Continuity, 4-wire, IR or hipot: what each test misses

No single electrical test covers a harness: low-voltage continuity finds opens, shorts and swapped pins but says nothing about insulation, while insulation resistance (IR) and hipot find insulation and spacing faults and show a wiring mistake only as an unexplained short at full voltage. None of the five reliably sees a few broken strands or a missing seal, though 4-wire can show heavy strand loss on short leads. For most harnesses, run continuity on every unit with every point fixtured, add IR and DC hipot when the voltage, pin spacing or contract calls for them and any components are protected, add 4-wire on conductors that carry real current, and use AC hipot only when a specification or the service requires it.

What each test detects and what it cannot

The table gives the logic of each test, from supplier guides and circuit theory, and Cirris warns that detection depends on severity.¹ A terminal not fully seated can pass visual inspection and standard continuity,² so it needs a retention check. A cable that fails only when moved needs a flex test with signals running.³

Why low-voltage tests miss damaged copper

Take 1 m of 0.5 mm² copper: 34.5 mΩ at 20 °C.⁴ Cut 9 of 19 strands over 5 mm and the strands rejoin the bundle beyond the break, so the resistance rises by 0.155 mΩ, or 0.45%. Warming the wire by 1.1 K does as much, because copper moves 0.393% per kelvin. With 18 of 19 cut the rise is 3.1 mΩ. On a short lead that is a larger share, 30% of a 300 mm lead and 180% of a 50 mm one, so 4-wire can show heavy loss if contact variation does not swamp it.³

Cirris cut a 38 in length of 7-strand 26 AWG one strand at a time with a 4-wire meter: 0.1290 Ω intact, 0.1320 Ω with one strand left. Warming part of the wire by hand moved it from 0.1297 to 0.1334 Ω, more than cutting six strands did.⁵ Passing does not clear them either: Molex lists cut or nicked strands among the things that lower pull force.⁶ Control strands at the strip and crimp, and check them with pull tests and by eye.

What a 2-wire reading hides

This illustration takes that metre of wire with two mated contacts at Harwin's 20 mΩ maximum⁷ and two fixture leads at the 0.1 Ω CAMI quotes for a 24 in lead.⁸ The 2-wire reading is 274.5 mΩ, and 87% of it is not the harness wire. Add a bad joint of 50 mΩ and it reads 324.5 mΩ, which a tester with 0.1 Ω resolution⁹ shows as 0.3 Ω both times. A 4-wire reading with its sense points at the fixture contacts goes from 74.5 to 124.5 mΩ, against about ±2.5 mΩ at 1 mΩ resolution and ±2%.

Basic go/no-go testers have one fixed threshold that Cirris puts anywhere from 100 to 10,000 Ω. Against the 274.5 mΩ reading, Cirris's own settings let through a joint of 0.51 Ω on Stringent (1.05 × actual + 0.5 Ω), 1.73 Ω on Good (greater of 2 Ω or 1.10 × actual + 1 Ω) and 4.73 Ω on Moderate (greater of 5 Ω or 1.20 × actual + 2 Ω).¹⁰ At 3 A, even the 0.51 Ω makes 4.6 W. For a conductor carrying amps, work back from the heat you could accept in a joint. DIT-MCO suggests 4-wire for paths at or below about 1 Ω and says it may be unnecessary when you only need to confirm continuity below 5 Ω.¹¹ It costs twice the test points, more fixture work and slower testing.¹²

Hipot finds gaps, not nicks in open air

Hipot fails a part when the voltage can break down a path or leakage passes its limit, and the reach in air is short. Cirris's arc-gap calculator gives about 0.05 mm at 500 V, 0.13 to 0.15 mm at 1,000 V and 0.23 to 0.43 mm at 1,500 V DC, so a pass at 1,500 V does not prove a 1 mm gap. Cirris warns that the tool is not for safety clearances.¹³ It lists nicked insulation among the faults hipot helps find,¹⁴ but our inference is that a nick with nothing metal nearby will pass, so it still needs inspection.¹

A 1 mA limit at 1,500 V DC trips only below 1.5 MΩ. NASA's public IR limit of 100 MΩ at 500 V DC is 5 µA,¹⁵ a resistance 67 times higher. So contamination that fails IR can pass hipot: Cirris calls IR the test sensitive to flux, oils and mould release, and hipot the one sensitive to arcs.¹⁴

Run them in NASA's order: continuity, then hipot, then IR. Continuity first spares a miswired unit the overstress, and IR last can reveal damage the hipot caused.¹⁵¹⁶

Why long cables trip AC testers

Under AC a good cable's capacitance still draws current on every cycle: I = 2πfCV. These figures use 1,000 V rms at 50 Hz, a round figure below the 1,200 V rms test voltage of LAPP's screened control cable, and its published 160 pF/m core to screen.¹⁷

Hipot raises each net against all the others in turn,¹⁸ so a screen sees every core at once. Treat 12 × 160 as a ceiling: every core is at the same voltage, so coupling between neighbours drops out, and the inner cores sit behind the outer ones. Our rough field calculation for a 3+9 lay, with assumed dimensions and no measurement, gives about 600 pF/m.

With a 1 mA limit at 1,000 V, a good cable fails on capacitive current alone at 19.9 m of core against the rest, and between 1.7 m and 5.3 m of screen net. At HSE's 3 mA those become 59.7 m and 5.0 m to 15.9 m. NASA says assemblies with large capacitance, generally runs over six feet or harnesses with shielding, should use DC to avoid false failures.¹⁵ Capacitance moves with wire spacing and coiling, so measure your own net.¹⁹

Under DC the current falls away once the cable is charged, but long nets stop at the ramp or the tester's charge limit. At NASA's 500 V/s the ramp to 1,500 V takes 3 s and a 192 nF net draws only 96 µA. The harder limit is charge: a tester that stops at 45 µC refuses anything above 30 nF at 1,500 V.²⁰¹⁹ Seaward suggests starting with a 30 s ramp when a DC test aborts on over-current,²¹ if your specification allows it: that is 50 V/s, below NASA's minimum of 500 V/s.

DC stresses the insulation in one polarity only, adds ramp and discharge time, and leaves charge on the part that can hurt someone.²¹ Cirris says AC can find problems DC misses, particularly with long application times, and that AC is known to degrade PVC insulation.²⁰ Replacing an AC figure with DC usually means 1.414 times the rms value,¹⁸ but only where your specification allows that swap. If you must run AC on a capacitive part, use a tester that reports real and total current separately, so that the capacitive current does not hide a defect.¹⁸

Which tests to put on the specification

Before any hipot or IR, list every component in the assembly: diodes, resistors, LEDs, capacitors, filter connectors, anything with a chip. Program each component and the wires on all its legs as one net so that no voltage appears across it, and have a test technician review the program, because a self-learn can miss parts.²²²³ First check that your tester can raise every point of such a net together: Cirris, comparing rival sets, warns that a tester applying voltage to one point at a time can damage the component.²² Nets linked this way are not tested against each other, so say so on the specification. Test with the assembly disconnected from any board or module. If a part cannot be protected this way, ask its maker what it will tolerate before you test. A misapplied voltage can damage a static-sensitive part.²³

Where the test voltage comes from

There is no universal test voltage. Take it from your customer's drawing or contract, then the product standard, and check it against the lowest limit in the data sheets of the wire, cable and connector you actually used. For panels the product standard decides, so start with which standard applies. If your customer names IPC/WHMA-A-620, use the values in your licensed copy.

None of these public examples is a figure to copy.

NASA's 1,500 V DC is above Harwin's 1,200 V, so the connector sets the limit. Treat the lowest data sheet proof figure as a ceiling, not a setting, and leave room for your tester's tolerance: at the ±10% Cirris allows,²⁰ a 1,200 V ceiling means setting no more than about 1,090 V DC. Harwin's figure is DC or AC peak, so on AC the ceiling is about 849 V rms and the setting no more than about 770 V rms. If the customer's figure is above that ceiling, raise it with them in writing before testing: it does not make the connector stronger. Do not keep rerunning a failed unit at full voltage: Cirris quotes MIL-STD-202 Method 301 advising against repeated application.²³

Who may run this, and what guards it needs

In Great Britain the law is the Electricity at Work Regulations 1989, and HSE calls regulation 14 particularly relevant to live testing.²⁵ You must carry out a risk assessment before testing begins, and record it if you have five or more employees.²⁶²⁵ Start with HSE's first question: can the test be done dead, or at a safe voltage or current? Then follow its order: eliminate the risk, use a safer voltage or current, control what you can by hardware, control the rest by a written safe system, and finally use competent workers.²⁶

HSE's leaflets name no qualification. They require competence for the task, meaning suitable training, skill and knowledge, or adequate supervision.²⁶ HSE also notes that production-line testers are often unskilled or semi-skilled electrically²⁵ and says not to rely on personal electrical competence as the main protective measure.²⁶ So the station has to protect the person, testers need thorough training,²⁵ and your written safe system says who is authorised.²⁶

• Choose proprietary test equipment where possible, with output current reliably limited to no more than 3 mA. HSE does not consider voltages above the low-voltage limits dangerous at that current, and notes older equipment may be limited only to 5 mA.²⁶ A purpose-built set must meet the same standards of safety as a proprietary one, and the way it connects to the part must be safe too.²⁶
• Do not rely on 3 mA alone. It limits what the tester can feed, and HSE's note gives it as AC.²⁶ Our reading is that it does not limit what a charged cable gives back through a hand. HSE warns that stored charges can also cause injury,²⁶ and that a shock can lead to a fall or other reaction injury.²⁵ Cirris says such limits must not be the sole basis of safe use.²⁷ Our advice goes further than HSE's wording: arrange the station so that nobody can touch the part while it is charged, and use the figures below as design checks, not as permission to touch.
• Move a long AC run to DC if the specification allows, rather than raising the current limit. It will not fit under 3 mA, and HSE keeps higher currents for where they are absolutely necessary and says most insulation testing can be done within the safe limits.²⁵²⁶ Several figures in the AC table exceed the 5 mA that HSE says can cause injury,²⁶ and Cirris quotes NFPA 70E that about 40 mA for a second or more may be fatal.²⁷
• Keep people out while a long DC net is charged. Moving to DC does not remove the hazard: it swaps running current for stored charge. If the net can hold more than the charge limit below, use an interlocked enclosure or a tester with a safety interlock,¹⁴²⁶ and remember that a long charged net can drive a destructive current through the fault if it breaks down.²³
• Check stored charge against the published limits. HSE says to prevent contact with stored charges above 350 mJ,²⁶ which at 1,500 V is 311 nF, or 162 m of screen net at the ceiling figure. Cirris publishes stricter figures, 45 µC and 35 mJ, which it attributes to IEC 61010-1:2010 and EN 50191:2010.²⁷ We have not read those paid standards, so until someone with a licensed copy confirms, design to the stricter: 45 µC at 1,500 V is 30 nF, or 15.6 m of screen net at the ceiling figure. Do not use the rough 600 pF/m for this limit. If a net may be near 30 nF, measure it, and count the tester's own contribution as Cirris does.²⁰
• Use HSE's extra precautions if a contract truly forces more than 3 mA: interlocked enclosures or heavily shrouded probes with retractable sleeving and a switch in the handle, and authorised people only.²⁵²⁶ An interlock should isolate the supply. Interlocking through the control system alone is generally not acceptable.²⁶
• Mark the test area, with fixed barriers, warning lights and emergency stops, and leave no exposed conductors at dangerous voltage on jigs.²⁵²⁶ A 30 mA RCD on the test equipment supply is supplementary protection. It does not prevent a shock.²⁶
• Keep hands off the part during the test and let the test finish before unplugging it.¹⁴ Treat the part as charged until it has been discharged. That includes a test stopped by the emergency stop, a fault or a power cut, so decide in your safe system how residual charge is discharged then.²⁷ HSE says it may be necessary to discharge stored energy safely before further contact,²⁵ and that where charge can be stored the equipment should be earthed before entry, preferably automatically, otherwise with a suitably insulated earthing tool. Where earthing is automatic, it says a manual earthing device is also essential before contact.²⁶ On long or shielded runs, have whoever signs off your safe system set the discharge time.
• Assess first aid needs.²⁶ Keep people with electronic implants, or a medical condition where they should not be startled, away from high-voltage testing, as Cirris advises.²⁷

Write the result down as a test specification: every net pair to be tested, the voltage and where it came from, AC or DC, ramp, dwell, current or resistance limit, the order of tests, and who may change them.

Notes

1. Cirris, Causes of bad insulation. Supplier guide. ↩ ↩²

2. Cirris, Connector contact retention. Supplier guide. ↩

3. Wiring Harness News, Electrically checking crimps, archive article built around a CableEye test system from CAMI Research. Supplier view. ↩ ↩²

4. National Bureau of Standards, Copper Wire Tables, Handbook 100. Used for the annealed copper resistivity and temperature coefficient at 20 °C. Stranding adds to this: LAPP's data sheet gives a maximum of 38.9 Ω/km for 0.5 mm², 13% higher. ↩

5. Cirris, Can I test for broken strands? Supplier experiment. ↩

6. Molex, Quality Crimping Handbook, order no. 63800-0029, copy hosted by DigiKey. Written for Molex terminals and tooling. ↩

7. Harwin, Component specification C005, Datamate M80 and M83 series, issue 42, October 2024, section 2.1.2. Check your own part. ↩ ↩²

8. CAMI Research, Improve cable reliability with milliohm resistance measurement, Connector Supplier. Supplier article. ↩

9. Cirris, Setting practical resistance specs for continuity testing. Supplier guide. It gives 0.1 Ω resolution in 2-wire mode and ±2% with 0.001 Ω resolution in 4-wire mode for most Cirris testers. ↩

10. Cirris, Guidelines for setting resistance test thresholds. Supplier guide. ↩

11. DIT-MCO, When should four-wire testing be considered? Supplier guide. ↩ ↩²

12. Cirris, 4-wire Kelvin testing. Supplier guide. ↩

13. Cirris, High voltage arc gap calculator. Supplier tool. ↩

14. Cirris, Testing cables with high voltage. Supplier guide. ↩ ↩² ↩³ ↩⁴ ↩⁵

15. NASA, NASA-STD-8739.4A with Change 4, paragraphs 18.3.1 to 18.3.5. A public document for NASA hardware, not the standard your customer may specify. ↩ ↩² ↩³ ↩⁴

16. Cirris, Can high voltage damage cables? Supplier guide. Its claims about its own testers' energy limits apply to those testers. ↩

17. LAPP, UNITRONIC LiYCY data sheet 0034302, valid from September 2006, copy hosted by Rapid. Check the current sheet for your cable. ↩

18. Cirris, AC hipot testing. Supplier guide. ↩ ↩² ↩³

19. Cirris, How to hipot highly capacitive cables. Supplier guide. ↩ ↩²

20. Cirris, Using voltage to detect insulation defects. Supplier guide. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

21. Seaward, DC flash support note. Supplier guide. ↩ ↩²

22. Cirris, High voltage testing of cables with components. Supplier guide. ↩ ↩²

23. Cirris, Does repeating the hipot test damage your cable? Supplier guide, quoting MIL-STD-202 Method 301. ↩ ↩² ↩³ ↩⁴

24. Cirris, Should connector shells and unused contacts be tested? Supplier guide. ↩ ↩²

25. HSE, Safety in electrical testing: products on production lines, EIS38 (rev 1), 2016. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

26. HSE, Safety in electrical testing at work, INDG354 (rev 1), 2013. HSE's GS38 covers probes, leads and instruments on circuits up to 1,000 V AC, which helps with a meter but does not set up a hipot station. HSE also names BS EN 50191 for test equipment, a paid standard not read for this article. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷ ↩¹⁸ ↩¹⁹

27. Cirris, Ideas for improving hipot safety. Supplier guide. Its 45 µC and 35 mJ figures are the supplier's account of paid standards, and its FAQ quotes 45 µC and 350 mJ from the 1993 EN 61010-1. ↩ ↩² ↩³ ↩⁴ ↩⁵