How a Connection Carries Electrical Current

Why this matters

An electrical connection is a mechanical device that happens to pass current. Everything that goes wrong with one goes wrong mechanically first, and the electrical symptom arrives later, which is why connections get replaced without anyone learning why they failed.

Understanding the mechanism changes three things in the field: what you measure, when you can trust a thermal scan, and why "I snugged it up" and "I torqued it to the label" are not two versions of the same action.

Before you measure anything on a connection

The measurement described later in this card is taken on a live circuit under load, and that is energized work.

• De-energize wherever the work can be done de-energized, under 29 CFR 1910.333(b)(2), and verify absence of voltage using the live-dead-live proving sequence in NFPA 70E-2021, 120.5, which binds you through your employer's electrical safety program or a contract rather than on its own.
• Working on or near exposed energized parts requires justification that de-energizing introduces additional hazards or is infeasible, plus the shock and arc flash risk assessment, boundaries, and PPE your employer's electrical safety program specifies under the NFPA 70E edition it has adopted. 29 CFR 1910.333(a)(1) is where the de-energize-first duty lives.
• Never re-tighten a loaded, energized connection. Every recommendation below about restoring contact force is a de-energized action. A lug that has been running hot is at reduced strength, the conductor may be annealed, and moving it under load can open the circuit at the point of highest resistance, in your hands, at full available fault current.
• A connection that has overheated may have damaged insulation, the terminal, and the conductor. Treat the conductor as suspect back to sound copper or aluminum, not just the lug.

Where the current actually crosses

Put two metal surfaces together and, however flat they look, they touch only at the high points. The real metal-to-metal contact is a scattering of microscopic spots, and their combined area is a small fraction of the apparent contact area you can see.

Current has to funnel through those spots. The lines of current crowd inward on approach, pass through the spot, and spread out again, and that crowding produces resistance in addition to the resistance of the bulk metal. That constriction resistance is most of what a connection's resistance is.

The size and number of the spots is set by force, not by area. Push harder and the high points deform, flatten, and grow, and new ones come into contact. Within the range that matters in practice, contact resistance falls as contact force rises, and it does so steeply at low force and then with diminishing return as the force gets high, which is why a connection goes from awful to good over a fairly small increase in clamping load and then stops improving.

That single fact reorganizes the subject: the electrical quality of a connection is the mechanical state of its clamping force. Everything that reduces the force, reduces the connection.

Films, and why aluminum is a different animal

The other thing between the two surfaces is whatever chemistry has grown on them.

Copper oxidizes, but its oxide is comparatively soft and breaks up readily under clamping force, and copper connections in a dry, undisturbed environment are generally well behaved.

Aluminum grows a hard, tenacious, electrically insulating oxide almost instantly on exposed surface, and it grows back within seconds of being disturbed. An aluminum termination therefore depends on the connector deliberately breaking through that film and then keeping air away from the interface, which is what a listed connector plus the manufacturer's specified compound is for. Aluminum also creeps under sustained compressive load more than copper does, so a joint that was correct on day one loses force over time by the metal flowing out from under the pressure.

Add temperature cycling, and both effects run together with a third: tiny relative movements at the interface repeatedly break and re-form oxide, grinding away contact spots. That is fretting, and it is why a connection can degrade with no one touching it and no obvious fault.

Do not mix aluminum and copper in a connector that is not listed for the combination, and use the compound the connector manufacturer specifies, in the way they specify it. This is a listing question, and it is enforced through the NEC as adopted by your authority having jurisdiction.

The feedback loop that makes this urgent

Here is why a connection that has started to go usually finishes going.

Resistance rises a little. Power dissipated at the connection is current squared times resistance, so heat rises with it. Heat accelerates oxidation, accelerates creep and relaxation in the clamping members, and drives more thermal cycling movement. Force falls, contact spots shrink, resistance rises further. Each turn of the loop feeds the next.

That is why connection failures look sudden. The mechanism is a positive feedback loop with a slow beginning and a fast end, and everything you do about connections is aimed at the slow part.

The measurement, and the comparison that makes it mean something

Resistance at a connection is far too small to read directly with an ordinary meter. What you can read is the millivolt drop across it while it carries current, because voltage drop is current times resistance.

The trick that makes this practical is comparison. Measure the drop across the connection in question and across an identical connection on the same circuit carrying the same current, at the same time. You do not need to know the correct absolute value for that lug, and you should not go looking for one, because it depends on the connector, the conductor size, and the load.

Two conditions on that comparison, both of which invalidate it if you ignore them:

• Same current. Drop scales directly with current, so comparing a lug at full load against one at a quarter of load compares nothing.
• Same physical connection type. A bolted lug and a spring-clamp terminal have different normal drops.

Worked example: two lugs on the same load

A three-conductor feeder, all three lugs the same type, same conductor size, all carrying about 60 A. Illustrative values used to carry the method.

The arithmetic: resistance is drop divided by current, so 0.003 V divided by 60 A is 0.00005 ohm, which is 50 microohms. Power is current squared times resistance: 60 x 60 x 0.00005 is 0.18 W. For lug C, 0.300 V divided by 60 A is 0.005 ohm, and 3,600 x 0.005 is 18 W.

Lug C is at 100 times the resistance of its neighbors and 100 times the heat, in a place designed to shed a fraction of a watt. Eighteen watts inside an enclosure, concentrated at a single point on a conductor, is a fire origin, and the two identical lugs beside it are what turn "300 mV, is that bad?" into a finding.

Now the trap, and it is the one that makes thermal scans lie. Heat goes as the square of current. Take the same lug C at 15 A instead of 60 A. Same 5,000 microohms, but 15 x 15 x 0.005 is 1.1 W, which is about one sixteenth of the heat. A scan taken on a light morning load may show lug C a few degrees above its neighbors and get written up as acceptable; the same lug on a design-day afternoon is at 18 W. So a thermal survey is only as good as the load it was taken at, and a survey report without the load current on it cannot be interpreted at all. The direction runs the other way too: a connection that shows a large temperature rise at light load is worse than it looks, not better, because you are seeing the low end of a squared relationship.

What the number does not tell you. A 300 mV reading says the contact is bad. It does not say why. Loose from day one, relaxed over years, fretted by cycling, corroded, a strand-damaged conductor, or a connector that was never listed for that conductor material all read the same on the meter. The repair is decided by what you find when it is open and de-energized, and if you cannot establish which of those it was, you have not finished: a lug re-terminated without correcting the cause reads perfect on the meter and comes back.

What was found on this one: an aluminum conductor in a connector listed for the combination, but with no compound applied and a torque that could not be confirmed. Re-terminated de-energized, on freshly prepared conductor, with the manufacturer's compound and their specified torque applied with a calibrated tool. Post-repair reading, taken at a comparable 58 A load: 3 mV, matching its neighbors within measurement error, and reported with the load current beside it so the next tech can compare like with like.

Torque is the input, force is the target

The clamping force is what the connection experiences and the torque wrench is an indirect way of asking for it, with friction taking a variable share along the way. A sibling article covers that conversion. Two consequences specific to electrical terminations:

• The manufacturer's torque value is not advisory. Where a numeric torque value is provided on the equipment or in the manufacturer's instructions, the NEC requires a calibrated torque tool to apply it, at 110.14(D) in the 2017 and later editions as adopted by your authority having jurisdiction, with the general termination requirements at 110.14. "Snug plus a bit" is a different value on every wrist in the shop.
• Whether a connection may be re-torqued at all is the connector manufacturer's call. Some designs are one-time, and re-applying torque to an already-seated connection can over-compress the conductor. Where re-torque is not permitted, the answer to a degraded connection is re-termination, not another pull on the screw.

How to verify you got this right

• Record the load current alongside every millivolt or thermal reading. A reading without its current is not data.
• Compare against a sibling connection on the same circuit at the same moment, rather than against a remembered value.
• Re-read after the repair, at a comparable load, and expect the repaired connection to sit with its neighbors rather than merely lower than before.
• Where the connection is aluminum, confirm the connector's listing for the conductor material and that the specified compound was used, and write both on the record. Those are the two facts that are impossible to recover once the cover is back on.
• If a breaker or fuse has been operating on the circuit, establish why before treating the connection as the whole story. A protective device that opened did its job, and a bad connection upstream of it and a genuine overload downstream of it are different findings.

References

• 29 CFR 1910.333(a)(1) and 1910.333(b)(2) for de-energizing and verification before work on electrical conductors and equipment, with NFPA 70E-2021, 120.5 for the live-dead-live proving sequence where your employer's electrical safety program adopts it
• NEC Article 110, including the termination requirements at 110.14 and the calibrated torque tool requirement at 110.14(D) in the 2017 and later editions, as adopted by your authority having jurisdiction
• Connector and equipment manufacturer documentation for torque values, listed conductor materials, antioxidant compound, and whether re-torque is permitted
• See related: How Connectors and Terminals Fail; Wire Termination Types Reference; What Torque Actually Controls and What It Does Not