Arc resistance is a stopwatch reading. You put two tungsten electrodes on a flat sample, strike a high-voltage, low-current arc discharge across the surface between them, make the arc progressively nastier every 60 seconds, and write down the second at which the material gives up. That number, in seconds, is the result. The standard sets no universal pass/fail limit and no unit other than time.
IEC 61621 is the method that defines how you run that stopwatch so two labs get comparable numbers.
Edition note. This article describes IEC 61621:1997, the edition currently in force. A second edition is in development under IEC TC 112 and reached DIS ballot in mid-2026. Check which edition your product standard or customer specification calls up before you quote a result.
Table of Contents
The ladder is the test
Everything else in the standard exists to serve one thing: a fixed seven-step severity ramp, 60 seconds per step, 420 seconds total if the sample survives to the end.
| Step | Arc current | Duty | Cumulative time |
|---|---|---|---|
| 1/8 | 10 mA | 0.125 s on, 0.875 s off | 60 s |
| 1/4 | 10 mA | 0.25 s on, 0.75 s off | 120 s |
| 1/2 | 10 mA | 0.25 s on, 0.25 s off | 180 s |
| 10 | 10 mA | continuous | 240 s |
| 20 | 20 mA | continuous | 300 s |
| 30 | 30 mA | continuous | 360 s |
| 40 | 40 mA | continuous | 420 s |
Two things are worth reading carefully.
The first three steps all run at 10 mA. Only the duty cycle changes. And the third step does not lengthen the on-time; it shortens the whole cycle to half a second, so you go from a quarter duty to a half duty with the same 0.25 s burst. The device doing this is a motor-driven or electronically operated interrupter, and it has to hold ±0.008 s. A manual switch is not an option.
The reason for the interrupted start is discrimination at the bottom end. A continuous arc from second one would burn through weak materials too fast to tell them apart. The reason the current does not go below 10 mA is stability: less than that and the arc wanders and flares instead of sitting flat on the surface.
Once you reach step four the ramp is pure current. 10, 20, 30, 40 mA, continuous, one minute each.
What counts as failure
This is where operators disagree most, so the standard pins it down.
Failure is a conducting path formed in the material. It is also failure if the arc sets the material burning and that burning keeps going after the arc is interrupted.
Five practical points sit underneath that definition:
You will usually hear it before you are sure you saw it. When the arc drops into the material, the circuit current shifts and the sound changes noticeably. That change is your cue to look hard.
Partial disappearance is not failure. On some materials the arc fades away in pieces over a fairly long stretch. The clock keeps running until the entire arc is gone.
Scintillation at the electrode tips does not count. Some materials keep sparkling near the tips after the arc itself has gone. That is not the arc, and you do not restart the clock for it.
Burning only counts if it persists between interruptions. If the flame dies each time the arc breaks, keep going until a conducting path actually forms.
First total disappearance wins. If the entire arc goes and then the material recovers and arcs again on the next cycle, the failure time is the first disappearance. You do not get to keep testing and report the later number.
Four ways materials fail
The standard groups the outcomes into four families, and knowing which one you are watching tells you a lot about the material.
Many inorganic dielectrics go incandescent. Hot enough, they conduct. Cool them down and they are insulators again, which makes the failure look alarming and the post-test sample look almost fine.
Some organic compounds simply catch fire, with no visible conducting path anywhere in the substance.
Others fail by tracking in the classic sense: a thin wiry line forms between the electrodes at the moment the arc disappears.
The fourth family carbonizes the surface progressively until there is enough carbon sitting there to carry the current.
If you are comparing two formulations, log the family as well as the time. A 150 s carbonizer and a 150 s burner are not the same material problem.
The rig
The supply is a transformer rated 15 kV open-circuit secondary and 60 mA short-circuit secondary, on line frequency between 48 Hz and 62 Hz, fed through a 1 kVA variable autotransformer. You calibrate the open-circuit voltage to 12.5 kV, read on the primary side through the known turns ratio.
Current is set by four current-control resistors in series with the transformer primary, designated R10, R20, R30 and R40. R10 stays in circuit at all times and provides the 10 mA base current; the others switch in to build the higher steps. All four need some adjustment range so you can trim the exact currents during calibration, which you do with the electrodes correctly spaced and resting on a ceramic block, draft shield closed.
A 15 kΩ ±1.5 kΩ resistor rated at least 24 W sits in the arc circuit with air-core inductors of 1.2 H to 1.5 H. Together they kill parasitic high frequency. Nobody builds that inductance as a single coil; the standard’s own suggestion is eight coils of 3,000 to 5,000 turns on non-metallic cores. Keep stray capacitance in the secondary wiring under 40 pF. Above that, the arc shape itself changes and so do your results.
Measurement is a true-RMS AC milliammeter covering 10 to 40 mA at ±5 %. Since it only earns its keep during setup, it is normally bypassed with a shorting switch during runs. Timing is a stopwatch or interval timer good to ±1 s.
Safety is built into the mechanics: lowering the draft shield closes a normally open microswitch, which pulls the contactor that connects the transformer. Raise the shield and the HV is gone.
Electrodes
2.4 mm ±0.05 mm tungsten rod, free of cracks, pits and rough spots, with at least 20 mm of free length. Tungsten welding rod works. Mount them in shanks so you can put the tip back at the right orientation after each sharpening.
The tip is ground and polished at 30° ±1° to the rod axis, which gives a flat elliptical face. A steel jig to hold the rod during grinding is worth building — the tip geometry is the part of this rig most likely to drift.
In the assembly, both electrodes sit in the same vertical plane, each inclined 35° ±1° from horizontal. That puts 110° ±2° between the two axes. The minor axes of the two elliptical faces run horizontal, and the gap between the tips is 6.35 mm ±0.1 mm. Each electrode presses down independently at 0.5 N ±0.05 N.
You need a clear view of the arc from just above the plane of the sample. Wear anti-UV glasses or work behind a UV shield.
Cleaning and re-sharpening
After every single test: acetone or ethanol on a lint-free lab tissue, then wet with deionised water and wipe dry with a clean dry lint-free tissue. If combustion residue survives that, run a continuous 40 mA arc for about a minute with no specimen in place to burn it off.
Check the tips at 15× magnification. They should still show the original elliptical face with no burrs or rough edges. If they do not, re-sharpen before the next run.
Chamber
A draft-protecting enclosure with no ventilation, no smaller than 300 × 150 × 100 mm. The point is to keep air movement off the arc, not to seal the sample in. The electrode assembly still has to let combustion products escape when a specimen smokes or gases during the run.
Specimens and conditioning
The specified specimen thickness is 3 mm +0.4/−0 mm. Any other thickness has to be reported with the result.
The tested surface must be flat, and the electrode assembly must sit at least 6 mm from any edge and at least 12 mm from anywhere you have already tested. Thin material is tested by clamping sheets tightly together to build up a stack as close to 3 mm as you can get.
For moulded parts, pick the location that actually matters in service and arc there. If you are comparing two parts, arc them in the same place.
Five tests per material, minimum, for a standard comparison.
Condition at 23 °C ±2 °C and 50 % ±5 % RH for at least 24 hours. That is standard atmosphere B. Remove dust, moisture and fingerprints first — and note what you used, because the cleaning agent itself can change the surface you are about to test.
Running it and reading it
Energise, start the clock, and watch the first arc of every step. If the first specimen behaves normally at a given step, you do not need to stare at the rest.
What you are checking early on is arc geometry. A healthy arc lies flat, close to the specimen surface. Three things say something is wrong:
- the top of the arc climbs to roughly 2 mm above the surface
- the arc runs up the electrode instead of staying on the tip
- the arc flares irregularly
Either your circuit constants are off, or the material is off-gassing hard. Fix the first before you blame the second.
At failure, kill the arc current immediately and stop the timer. Record the seconds. Repeat until you have five numbers.
Report the central value plus the minimum and maximum, not a mean. Add the material identity, the thickness, what you did to clean and condition it, and anything odd you saw, especially burning and softening.
The stage-boundary effect
Materials tend to fail in the first seconds after the severity increases. That makes the arithmetic gap between two results misleading.
178 s and 182 s straddle a step change. One material survived the jump to a higher stage, the other did not. That is a real difference in behaviour.
174 s and 178 s sit inside the same stage. Four seconds of the same stress. That is close to noise.
Weight your comparisons accordingly.
Where this test does and does not apply
The honest scope is narrow, and the standard says so plainly.
Thermosets, yes. Reproducibility has been acceptable.
Thermoplastics, generally not recommended. Several labs reported variation large enough that the standard recommends against using the method on them. The suspected cause is uncontrolled electrode pressure and penetration depth as the tip sinks into softening material. Without controlling for that, testing many thermoplastics is not meaningful enough to be worth doing.
Other arc types, no. Results here do not let you rank materials for arcs of a different character.
Wet tracking, no. Ranking by IEC 61621 can differ from the ranking you get in IEC 60112 or the IEC 60587 inclined plane test, and from what the material does in service. Those tests run wet, at lower voltage, with an electrolyte doing the work. This one runs dry, at 12.5 kV, with the arc doing the work. Different mechanism, different answer.
Use it for what it is good at: screening candidate materials early, catching what a formulation change did, and quality control on incoming material.
FAQ
Is IEC 61621 the same test as ASTM D495? They are technically equivalent methods, and IEC 61621 was developed from ASTM D495. They are not interchangeable.
The main difference is the electrodes. ASTM D495 describes two electrode systems: stainless steel strip and tungsten rod. It treats the stainless steel strip as the preferred technique for materials in the poor-to-moderate range, up to about 180 s, and recommends the tungsten rod above that, because the strip corners erode badly under long exposure. IEC 61621 describes only the tungsten rod. D495 says so itself.
D495 also defines normal and inverted specimen orientation, with the electrodes on the upper or the under surface, and requires you to report which you used. Inverted is the more severe condition. IEC 61621 has no equivalent. The specimen clauses differ too: IEC calls for 3 mm +0.4/−0 mm with 6 mm edge clearance and 12 mm from a previously tested area, while D495 works in inch-pound units at 0.125 ± 0.010 in (3.17 ± 0.25 mm), with 1/4 in and 1/2 in clearances. Conditioning, cleaning cadence and reporting all differ as well.
If a datasheet, product standard or qualification document names one of them, test to that one. Do not assume a number produced under the other is equivalent. The full breakdown is in IEC 61621 vs ASTM D495.
Why can’t I test thermoplastics? The electrode tip sinks into softening material, and the depth it sinks to is not controlled by the method. Contact geometry changes during the run, so the arc changes, so the time changes. Between labs the scatter gets large enough that the comparison stops meaning anything.
What is a good arc resistance value? The standard sets none. It is a comparative screening test. A number only means something against another number produced the same way, at the same thickness, on a similar material.
Does a good CTI mean good arc resistance? No. Tracking resistance and arc resistance are different failure mechanisms and the rankings do not track each other. A material can do well on one and badly on the other. See CTI values of common insulating plastics for where the tracking numbers land.
How many specimens do I need? Five, minimum, for a standard material comparison. Report the central value with the minimum and maximum.
Can I test film or thin sheet? Only by clamping layers tightly together to build a stack near 3 mm, and you report the thickness you actually tested.
My arc keeps climbing the electrode. What’s wrong? Either the circuit constants are out of spec — check stray capacitance, the suppression resistor and the inductors — or the material is releasing gas fast enough to push the arc off the surface. Verify the rig on a ceramic block first.
Do I have to clean the electrodes between every test? Yes. Solvent, then deionised water, then dry. And inspect the tips at 15× for burrs before the next run.