Why a Loose Electrical Connection Heats and Then Fails

Why this matters

Two techs open two panels on the same morning. Both find a terminal reading 120 millivolts across it while the circuit is carrying its normal load. One of those joints is a note in the report. The other is a fire that has not happened yet, and it will not announce itself before it happens. Nothing in the 120 millivolts tells you which is which. Worse, the instinct that higher resistance means a worse joint points the wrong way in this pair: the harmless joint has roughly eighteen times the contact resistance of the dangerous one. Read it the usual way and you spend the afternoon re-terminating the joint that was never going anywhere, and close the door on the one that was.

Before you measure: what this task actually is

A drop reading across a terminal has to be taken with the circuit energized and carrying load. That is energized diagnostic work, not maintenance. Do it only under an energized-work justification, inside the shock and arc-flash boundaries and with the PPE your employer's electrical safety program specifies. That program is normally built on NFPA 70E in the edition your employer adopted (NFPA 70E-2021 is the common current basis); the standard binds you through that program or through a contract, not on its own authority.

The instant the job stops being reading and starts being touching - stripping, re-landing, re-torquing, pulling a lug, backing out a screw - the circuit gets de-energized, locked, tagged and proven dead under 29 CFR 1910.333(b)(2) for general-industry work or 29 CFR 1926.417 for construction, using the live-dead-live proving sequence at NFPA 70E-2021, 120.5. 29 CFR 1910.147 is the wrong standard to cite for this; it carves out exposure to electrical hazards from work on electric utilization installations at (a)(1)(ii)(C) and sends you to Subpart S.

Two more, because this article tells you to do both: do not re-torque a joint that is hot and loaded, because the fastener, the conductor and the terminal are all at different temperatures and the torque you set will not be the torque that exists when it cools. And if you take an infrared image, opening an enclosure door to see the terminals puts you at full arc-flash exposure for the whole time the door is open; that is the reason permanently installed infrared windows exist.

What the joint is doing when it is right

A termination is two pieces of metal squeezed together hard enough that the real contact happens on a small number of tiny high spots where the surface oxide has been broken through. Those spots are what carries the current. The clamping force is what makes them, holds them, and keeps air out of them. Everything a good termination does traces back to holding that force.

So the load path through a joint has a weakest element, and it is almost never the conductor. It is the contact interface. Torque, tin plating, antioxidant compound, spring washers and crimp geometry are all just different ways of managing that one element.

The gate: watts, not millivolts

The millivolt drop across a joint is not the hazard. The hazard is the power the joint turns into heat, and where that heat can go. Power is the current through the joint times the drop across it. Two numbers, and most techs only take one of them.

Case A, a control-circuit terminal. The circuit carries 3 amps. The drop across the terminal is 120 millivolts. Joint resistance is 0.120 V divided by 3 A, which is 0.040 ohms, or 40 milliohms. Power dissipated is 0.120 V times 3 A, which is 0.36 watts. Against a 120-volt control supply, 120 millivolts of loss is one tenth of one percent, so nothing downstream even notices, and roughly a third of a watt in a terminal block barely lifts it above the enclosure air.

Case B, a feeder termination. The circuit carries 55 amps. The drop is the same 120 millivolts. Joint resistance is 0.120 V divided by 55 A, which is 0.0022 ohms, or 2.2 milliohms. Power dissipated is 0.120 V times 55 A, which is 6.6 watts.

The safe joint has 40 milliohms, the dangerous one has 2.2, and 40 divided by 2.2 is about 18. The joint with eighteen times the resistance is the one you can leave. The joint that is nearly a short by comparison is dumping over six watts continuously into a lug about the size of your thumb, in still air, inside a box.

Six and a half watts sounds small until you notice it has nowhere to go. That is roughly what a small night light draws, concentrated in a lump of metal with maybe a couple of square inches of surface, surrounded by air that is already warmed by everything else in the enclosure, often with the terminal insulated on three sides by the conductor's own jacket and the adjacent terminals.

Why it accelerates instead of settling out

This is what separates a loose termination from most other faults. It is a closed feedback loop, and every term in the loop pushes the same way.

Heat the joint and three things happen at once. Copper's resistance rises with temperature at roughly four tenths of one percent per degree Celsius near room temperature (that coefficient is for annealed copper measured from a 20 C base, and it applies to the bulk conductor, not to the contact interface). Oxide growth on the exposed faces of the contact accelerates with temperature, and oxide is an insulator, so the current is pushed onto fewer and smaller contact spots. And the clamping force relaxes, because the fastener, the terminal body and the conductor all expand and contract at different rates through every load cycle, and a threaded connection under repeated differential movement loses preload.

Higher resistance at the same current means more watts. More watts means more heat. More heat means higher resistance. There is no stabilizing term anywhere in that loop, which is why a joint does not sit in a mildly degraded state waiting for your next annual inspection.

The direction at the other end of the range matters just as much, and it is the reason this fault fools experienced people. Drop the current and the loop opens. The same physically loose joint on a circuit that only pulls real load for two months a year will sit for seasons with no symptom, no smell and no discoloration, because 0.4 watts does not drive oxide growth fast enough to close the loop. Then the load comes back and the joint that has been "fine for three years" fails inside one season. "It has been like that a long time" is evidence about the load history, not about the joint.

Scaling a reading taken at the wrong load

You will often be able to measure only at whatever load happens to be running. Scale it, and scale it as a square, because at a fixed joint resistance the drop rises with current and the power rises with current squared.

Say you read 30 millivolts across a lug while the circuit is pulling 15 amps. Joint resistance is 0.030 divided by 15, which is 0.002 ohms, or 2 milliohms. At that moment it is dissipating 0.030 times 15, which is 0.45 watts, and it feels like nothing. The equipment's full-load current is 60 amps, four times what you measured at. At 60 amps the drop becomes 0.002 times 60, or 0.120 volts, and the power becomes 0.120 times 60, which is 7.2 watts. Four times the current, sixteen times the heat: 7.2 divided by 0.45 is 16.

That is the same 2.2-milliohm-class joint as Case B, and you found it looking harmless. A tech who records "0.45 watts, monitor next visit" has recorded a true number about a condition that does not exist most of the time.

The same scaling is why an infrared survey has to say what the load was. A hot spot photographed at a fifth of full load understates the full-load rise badly, and infrared readings also depend on the emissivity of the surface you aimed at, so bare bright copper reads far cooler than it is.

What the runaway leaves behind

Once the loop has been running, the evidence is in the metal and it is worth reading before you clean anything up.

• Discoloration that is darkest at the contact face and fades along the conductor tells you the heat was generated at the interface, not delivered to it from somewhere else. Heat arriving from a neighbouring source makes the pattern run the other way.
• A conductor annealed soft for the first half inch behind the lug means sustained high temperature, not a single event. Compare it by feel against the same conductor a foot back.
• A terminal screw that turns freely and then bites is a joint that already relaxed. The bite is the fastener finding metal that has crept out from under it.
• Green or white powder at the interface with no water source nearby is oxide and corrosion product driven by heat, which is the loop's own signature rather than a wet-environment problem.

Any of these means the termination is not re-torquable. The conductor end has lost temper, the plating is gone, and the terminal has been through a heat cycle its listing did not contemplate. Cut back to sound conductor and replace the terminal or the device. A re-torque puts a fresh clamping force onto damaged metal and buys weeks. The sibling article on why a loose connection gets hot follows one such call from the smell through the repair; this one is about deciding which joint deserves that call.

Confirming you fixed the right element

Re-measure the drop under the same load you measured it at the first time, and state both numbers together. A drop that fell from 120 millivolts to 4 millivolts at 55 amps has gone from 6.6 watts to 0.22 watts, a factor of thirty, and that is a real fix. A drop that fell to 60 millivolts is still 3.3 watts and the loop will close again; something in the load path is still wrong, and it is usually that the damaged conductor was re-used or the terminal is rated for a different conductor material than the one landed in it.

Then check the neighbours. A joint that ran hot for a season cooked the terminals either side of it, and they will show up on the next survey as new faults if you only measured the one you came for.

References

• 29 CFR 1910.333(b)(2), general industry, for de-energizing and verifying before contact work on electric utilization equipment; 29 CFR 1926.417 is the construction counterpart for lockout and tagging of circuits.
• NFPA 70E-2021, 120.5, for the live-dead-live proving sequence; it binds through the employer's electrical safety program or a contract in the edition the employer adopted, not on its own.
• NFPA 70 (National Electrical Code), Article 110, in the edition your authority having jurisdiction adopted, for conductor termination requirements including the use of connectors identified for the conductor material and installation to the listed torque.
• See related: Why a Loose Connection Gets Hot; How a Connection Carries Electrical Current; Why Aluminium and Copper Terminations Differ.