across tests taken at different ambient conditions, where a direct comparison of IR₁ readings corrected to 40°C might be unreliable. It is one reason the PI is often more useful than a single IR reading for trending purposes, particularly on machines where controlling test temperature is difficult.
The one exception: a machine tested immediately after being shut down from a high operating temperature. If the winding is cooling rapidly during the 10-minute test, the IR may rise between 1 minute and 10 minutes due to temperature drop rather than purely absorption current behaviour. The result is an artificially high PI. If you suspect this is happening, let the machine cool below 40°C and repeat.
Acceptance Thresholds
IEEE 43-2013 sets the minimum acceptable PI based on the thermal class of the insulation:
| Insulation Thermal Class | Minimum PI |
|---|---|
| Class 105 (A) | 1.5 |
| Class 130 (B) and above | 2.0 |
Most motors manufactured after the mid-1960s use Class B (130°C), Class F (155°C), or Class H (180°C) insulation. For all of these, the minimum PI is 2.0.
The threshold of 2.0 is derived from field experience. When the PI is below 2.0, the leakage and conduction currents are large enough relative to the absorption current that electrical tracking on the end turns becomes a real risk. When the PI exceeds 2.0, experience shows that contamination-driven tracking is unlikely.
A PI result and an IR₁ result should always be interpreted together. IEEE 43 states that a winding must satisfy both the minimum IR₁ threshold (corrected to 40°C) and the minimum PI to be cleared for energisation or high-potential testing. Passing one but failing the other is not sufficient.
Interpreting the Full Range
PI below 1.0 is a serious flag. The 10-minute IR is lower than the 1-minute IR. This is not physically impossible — it can happen if the winding temperature is rising during the test, or if there is an instrumentation problem. If neither of those applies, severely degraded insulation or a massive contamination problem is the likely explanation.
PI of 1.0 to 1.5 indicates that surface leakage or conduction current is dominating. The winding is wet, contaminated, or both. This does not necessarily mean the insulation is permanently damaged — many windings at this PI level respond well to cleaning and drying. The insulation itself may be intact; the test is simply reflecting the surface condition.
PI of 1.5 to 2.0 is borderline. The result is marginal relative to the IEEE 43 minimum. On an older thermoplastic winding, this may indicate that the insulation is in reasonable condition with some surface contamination. On a modern epoxy-mica winding, anything below 2.0 warrants investigation, since a clean modern winding should comfortably exceed this threshold. Context — machine history, recent environment, whether the winding has been idle — matters heavily at this level.
PI of 2.0 to 4.0 is acceptable. The absorption current is dominant. The surface and bulk conditions are unlikely to be driving immediate failure risk.
PI above 4.0 on a modern epoxy-mica winding is normal and expected. These insulation systems have high absorption current and low surface leakage when clean and dry. Readings in the range of 4 to 10, or higher, are routine.
PI above 6 on an older thermoplastic (asphaltic-mica) winding is a specific case requiring a different interpretation. As asphaltic insulation ages thermally, the asphalt component changes character — it may flow out of the groundwall under heat, leaving the structure more porous and the insulation more brittle. Thermally aged asphaltic insulation sometimes shows very high PI values because the molecular structure has fundamentally changed and the absorption current behaviour shifts. Stone et al. note that a PI above 6 on an older stator may indicate thermal deterioration of the insulation rather than good condition. Physical inspection — tapping the insulation to check for brittleness, looking for surface cracking or delamination — is warranted when this is observed.
The Hard Limits: Where PI Gives No Useful Information
When IR₁ exceeds 5000 MΩ. IEEE 43-2013 states this directly: when the 1-minute IR reading is above 5000 MΩ, the PI should not be used as a diagnostic tool. At this resistance level, the total current flowing is in the sub-microampere range. At that level of sensitivity, small fluctuations in supply voltage, ambient humidity, surface condition of the test leads, and even proximity to metallic objects can cause measurable changes in the recorded current. The difference between the 1-minute and 10-minute readings can be dominated by these noise sources rather than by genuine absorption current behaviour. The PI calculation will produce a number, but that number is not a reliable indicator of insulation condition. If you encounter this situation, the appropriate response is to note the very high IR₁ as a positive finding and not calculate or interpret the Polarization Index.
This situation is common with modern form-wound stator windings in good condition. A clean, dry, modern epoxy-mica winding at room temperature may read several hundred MΩ to several GΩ. At that level, stop the test at 1 minute, record the IR₁, and move on.
Non-insulated field windings. Large turbine generators and some salient pole machines use field windings where the copper conductors are not individually encapsulated in insulation. The conductors are isolated from ground through insulating strips and barriers, but the copper surface area exposed to those surfaces is enormous. Surface leakage current (I_L) dominates the measurement completely. The absorption current component is negligible by comparison. The IR barely changes between 1 and 10 minutes regardless of winding condition — PI will read close to 1.0 for a machine in perfect condition. Using this result as a diagnostic number is meaningless.
DC machine armatures with exposed commutators. The commutator copper surface creates the same problem as bare field windings — excessive surface leakage current swamps the absorption current. PI is not applicable.
Machines with surge capacitors still connected. A surge capacitor with an internal discharge resistor — typically around 10 MΩ — connected to the motor terminals during the test will present that resistor in parallel with the winding insulation. The resistor draws constant current that is large relative to the absorption current of the winding. The IR will appear low and flat over time. PI reads near 1.0 regardless of actual winding condition. Always disconnect and ground surge capacitors before testing.
Testing from the switchgear without isolating cables. When the motor cannot be isolated at its terminals, the IR and PI will reflect a combination of the cable insulation and the motor winding insulation in parallel. If the result is good, both are probably acceptable. If the result is poor, the test must be repeated at the motor terminal box with cables disconnected to determine which component is responsible.
Continuous stress control coating across the entire endwinding. Some generator designs apply stress control material across the entire endwinding overhang in a way that makes electrical contact with the bare copper at the ends of the bars. The resulting surface leakage current is very large relative to the absorption current, and the Polarization Index will read close to 1.0 even on a machine in perfect condition. This is a design characteristic of those specific windings, not a fault. IEEE 43-2013 identifies this as a situation where the PI threshold values do not apply.
Small random-wound motors. In random-wound stator windings, the groundwall insulation is relatively thin and the absorption current decays much faster — often within 2 to 3 minutes rather than 10 minutes. The standard 10-minute test may show very little difference between IR₁ and IR₁₀ not because the winding is contaminated, but because the absorption process is essentially complete long before 10 minutes. IEEE 43 acknowledges this and describes shorter-interval variants of the PI — such as IR₁/IR₃₀ₛ or IR₅/IR₁ — as alternatives. However, no standardised acceptance thresholds exist for these variants.
Rotors vs Stators
Rotor windings behave differently from stator windings in IR and PI testing, and the same numerical thresholds should not be applied without adjustment.
Round rotor field windings and salient pole windings with strip-on-edge construction often have significant exposed copper surface area. This drives surface leakage current up and keeps PI values lower than you would expect from a comparably conditioned stator. Rotor windings also tend to have thinner insulation and more complex geometry, which reduces the ratio of absorption current to leakage current.
As a practical matter, rotor PI values tend to run lower than stator PI values even on healthy machines. Historical comparison within the same machine is more useful than applying a universal threshold. The IEEE 43 minimum PI values apply to rotor windings with insulation encasing the conductors — they are not applicable to non-insulated field windings as noted above.
Carbon dust accumulation around slip rings is a common cause of low IR and PI on rotors. This is the first location to check and clean before interpreting a poor rotor result as evidence of insulation damage.
The Dielectric Absorption Ratio
The DAR is a faster variant of the PI using 30-second and 60-second readings instead of 1-minute and 10-minute readings:
DAR = IR₆₀ₛ ÷ IR₃₀ₛ
It produces a result faster than the full PI test and requires less discharge time afterward. The trade-off is sensitivity — the shorter interval captures less of the absorption current decay, so the ratio has less contrast between clean and contaminated insulation.
The DAR is used in situations where a 10-minute test is not practical: large numbers of machines to screen in limited time, confined working conditions, or maintenance windows too short for the full PI procedure. It is a screening tool. A poor DAR result warrants a full PI test before any decision is made. A good DAR result provides reasonable assurance but is less definitive than a full PI at the IEEE 43 thresholds.
No DAR-specific acceptance values appear in IEEE 43-2013. Common practice uses approximately 1.25 as a marginal threshold and 1.6 as indicating acceptable condition, but these are field conventions derived from experience with the PI thresholds, not values defined in the standard.
Testing Procedure
The PI test requires no additional equipment or connections beyond what the IR₁ test already uses. The megohmmeter applies voltage and the same test session covers both results.
Apply the DC test voltage appropriate for the machine’s rated voltage (from IEEE 43 Table 1). Take the IR reading at exactly 1 minute. Continue applying voltage and take the second reading at exactly 10 minutes. The PI is IR₁₀ ÷ IR₁.
To improve measurement accuracy around the 1-minute point — where the transition from absorption current to the more stable region is steepest — IEEE 43 recommends taking readings at multiple intervals: 15 seconds, 30 seconds, 45 seconds, 1 minute, 1.5 minutes, 2 minutes, 3 minutes, 4 minutes, and continuing to 10 minutes. This allows the data to be plotted on a log-log scale, which shows the characteristic shape of the IR curve and makes it easier to identify abnormal behaviour that a two-point ratio alone might not reveal.
After the test, remove the high-voltage lead and immediately ground the winding through a discharge resistor. The minimum grounding time after a PI test is 40 minutes — four times the 10-minute test duration. Do not lift the ground until return voltage is below approximately 20 V. Residual charge after a 10-minute application of high DC voltage can persist for hours on large windings and constitutes a genuine shock hazard.
What the Test Cannot Find
The PI and IR tests share the same fundamental limitation: they are sensitive to surface contamination and moisture, and to severe cracking that provides a continuous conductive path from conductor to ground. They are not sensitive to internal defects within the bulk of the insulation.
In modern epoxy-mica form-wound windings, mica has essentially infinite DC resistivity. A void or delamination within the groundwall — from thermal deterioration, inadequate VPI impregnation, or thermal cycling — is electrically invisible to DC testing. The current does not flow through the void; it flows around it. The IR and PI will appear normal while a significant internal defect is present.
This is not a failure of the test. It is a fundamental physical characteristic of DC testing on mica-based insulation. Detection of internal bulk degradation in modern form-wound stators requires AC-based testing: power factor and capacitance tip-up testing, or offline partial discharge testing at rated voltage. The PI should be understood as a surface and contamination test — thorough and reliable for what it detects, blind to what it does not.
Documentation
Every PI test record should include: date, machine identification, winding temperature, ambient temperature and relative humidity, test voltage, IR₁ reading, IR₁₀ reading, PI value, and whether all external equipment was disconnected. Note anything unusual — recent rainfall, recent process contamination event, whether the machine was tested warm or cold.
Trend the PI alongside IR₁ corrected to 40°C. A PI that is declining over successive tests — even if it remains above 2.0 — is a machine that is moving in the wrong direction. A PI that is stable or improving is a machine that is being maintained well.
Quick Reference
| Parameter | Value |
|---|---|
| Formula | PI = IR₁₀ ÷ IR₁ |
| Minimum PI — Class A insulation | 1.5 |
| Minimum PI — Class B and above | 2.0 |
| PI not applicable | IR₁ > 5000 MΩ; non-insulated field windings; exposed commutators; some stress-control coatings |
| PI may be unreliable | Winding temperature dropping rapidly during test |
| PI close to 1.0 on modern winding | Surface contamination or moisture — not necessarily permanent damage |
| High PI (>6) on old asphaltic winding | May indicate thermal deterioration — inspect physically |
| Test duration | 10 minutes voltage application |
| Post-test discharge time | Minimum 40 minutes (4× test duration) |
| DAR formula | IR₆₀ₛ ÷ IR₃₀ₛ |
| DAR use case | Fast screening when full PI not practical |