The coil that passes every test on the bench and dies in the engine bay. Same part number, different physics — and the cold test never sees it.
An ignition coil is a transformer: a primary winding charged from system voltage, a magnetic field, a collapse that induces tens of thousands of volts in the secondary. Every one of its failure mechanisms is temperature-sensitive, because the coil is wound copper, insulation, and (in smart coils) driver electronics — and heat changes all three.
A coil that fails hot but tests good cold is not a mystery and not a fluke. It is a coil whose defect only exists — or only matters — at operating temperature. The bench test at room temperature measures a different component state than the engine bay at 90°C. The cold reading is real data about the wrong condition.
Three mechanisms, and they often stack:
Service information gives primary and secondary resistance specifications measured at room temperature — typically 20°C. Those numbers are correct and useful, and they do not apply at operating temperature. Copper's resistance rises with temperature at about 0.393% per degree Celsius:
Rexpected hot = Rcold × [1 + 0.00393 × ΔT°C]
A primary that reads 0.8 ohms cold at 20°C should read about 1.0 ohm at 90°C — that is normal physics, not a fault. Judging a hot resistance reading against the cold spec condemns good coils. The corrected value is supporting evidence only: it tells you whether the winding is behaving like copper at temperature. What condemns or clears the coil is its output performance at temperature — spark energy, pattern, the ability to fire under load — not its resistance number.
Resistance is a static measurement of the winding. It cannot see insulation that leaks only under high voltage, connections that open only under thermal expansion, or a driver that only drifts when hot. All of those defects live in the coil's dynamic behavior — what it does when it is actually firing at temperature.
The dispositive test is always a performance test at the failing thermal state: does the coil produce adequate spark energy, with a clean pattern, under the load conditions where the symptom occurs? A coil that resists correctly hot but cannot fire the plug hot is failed. A coil that fires correctly hot is good across the tested range, whatever its resistance number says. Performance at temperature is the verdict; resistance is context.
It is not "coils just do that." Coils fail hot for specific physical reasons — insulation, expansion, driver drift — and each one is diagnosable. Shrugging and replacing the coil on a hot-symptom pattern is a guess that happens to be right often enough to feel like knowledge. The reproduce-isolate-recover method exists to replace the feeling with proof.
It is not a reason to skip the circuit. A coil that fails hot is only condemned after the feed, ground, and command are proved good at that same thermal state. Heat moves connectors, shifts grounds, and stresses drivers — the coil is one thermal suspect among several, and the isolate step is what earns the verdict.
It is not what the heat gun does to potting. Indiscriminate heating — a heat gun blasting the coil body — can exceed the service temperature range and damage insulation that was fine, creating the failure you are trying to diagnose. Controlled heating, monitored, within the service range: engine to operating temp first, supplemental heat only as needed. If you cooked it past reality, you proved nothing.
SpeedNeeds diagnostic procedures apply this physics directly. The procedure requires the failure reproduced at a defined thermal state, isolates the coil at that state — command verified good hot (amplitude, dwell/frequency, and cleanliness on smart-coil triggers, not just "present"), feed and ground verified adequate under load — and confirms recovery that beats the time confound. The copper correction formula appears as supporting evidence: compute the expected hot resistance, compare, and use it for context — while the verdict rides
The copper correction formula appears in B4 as supporting evidence: compute the expected hot resistance, compare, and use it for context — while the verdict rides on output performance at temperature. And the suspect-cylinder gate up front ensures the thermal work happens on the right coil, identified from misfire data or dead-cylinder isolation before any heating begins.
The copper coefficient is copper-standard. Aluminum-winding or exotic coil constructions have different temperature coefficients — service information overrides the 0.00393 figure when it specifies otherwise. The principle (resistance rises with temperature, judge hot readings against corrected expectations) holds; the number may not.
Intermittent thermal opens can be maddeningly state-specific: a connection that opens at 85°C and closes at 80°C gives you a five-degree window to catch it in. If the symptom is intermittent even when "hot," define the thermal state tighter — measure the actual component temperature when the symptom is active, not the coolant temp or the underhood air. The coil's temperature is the thermal state; everything else is a proxy.
And remember the driver: on smart coils, a hot failure with a clean PCM trigger and good feed/ground points at the coil's internal electronics, not its windings. The three-element proof still applies — reproduce, isolate, recover — but the "coil" being condemned is the assembly including its driver, and no winding resistance test will ever see a semiconductor drift.
Cold specs describe a cold coil. The failure lives at temperature. Test where the symptom lives, judge performance not just resistance, and never let a cold in-spec reading clear a hot symptom.
Stop guessing. Prove the next move. A coil that fires correctly at the failing thermal state is good. One that does not — proved by reproduce, isolate, and recover — is a verdict, not a guess.
Stay in the loop
New QTS builds, tools, and garage stories — straight to your inbox. No spam. Unsubscribe anytime.
Read the Privacy Policy.