--- title: "Insulation resistance test values: what passes, and when to reject" description: "What a megger reading means: IS 732 and IEEE 43 limits, temperature correction, polarisation index, and the three things that fake a failure." date: "2026-09-16" author: "Divakar B" source: "https://energycalchq.com/blog/insulation-resistance-test-values" license: "© 2026 Divakar B. Quote with attribution to https://energycalchq.com/blog/insulation-resistance-test-values" --- Every commissioning sheet has a column for it. A number goes in, somebody initials the box, and the file goes into the handover folder. Six months later a motor fails and the number in that box turns out to have meant nothing at all — because it was taken at the wrong voltage, at the wrong temperature, with the drive still connected, and nobody wrote down what the reading had been the year before. An insulation resistance test is cheap and fast, which is exactly why it gets done badly. This is what the reading actually measures, what the standards require you to accept or reject, and how to tell a genuine insulation fault from the three things that most often fake one. ## What the meter is actually measuring An insulation tester applies a DC voltage between a conductor and earth — or between two conductors — and measures the current that leaks across. Ohm's law gives you a resistance. The instrument is doing nothing cleverer than that. The leakage it sees is three currents in parallel, and they behave differently: - **Conduction current** through the bulk of the insulation. Small, steady, and the one you actually care about. It rises when the insulation is wet, contaminated or degraded. - **Surface leakage** across the outside of terminations and end windings. Driven by dirt, salt, carbon dust and humidity — not by the condition of the insulation itself. - **Absorption (polarisation) current** as the dielectric's molecules align with the applied field. Large at first, decaying over minutes. That third current is why the reading climbs while you watch it, and why a one-second reading is worthless. It is also the basis of the polarisation index, which is covered further down. A fourth current — capacitive charging — exists for the first few seconds and then disappears. On a long HV cable it is large enough to matter for safety; see the discharge warning below. ## Choosing the test voltage Too low and you learn nothing about insulation that will sit at working voltage. Too high and you can puncture insulation that was serviceable, which turns a test into a repair job. ![An insulation tester connected between one motor winding terminal and the frame earth bolt, with the copper shorting links removed and lying on the plinth beside the terminal box](https://hhfiunqrrmpuctcnlwnk.supabase.co/storage/v1/object/public/blog-images/insulation-test-motor-winding-to-earth.jpg) For fixed installations, IEC 60364-6 — adopted in India as IS 732 — sets both the voltage and the pass mark: | Circuit nominal voltage | DC test voltage | Minimum insulation resistance | | --- | --- | --- | | SELV and PELV | 250 V | 0.5 MΩ | | Up to and including 500 V, including FELV | 500 V | 1.0 MΩ | | Above 500 V | 1 000 V | 1.0 MΩ | For rotating machines, IEEE 43 sets it by machine rating rather than by circuit: | Machine rated voltage | DC test voltage | | --- | --- | | Below 1 000 V | 500 V | | 1 000 to 2 500 V | 500 to 1 000 V | | 2 501 to 5 000 V | 1 000 to 2 500 V | | 5 001 to 12 000 V | 2 500 to 5 000 V | | Above 12 000 V | 5 000 to 10 000 V | A 415 V motor gets 500 V DC. Not 1 000 V, however tempting the range switch is — the extra stress buys no extra information at that rating, and on an old random-wound winding it can find a weak spot and finish it off. ## The numbers that constitute a pass Here is where most sheets go wrong: **the 1 MΩ figure from IS 732 is an installation wiring limit, and applying it to a motor is far too generous.** A 415 V motor reading 1.2 MΩ has passed the wiring rule and is in serious trouble as a machine. IEEE 43 gives the minimum one-minute value, corrected to 40 °C: | Winding type | Minimum IR at 1 minute, 40 °C | | --- | --- | | Most windings built before about 1970, and most DC armature and field windings | (rated kV) + 1 MΩ | | Form-wound stator windings built after about 1970 | 100 MΩ | | Random-wound stators and form-wound coils rated below 1 kV | 5 MΩ | So the realistic floor for an ordinary 415 V squirrel-cage motor is **5 MΩ**, and for a modern 6.6 kV machine it is **100 MΩ** — not 7.6 MΩ, which is what the kV + 1 formula would have given you. The formula is the legacy case, not the default. In practice a healthy new LV motor reads in the hundreds of megohms to low gigohms. A serviceable 6.6 kV machine after years in a humid plant might sit around 500 MΩ. If a winding you know to be sound reads 8 MΩ, something is wet, not worn. ## Correct to 40 °C, or the number is fiction Insulation resistance roughly halves for every 10 °C rise in temperature. That is not a small correction — it is the difference between a pass and a reject. To refer a reading taken at temperature *T* to the 40 °C reference: ``` R₄₀ = R_T × 2^((T − 40) / 10) ``` | Winding temperature | Multiply the reading by | | --- | --- | | 20 °C | 0.25 | | 25 °C | 0.35 | | 30 °C | 0.50 | | 35 °C | 0.71 | | 40 °C | 1.00 | | 50 °C | 2.00 | | 60 °C | 4.00 | A motor tested cold on a 20 °C morning, reading 40 MΩ, is really **10 MΩ** at the reference temperature. Still a pass, but a much less comfortable one than the raw number suggested. The same machine tested hot straight off load at 60 °C, reading 15 MΩ, is 60 MΩ corrected — and perfectly healthy. Write the winding temperature on the sheet next to every reading. A reading without a temperature cannot be compared with anything, including its own history, which makes it the one thing an insulation test is most useful for. ## Polarisation index and DAR A single number tells you the condition today. The **shape of the curve** tells you whether the insulation is contaminated or merely old. **Polarisation index (PI)** is the ten-minute reading divided by the one-minute reading. Clean, dry insulation keeps absorbing charge, so the resistance keeps climbing and PI comes out high. Wet or contaminated insulation reaches its (low) steady value almost immediately, so PI approaches 1.0. IEEE 43 asks for a **minimum PI of 2.0** for thermal classes B, F and H, and 1.5 for class A. Two caveats that field engineers get wrong regularly: - **If the one-minute reading is above about 5 000 MΩ, PI is meaningless.** The currents involved are so small that instrument noise dominates the ratio. IEEE 43 says the test need not be performed in that case. A PI of 1.1 on a winding reading 12 GΩ is not a failure — it is a winding so dry there is nothing left to polarise. - PI is a ratio, so it needs no temperature correction — provided the temperature holds steady for the full ten minutes. On a machine cooling rapidly after shutdown, it does not, and the PI drifts high for the wrong reason. Test after the machine has settled. **DAR**, the dielectric absorption ratio, is the sixty-second reading over the thirty-second reading. It is the one-minute version for when you have not got ten. Below about 1.25 is poor, above about 1.6 is good. It is field practice rather than a standard requirement, so use it as a screen and follow up with a full PI where it looks marginal. ## A worked example A 55 kW, 415 V, random-wound motor on a cooling water pump. Annual test, with the cables lifted at the motor terminal box. - Test voltage: **500 V DC**, from the IEEE 43 table. - Winding temperature, from the RTD: **30 °C**. - Reading at 30 seconds: 210 MΩ. At 60 seconds: **340 MΩ**. At 10 minutes: **760 MΩ**. Work it through: - Corrected one-minute value: 340 × 0.50 = **170 MΩ at 40 °C**. Against a 5 MΩ floor for a random-wound winding, that is a comfortable pass. - DAR: 340 ÷ 210 = **1.62**. Good. - PI: 760 ÷ 340 = **2.24**. Above the 2.0 requirement. Now the part that matters. Last year the same motor, at the same terminals, gave 620 MΩ corrected. This year it gives 170 MΩ. Both readings pass every limit in every table above — and the machine has lost roughly three quarters of its insulation resistance in twelve months. **That trend is the finding.** A number that passes today and is falling by a factor of three or four a year will not pass in two years, and the failure will happen at whatever moment is least convenient. This is why IEEE 43 is explicit that the trend over time carries more weight than any single reading, and why a test record that is not comparable year on year has wasted the time it took to collect. Suspect moisture ingress at the terminal box gland or a failing shaft seal letting water onto the windings. Check the gland and the drain plugs before the next run, and repeat the test in three months rather than waiting a year. ## Three things that fake a failure Before condemning a winding or a cable, rule these out. In the field they account for most low readings. **1. Something is still connected.** Variable speed drives, soft starters, surge protective devices, RCDs, electronic ballasts, PLC input cards and power factor capacitors all present a deliberate low-impedance or non-linear path to earth. They will pull a reading down to a few hundred kilohms and read as a catastrophic fault. Worse, the test voltage can destroy them: a 500 V or 1 000 V DC impulse across a drive's DC bus or an SPD's metal-oxide varistor is outside anything they were designed for. Disconnect them. If a device genuinely cannot be isolated, IEC 60364-6 permits measuring line and neutral bonded together against earth — which protects the device, at the cost of no longer testing between conductors. **2. The neutral-earth bond is still in.** Test an installation with the link at the main earth bar still made and you are measuring the resistance of your own earthing system, which is by design close to zero. Lift the link, test, and put it back — and put it back before energising, every time. **3. Surface contamination and humidity.** Salt-laden coastal air, cement dust, carbon brush dust and simple condensation form a conductive film over terminations. The insulation underneath is intact; the surface is not. The signature is a low reading with a **poor PI, under about 1.5**, together with a reading that improves markedly after cleaning and drying the terminations. Clean them, allow them to dry, and retest before ordering a rewind. A fourth, less common: on a long HV cable run, a reading that is genuinely capacitive-dominated will still be climbing at ten minutes. That is the cable's size, not a fault. ## Safety: discharge before you touch anything The test charges the circuit capacitance to the test voltage, and long cables hold a serious amount of energy. Modern testers discharge automatically when the test ends, but this is not something to take on trust. Leave the leads connected after the test and allow a discharge time of **at least four to five times the duration for which the voltage was applied.** A ten-minute PI test on a cable therefore needs the better part of an hour, or a verified discharge with an earthing stick and a confirming voltage check. Then prove dead with an approved tester before you touch the conductors. Insulation testers are also live-circuit hazards in the other direction. Verify the circuit is isolated and dead before applying the test voltage — most instruments will refuse to test into a live circuit and tell you so, but the warning is a backstop, not a procedure. ## What to do when a reading genuinely fails A low reading is a symptom, and the useful next step depends on where the circuit is low: - **Whole installation low, individual circuits fine.** The fault is in the common conductors or in something you have not disconnected. Split the board and halve your way to it. - **One circuit low, all phases equally.** Suspect water in a duct, a buried cable damaged by later civils work, or a flooded junction box. A reading that is identical on all three phases points at a shared environment, not a conductor fault. - **One phase low, other two healthy.** A single-point defect: a nicked core at a gland, a crushed cable, a failed termination. Localise before you replace — the phase that reads low tells you which core to follow. - **Motor low, cable healthy when tested separately.** The winding is the problem. Get the temperature correction and the PI before you commit to a rewind — a contaminated but sound winding cleans and dries, and a rewind on a large machine is weeks of lead time. Always test the cable and the machine separately. Testing them together and condemning the expensive item is a common and costly mistake. ## Frequently asked questions **What is a good megger reading?** For 415 V installation wiring, anything at or above 1 MΩ passes IS 732 and IEC 60364-6, though healthy new wiring reads far higher — typically tens to hundreds of megohms. For a 415 V motor, the floor is 5 MΩ at 40 °C under IEEE 43, and a sound machine will normally read in the hundreds of megohms. **Should I test at 500 V or 1 000 V?** Use 500 V DC for anything rated up to 500 V, including 415 V motors and installation circuits. Use 1 000 V for circuits rated above 500 V. Going higher than the standard requires adds no information and risks damaging serviceable insulation. **What is an acceptable polarisation index?** 2.0 or above for thermal class B, F and H insulation; 1.5 for class A. If the one-minute reading exceeds roughly 5 000 MΩ, disregard the PI — the ratio is dominated by noise at those currents. **Can I megger a circuit with a VFD connected?** No. Isolate the drive at its terminals first. The DC bus capacitors and the input filtering present a low-impedance path that will read as a fault, and the DC test voltage can damage the drive's semiconductors outright. **Why did my reading go up after I cleaned the terminals?** Because the original reading was surface leakage across contamination, not insulation degradation. A low reading with a PI under 1.5 that recovers after cleaning and drying is the standard signature of that, and it does not mean the insulation needs replacing. **How often should insulation resistance be tested?** Annually for most industrial plant, and at every commissioning and every re-energisation after a shutdown. Increase the frequency to quarterly for any machine whose corrected reading is falling faster than about half per year, regardless of whether it still passes. ## The point of the number An insulation test is one of the few pieces of condition data you can take in five minutes, on a machine you cannot open, that predicts a failure before it happens. That only works if the record is comparable — same test voltage, same terminals, same duration, temperature written down, and the trend plotted rather than filed. A single reading tells you whether to energise today. The series tells you when the machine is going to fail.