What an Insulation Resistance Test Actually Tests

Why this matters

An insulation resistance test is one of the few instruments a field tech carries that generates its own hazard on purpose. It applies a test voltage far above what your multimeter puts out, it charges everything it touches, and it hands back a single number in megohms that gets written on a ticket and treated as a verdict on the equipment.

It is not a verdict on the equipment. It is a leakage measurement across whatever was electrically in the circuit at the moment the button was held, at one applied voltage, at one temperature. The distance between those two readings of the same number is where the expensive mistakes live: a motor that passed and failed on the next start, and a motor that was condemned for a fault that lived in the cable.

Before the test set comes out of the bag

The instrument is a source of hazardous voltage. Everything below is what you do first, in order.

Isolate and lock. This is dead-circuit work on an electrical installation, so the controlling standard is 29 CFR 1910.333(b)(2), not 29 CFR 1910.147, which expressly carves out exposure to electrical hazards from work on conductors and equipment in electric utilization installations at 1910.147(a)(1)(ii)(C). Open the disconnect, apply your lock and tag, and verify the absence of voltage with a tester you proved live, then dead, then live again on a known source (NFPA 70E-2021, 120.5, which binds you through your employer's electrical safety program or your contract in whatever edition that program has adopted).

Control both ends. A cable run has a far end, and someone standing there can be contacted through the conductor you are energizing. Guard or barrier both ends, and post a person where you cannot see the far end yourself.

Disconnect anything with electronics in it before you apply test voltage. Drives, soft starters, control boards, capacitors, surge suppressors and electronic overloads are destroyed by a test voltage they were never built to withstand. This is a hazard your own command creates: the instruction to press the button is the instruction that kills the drive. Land the leads on the motor terminals with the machine leads lifted, not on a panel that still has the drive connected.

Discharge and verify before you touch anything. The test charges winding and cable capacitance and that charge does not leave when the reading ends. Let the instrument's discharge circuit work, then apply a grounding device and confirm zero before hands go anywhere. On a long run or a large machine the stored charge is enough to hurt you well after the display goes blank.

What the instrument is actually doing

It applies a DC voltage between the conductor and ground and measures the tiny current that flows. Then it divides: resistance is the applied voltage divided by the measured current. That is the entire mechanism. The megohm number on the display is arithmetic performed on a leakage current in the microamp range.

Two consequences follow immediately, and both matter more than the number itself.

First, the result belongs to everything that was in the circuit, because parallel leakage paths add in conductance. Two paths of 156 megohms and 13 megohms in parallel do not average; their conductances add and the pair reads about 12 megohms, dominated by the worse one. The instrument cannot tell you how many paths it found or where they were.

Second, the current it reads is three currents added together, and they behave differently over time. Capacitive charging current is large at first and dies away in seconds as the geometry charges. Absorption current, from the slow polarization of the insulating material, decays over a much longer span, typically minutes. True leakage current, the conduction through and across the insulation, is the only one of the three that is steady and the only one that is telling you about condition. This is why the reading climbs while you hold the button, and why a reading taken at ten seconds and a reading taken at sixty seconds on the same machine are two different measurements, not one measurement taken twice.

What the test is genuinely good at

It finds distributed problems in the ground wall: moisture, contamination, carbon tracking across a surface, an insulation system that has absorbed water, a path to ground through a wet terminal box or a chafed conductor. These share a shape. They are conductive paths from a conductor to ground that exist at low voltage and get worse with wetness and dirt, and the test drives current down exactly those paths.

It is also a good comparator against itself, at the same voltage, duration and corrected temperature. The sibling article on reading an insulation resistance result over time owns that method.

What the test cannot see, and why the blindness is structural

A turn-to-turn fault. Two adjacent turns of the same winding sit at nearly the same potential with respect to ground. A DC test between the winding and ground puts almost no voltage across the insulation between them, so almost no current flows through the fault, so the instrument does not see it. The current that destroys a shorted turn is driven by transformer action once AC excitation is applied, which is a condition the test never creates. A machine with a shorted turn can read at the top of the instrument's scale and fail on the next start. Turn insulation is a different question and it needs a different instrument, typically a surge comparison test, which is not a hand-tool measurement.

A weakness that only breaks down above the test voltage. A void, a crack or a thin spot that holds at the applied stress and punctures at operating peak returns a passing number. The result is a statement about behavior at the test voltage and nothing more. Raising the test voltage to chase this belongs to the manufacturer's service literature or the governing specification, because the wrong test voltage damages sound insulation.

Anything that only appears under duty. Heat, vibration, centrifugal force on end turns, thermal expansion opening a crack. The machine is stationary and cold when you test it, so a fault that requires it to be hot and turning is outside the test's reach by construction.

Where the leakage is. The reading is one number for all paths in parallel. Locating requires disconnecting and re-testing the pieces separately, which is a procedure, not a reading.

The gate this card is built on

A passing insulation resistance result licenses exactly one conclusion: that the total leakage of everything in the test circuit, at the applied test voltage, at that temperature, at that moment, met the acceptance figure in the governing document. It licenses no conclusion about turn insulation, about behavior above the test voltage, or about which part of the test circuit produced the reading.

The acceptance figure itself is not something to invent. It comes from the equipment manufacturer's service literature, or from the consensus standard your specification calls out (IEEE 43 for rotating machinery, in the edition your specification or the machine's manufacturer references, which binds through that document rather than on its own authority). What follows runs that one gate against two jobs that resolve in opposite directions.

Outcome one: the reading passed and the machine was still bad

A three-phase pump motor tripping its overload on start, intermittently. Locked out, leads lifted, tested at the voltage the manufacturer's literature specifies for that machine. Reading was at the top of the instrument's scale on all three phases to ground, steady for a full minute, no downward drift.

By the gate, that result says the ground-wall leakage met acceptance. That is all it says. The tech who reads it as "the motor is good" has substituted a conclusion the test cannot support, and the substitution is easy to make because the number looked so clean.

The fault was between turns in one phase. Both sides of that fault sit at essentially the same potential relative to ground, so the DC ground test drove almost no current through it. Under AC excitation the shorted turn behaved as a shorted secondary, drew heavy circulating current, heated locally, and pulled enough phase current on start to open the overload.

What the pass was worth. Not nothing. It eliminated the ground wall, the cable and the terminal box in one measurement. The error was writing "megged good" instead of "ground wall and cable clear at the test voltage, turn insulation not evaluated."

Outcome two: the reading failed and the machine was fine

Same size of machine, a fan motor on a rooftop unit, reading 12 megohms phase-to-ground at 500 V DC with everything connected: motor, machine leads and the cable back to the disconnect. Well under the acceptance figure the specification called for. The reflex is to condemn the motor.

Run the gate instead. The result belongs to everything in the circuit, so the first job is to reduce the circuit. Lift the leads at the motor terminal box and test the two pieces separately, same voltage, same duration.

• Motor alone: 156 megohms.
• Cable and leads alone, motor disconnected: 13 megohms.
• Check the pair against the combined reading: conductances add, so 1 divided by 156 is 0.00641, 1 divided by 13 is 0.07692, and the sum 0.08333 inverts to 12.0 megohms. That matches the reading taken with everything connected, which confirms nothing was missed and no third path was hiding.

Convert to leakage current, because current is what the instrument actually measured and it makes the share obvious. At 500 V applied:

• Total: 500 divided by 12 megohms is about 41.7 microamps.
• Through the cable: 500 divided by 13 megohms is about 38.5 microamps.
• Through the motor: 500 divided by 156 megohms is about 3.2 microamps.
• The cable carried about 38.5 of 41.7 microamps, roughly 92 percent of the total leakage.

The motor was not the fault. Water had entered a fitting on the cable run and the leakage was surface conduction along a wet path. Drying and resealing that fitting brought the combined reading back up.

The failure mode this case prevents. A motor gets replaced, the new one goes on the same wet cable, the combined reading is still low, and now there is a new motor on the ticket and an unresolved fault on the job. The number was real; the interpretation attributed it to the most expensive item in the circuit because nobody reduced the circuit first.

What would change the answer. If lifting the leads had produced 13 megohms on the motor and 156 on the cable, the same arithmetic points the other way and the motor is the finding. Only the location of the dominant conductance changes. And if both pieces read low, you have two problems, and the combined-reading cross-check is what tells you so.

How to verify you got this right

• Did you state the test voltage and the duration with the number? Without both, the figure is not comparable to anything, including your own reading next year.
• Can you name what was in the circuit? Write down every piece the leads reached. If you cannot list it, the reading has no defined subject.
• When the pieces were tested separately, do their conductances add back to the combined reading? If they do not, something was still connected or a path was missed.
• Does your written conclusion stop where the gate stops? "Ground wall clear at the test voltage" is defensible. "Motor good" is not.

References

• 29 CFR 1910.333(b)(2) - lockout and tagging of circuits for work on electrical installations
• 29 CFR 1910.147(a)(1)(ii)(C) - the carve-out that sends electrical utilization work to Subpart S rather than the general lockout standard
• NFPA 70E-2021, 120.5 - verification of an electrically safe work condition, binding through your employer's electrical safety program or your contract in the edition adopted
• IEEE 43, in the edition referenced by your specification or the machine manufacturer's service literature, for acceptance values and test voltages on rotating machinery
• See related: How to Read an Insulation Resistance Result Over Time; What a Resistance Reading Tells You and What It Hides