Somewhere right now an electrician is standing in front of a steel junction box that is warm, and they cannot work out why. Every lug is torqued. Every splice is tight. The breaker is loaded to barely half its rating. The thermal camera says the box is running hotter than the wire inside it, which makes no sense, because heat is supposed to come from resistance and the box is not carrying any current at all. Except it is. Not through a wire. Through the steel itself. The circuit was run the wrong way, with the hot in one pipe and the neutral in another, and the conduit has quietly become part of the load.

This is one of those failures that never announces itself. It does not trip anything. It does not flicker. It just sits above the ceiling, cooking a little, for years. And the rule that prevents it is one most of us learned as a sentence to memorize rather than a thing to understand: all conductors of the same circuit go in the same raceway. Here is why.

A wire carrying AC is a magnet that never holds still

Any conductor carrying current is wrapped in a magnetic field. With direct current that field is steady, and a steady field does almost nothing to the steel around it. Alternating current is different. At 60 hertz the current reverses direction 120 times a second, and the magnetic field around the wire collapses and rebuilds with it.

Michael Faraday worked out the consequence in 1831: a changing magnetic field induces a voltage in any conductor it passes through. Steel conduit is a conductor. So is a steel box, an enclosure wall, a length of strut. Put a single AC conductor inside a steel pipe and the pipe becomes, in effect, the shorted secondary of a transformer whose primary is your hot wire.

Two things then happen inside that steel. The induced voltage drives circulating currents through the metal, loops that go nowhere useful and simply heat the pipe through its own resistance. These are eddy currents, named for the way they swirl, studied by Léon Foucault in the 1850s. Second, because steel is ferromagnetic, its internal magnetic domains flip back and forth to follow the field, and each reversal costs energy that leaves as heat. That is hysteresis loss. Aluminum conduit gets the eddy currents but not the hysteresis, which is why the Code's language is specifically about ferrous metal.

Neither effect needs a fault. Neither needs anything to be loose. A flawless installation with one conductor alone in a steel raceway will heat that raceway in proportion to the current, all day, for as long as the load runs.

The cancellation trick

The fix is almost embarrassingly simple, and you have already watched it work if you have ever used a clamp meter.

Clamp around the hot conductor of a running circuit and you read the load current. Clamp around the hot and the neutral together and you read zero. The current is still flowing. But at every instant, the current going out on the hot equals the current coming back on the neutral, in the opposite direction. Their magnetic fields are equal and opposite, and outside the pair they cancel. The sensing coil sees no net changing field, so it reports nothing.

Steel conduit is just a very large, very dumb clamp meter. Run the hot and the neutral through it together and the net current through any cross-section of the pipe is zero. No net field, no induced voltage, no eddy currents, no hysteresis. The same holds for a 240-volt pair, for a three-phase set whose three currents sum to zero, and for a multiwire branch circuit with its shared neutral. What matters is that every conductor carrying current for that circuit, including the grounded conductor, travels in the same raceway.

This is also, incidentally, exactly how a GFCI works. It passes hot and neutral through a sensing toroid and trips the moment the net current is no longer zero. The wiring rule and the safety device are leaning on the same piece of physics.

What NEC 300.3(B) and 300.20 actually say

Section 300.3(B) is the general rule: all conductors of the same circuit, including the grounded conductor and all equipment grounding and bonding conductors, must be contained within the same raceway, cable, trench, cord, or cable tray, unless one of the listed exceptions applies. The exceptions are narrow, covering things like paralleled conductors arranged in matching sets, isolated-phase installations in nonmetallic underground raceways, and specific grounding conductor situations.

Section 300.20(A) is the physics made explicit. Where conductors carrying alternating current are installed in ferrous metal enclosures or raceways, they must be arranged to avoid heating the surrounding metal by induction, which means grouping all phase conductors, the grounded conductor where used, and all equipment grounding conductors together.

Section 300.20(B) handles the case where you genuinely must pass a single conductor through a piece of steel, such as one conductor per opening on a large enclosure. The Code's answer is to cut slots in the metal between the individual holes, or pass the conductors through a wall of insulating material, so the steel cannot form a closed magnetic ring around any one conductor. You will see this on big service gear where phase conductors land through separate openings: a saw cut joining the knockouts, or a nonferrous plate in place of the steel.

The second problem: a lonely conductor is a worse conductor

Heat is the obvious consequence. The quieter one is impedance.

A circuit's inductance depends on the area enclosed by the loop of current flowing out and back. Hot and neutral in the same pipe, lying against each other, form a tiny loop. Hot in one pipe and neutral in another twenty feet away form a huge loop, and the inductive reactance of the circuit climbs with it. Under normal load that shows up as extra voltage drop no wire table will explain. Under a fault it is worse: the higher impedance limits the fault current, and a breaker that should have tripped magnetically within a cycle may instead sit in its thermal range for seconds while the arc does its work. Splitting a circuit's conductors does not just heat the pipe. It slows down the one device whose entire job is speed.

Where this goes wrong in the field

Nobody sets out to run the neutral in a separate pipe. It happens at the edges.

A remodel adds a circuit and the existing conduit is full, so someone pulls the new hot through a convenient empty raceway and borrows a neutral from a box across the room. A service upgrade lands each phase through its own knockout in an unslotted steel pull box. A single large conductor leaves a transformer enclosure through a hole punched in a steel wall. A parallel feeder gets one full set of A, B, C, and N in one pipe and a second set that is missing its neutral. Every one of these passes a continuity test, powers the load, and gets inspected on a cold day.

Induction heating scales with current, so a lighting circuit drawing three amps in the wrong configuration will barely register. The same mistake on a feeder drawing two hundred amps is a different story. If you ever find a warm raceway, a warm enclosure wall, or a warm pull box with no loose connection to explain it, induced current belongs at the top of your list of suspects.

Your next moves

  • Clamp the whole circuit, not just the hot. Next time you have a clamp meter out, put the jaws around the hot and neutral together on a running circuit and watch it read zero. Once you have seen cancellation with your own eyes, the rule stops being a sentence and becomes a picture.
  • Walk a service enclosure and look for the slots. Open a piece of gear where large phase conductors enter through separate openings and check for a saw cut joining the knockouts or a nonferrous plate. If each conductor has its own unbroken ring of steel around it, you have found a 300.20(B) problem.
  • Before you pull a new circuit into a full raceway, trace the return path. Decide exactly where the grounded conductor and the equipment grounding conductor for that circuit will physically run. If they will not share the raceway with the ungrounded conductor from source to load, stop and plan a different route or a larger pipe.
  • Add a touch test to your troubleshooting routine. When hunting an unexplained heat source, put an infrared thermometer on the raceway and enclosure walls, not just the terminations. Warm steel with cool connections is a signature, and it points away from torque and toward induction.
  • Audit any parallel run for a missing conductor. Every raceway in a parallel set must carry a full set: all phases, the neutral if there is one, and the equipment grounding conductor. One pipe missing its neutral breaks cancellation for the entire run.

Most separate-raceway mistakes start with the same sentence: the pipe is full. The honest answer in that moment is a conduit fill calculation for the next trade size, a voltage drop check for the longer route, and the ampacity table for whatever derating the new count triggers, all of which take longer to look up in a truck than they take to decide against. Voltly puts conduit fill, voltage drop, box fill, and the NEC ampacity and bending references on your phone, fully offline, so the right route stays faster than the shortcut. If you spend your days in steel pipe, it is worth a look at voltly.lumenlabs.works.