How to Insulation Test a Feeder or Branch Circuit

Why this matters

An insulation resistance test is the only routine check that sees damage you cannot: a conductor scraped on a burr during the pull, water in a buried run, a staple driven through a jacket. It is also the test most often run in a way that means nothing. Test with a dimmer still on the circuit and you read the dimmer. Forget to lift the neutral and the neutral-to-ground reading is zero, along with any phase reading taken through a connected load. Forget an AFCI breaker and you may destroy it. This card is the procedure for a feeder or a branch circuit, built around the record you fill in, because a test with no record is a test nobody can trend or defend. Reading results across years is 'How to Read an Insulation Resistance Result Over Time'.

Two hazards: the circuit and the tester

The circuit. Open the overcurrent device at the supply end, lock and tag it under your written program (NFPA 70E Article 120, 120.5 in the 2021 edition and renumbered in 2024; work practices at 29 CFR 1910.333(b)(2), and 29 CFR 1926.417 on construction work), and prove dead with live-dead-live on a CAT III or better meter, proved on a known live source before and after. On a feeder, do it at both ends: the far end can be back-fed by a generator, a second feed, or a PV system.

The tester. A megohmmeter puts hundreds to a thousand volts DC on the conductors. That is a shock hazard at the far end of the run, where you are not standing. Before every test, someone at the far end or a lock and a sign on the far enclosure says the conductors are under test. A long cable stores charge like a capacitor and keeps it after the test stops, so every pair gets discharged and proved at zero before anyone touches it.

What comes off the circuit first

The test measures insulation. Anything electronic or anything that connects conductors to each other or to ground will either be measured instead of the insulation or be damaged by the test voltage. Isolate:

• Surge protective devices. They are designed to conduct above a threshold your test voltage can exceed.
• AFCI, GFCI and dual-function breakers. Lift the circuit conductors off the breaker and test the wiring alone. The breaker's electronics and its neutral pigtail sit across the conductors you are testing.
• Dimmers, smart switches, occupancy sensors, timers, smoke alarms, LED drivers and anything with a power supply. Remove or disconnect them. Plug-in loads get unplugged.
• The neutral at the bar. In a service panel the neutral bar is bonded to ground, so a phase-to-ground and neutral-to-ground test is meaningless until the circuit's neutral is lifted. In a subpanel, lift it too: the bar ties this circuit's neutral to every other circuit's neutral and their loads.
• Motors and transformers, when the acceptance basis is the wiring. A motor can be tested separately at its own voltage.

On branch circuits, switches go on. A switch left off isolates the switched leg, so that piece of wire is never tested. Turn every switch on after the lamps and drivers are disconnected.

The test voltage is a decision you record, not a habit

The voltage comes from the instrument maker and the acceptance basis of the job: an engineer's specification, an owner's commissioning standard, the equipment maker's instructions. Never test above what the insulation, or anything left connected, is rated for, and never assume the last job's voltage. If the job has no written basis, say so in the record and write what you used and why. The record below carries a spec number, which is the only reason its minimum resistance has any authority.

The record, then the steps that fill it

INSULATION RESISTANCE TEST RECORD
Circuit: feeder, main sp 1-3 to garage sub
Cond: 3 x 6 Cu THWN + 10 Cu EGC, 140 ft
Raceway: EMT on house wall, PVC buried
Isolated: 60A 2P off, locked, tagged; sub
  main off; L1 L2 N lifted both ends; SPD off
Instrument: IR tester, cal date checked
Test: 1000 V DC, 60 s per pair
Basis: job spec section 26 05 19, 100 MOhm
Temp: 18 C ambient, run buried

PAIR       RUN 1        RUN 2 (after repull)
L1-EGC     >2000 MOhm   >2000 MOhm
L2-EGC     2.4 MOhm     >2000 MOhm
N-EGC      >2000 MOhm   >2000 MOhm
L1-L2      >2000 MOhm   >2000 MOhm
L1-N       >2000 MOhm   >2000 MOhm
L2-N       1850 MOhm    >2000 MOhm
Discharged to 0 V each pair: yes / yes
Result: RUN 1 FAIL (L2-EGC); RUN 2 PASS

The >2000 entries are the instrument's over-range: it could not measure higher at that voltage. Write the symbol, not a number, because an over-range is not a value you can trend.

1. Fill the header before you lift anything. Circuit, conductors, length, raceway. The length matters: insulation resistance falls as a run gets longer, because more insulation in parallel leaks more total current, so a 140 ft feeder and a 20 ft one should not be compared directly.

2. Isolate and record what you isolated. Lock out, prove dead at both ends, then lift L1, L2 and N at both ends so each conductor is free at the panel and at the subpanel. The "Isolated" line is how the next tester knows whether a low N-EGC reading meant a fault or a forgotten bonding screw.

3. Prove the conductors are dead with the tester itself. Most insulation testers read voltage on their leads and refuse to start above a low threshold. Read it on each conductor before the first test.

4. Test each pair for the stated time. Apply the test, hold for the 60 seconds the basis calls for, and record the reading at the end of that time, not the first number that flashes up. Six pairs on a feeder with a neutral: each phase to the EGC, the neutral to the EGC, and the three conductor-to-conductor pairs. On a branch circuit with fixtures still in place, a conductor-to-conductor test reads the fixture, so test each conductor to the EGC and write that the phase-to-neutral pair was not tested and why.

5. Discharge every pair. Most testers discharge through their own circuit when the test button is released; hold the leads on until the voltage display returns to zero, and on a long cable give it at least as long as the test lasted. Tick the discharge line only after you have seen zero.

6. Record the temperature. Insulation resistance drops as temperature rises, by roughly half for every 10 C on many insulation types, so a reading taken in July and one taken in January are not the same measurement. Record the ambient, and note where the run is (buried, attic, outdoor). Trending them belongs to the sibling article; the field tech's job is to make trending possible.

7. Write the result against the basis, then restore. Pass or fail against the minimum the basis named. Then reland every conductor, torque each termination to the value on the equipment with a calibrated tool (NEC 110.14(D), 2023 edition as adopted), reinstall the SPD and every device you removed, and remove your locks. When you restore, stand to the side of each enclosure as its breaker closes, and confirm the subpanel's loads and the SPD's status indicator come on.

The worked feeder: the failure, then the find

Run 1 failed one pair: L2 to EGC at 2.4 MOhm against a basis of 100 MOhm, about one-fortieth of the minimum, while the other five pairs read over-range. L2-N read 1850 MOhm, just under the instrument's range and far above 100 MOhm, so it passed on its own. One phase low to ground and normal to everything else points at a local insulation break on that one conductor near something grounded.

Sectioning the run. The house-wall portion is EMT, which is itself bonded to the EGC system. The first place to look is where the conductors change direction against metal. With every conductor still isolated and locked out, the tech opened the LB at the top of the wall, where the run turns into the house. L2 had a scrape through its insulation where the pull dragged it over the edge of the fitting's opening, sitting against the fitting with condensation in it. A clean metal contact would read near zero; a damp scrape reads in megohms, which matches 2.4.

The repair. Tape does not restore a conductor's insulation rating inside a raceway. The L2 conductor was pulled out and a new one pulled in, with a bushing checked at the fitting and the pull lubricated. Run 2: all six pairs over-range, recorded in the second column of the same sheet. That is the record a reviewer wants: the failure, the find and the retest on one page, not a clean sheet that hides a repull.

What changes the answer

• The basis gives a different minimum or voltage. A maker's motor test, an owner's lower-voltage standard for control wiring, or a utility commissioning sheet each sets its own. Use the one the job runs on and write which.
• One pair passes but sits far below its neighbors. Say 150 MOhm on one pair with the others over-range, on a 100 MOhm basis. It passes, and it is the pair to record, flag and retest at the next visit, because a local break often reads like that before it reads like run 1 above.
• No basis exists. A residential service call has no spec. Record what you used, state it as a baseline rather than a pass, and compare next time.

References

• NFPA 70E, Standard for Electrical Safety in the Workplace, current edition, Article 120; 29 CFR 1910.333(b)(2); 29 CFR 1926.417 for construction work
• NFPA 70, National Electrical Code, 2023 edition as adopted, 110.14(D) for retorquing terminations after the test
• Insulation tester manufacturer's operating instructions (test voltage, timing, discharge)
• See related: How to Read an Insulation Resistance Result Over Time
• See related: What an Insulation Resistance Test Actually Tests