What a Relay Does That a Switch Does Not

Why this matters

A switch and a relay both make and break a circuit, so techs treat them as the same device with a different trigger. They are not. A relay exists to keep two energy levels apart, and that separation is the whole point: a decision made with a fraction of an amp governs a circuit carrying tens of amps. The practical consequence is that a relay has two independent halves that can fail without each other, and from outside the box they produce the identical complaint. Every wasted relay swap in the trade comes from someone answering "did it work" instead of "which half."

Before you put probes on either half

The contact side of a relay is at line potential whenever the circuit ahead of it is energized, including when the relay is de-energized and sitting open. Open and lock the disconnect under 29 CFR 1910.333(b)(2) for anything you can do dead, including removing the relay, checking terminations and inspecting the contacts. Reading a relay under load genuinely requires it to be energized, which is the troubleshooting exception at 29 CFR 1910.333(a)(1); take that exception with a meter and leads rated at least CAT III for the highest voltage in the enclosure, inspect the meter and leads for damage before use per 29 CFR 1910.334(c)(2), and prove the meter on a known live source before and after per the live-dead-live sequence in NFPA 70E-2021, 120.5, and keep your free hand out of the cabinet.

Two more, specific to this device. Before you touch the load-side terminals of any relay serving a motor, discharge the run capacitor through a proper resistive discharge tool and verify it reads near zero, because a capacitor holds a lethal charge after the disconnect is open and shorting it with a screwdriver damages the plates and throws metal. And never jumper a contact that serves a protective function, which on this equipment means any limit, rollout, pressure, float, interlock or flame-proving contact. If you need to prove a protective contact, meter across it. Bridging it removes the protection while you stand in front of the appliance.

What the relay actually adds

Strip the device to its function and it adds five things a switch does not have:

• Separation of energy levels. The coil circuit and the contact circuit share no conductor. A control transformer of a few tens of volt-amperes commands a circuit sized in tens of amps.
• Remote actuation. The decision can be made anywhere a small conductor reaches, which is what makes thermostats, boards and interlocks possible at all.
• Multiplication. One coil can drive several poles, so one decision closes three circuits at the same instant with no mechanical linkage between them.
• A chosen failure position. Normally open or normally closed is a design decision about where the circuit should sit when the command disappears. A switch's position on power loss is wherever a person last left it.
• A memory of nothing. A relay forgets on power loss, which is a feature when you want a machine to restart in a known state.

And it adds one liability a switch does not have: a coil. The coil is a second component, with its own voltage window, its own current, its own thermal limit and its own way of dying, sitting in series with the decision.

Two failure domains that look identical from the outside

The customer complaint for both halves is "it does not come on." The equipment gives you nothing to separate them, because the relay's state is inferred rather than seen. A switch shows you its handle position; a sealed relay shows you a plastic case.

• Coil side failures: coil open, coil shorted turn to turn, supply voltage below the coil's pull-in threshold, a control path that never closed, an armature or plunger mechanically bound.
• Contact side failures: contacts fully open despite a pulled-in armature (worn contact rivets, a broken return, welded-then-torn material), contacts pitted or oxidized into a high resistance, contacts welded closed so the load never stops.

Note that the last one produces the opposite complaint: the equipment will not turn off. Welded contacts are the one relay failure that presents as too much operation rather than too little, and they are the reason "the compressor ran all night" is a relay question before it is a thermostat question.

The gate that separates them, stated as a rule

Take two readings at the relay itself, per relay, during a single active call, with the load's normal current flowing:

1. Coil voltage, measured across the coil terminals at the moment the call is active.
2. Contact drop, measured across one closed pole, line terminal to load terminal, at the load current you also measure with a clamp.

The gate is reading one. If coil voltage is at or above the coil's own listed pull-in voltage, the command arrived and the contact reading is meaningful. If coil voltage is below that listed value, the contact reading tells you nothing yet, because you are looking at a relay that was never properly asked to close. Resolve the control side first.

Once the gate passes, the contact reading resolves the second half. A healthy closed power contact drops on the order of tens of millivolts at its rated current. Read it on the millivolt range, not the volt range: a healthy contact drops tens of millivolts and a 0.05 V resolution cannot resolve that at all and treat 0.2 volts across a single closed pole at measured load current as the action threshold: below it, the contact is doing its job; at or above it, that pole is a resistance in series with the load and needs replacement, not cleaning. Full line voltage across the pole means the contact is not making at all.

Two appliances, one gate, opposite answers

Both are 240 volt single-phase motor loads that will not start. Both have a 24 volt coil listing a minimum pull-in voltage of 18 volts, read off the part.

The first unit. Coil measures 23.8 volts with the call active. That is above 18, so the gate passes and the command arrived. Across the closed pole, the meter reads 241 volts, which is full line, so the contacts are not making. A clamp on the load conductor confirms zero amps. This relay's coil side is healthy and its contact side is dead. Replacing the thermostat, the board or the control transformer here fixes nothing, and a tech who never took reading two will replace one of them because 23.8 volts at the coil looks like proof the relay is fine.

The second unit. Coil measures 4.1 volts with the call active. That is below 18, so the gate fails and the contact reading is not yet interpretable, whatever it says. The relay is behaving correctly: it was never commanded. Work backwards along the control path for the open. The trap here is the mirror image of the first unit: a tech who reads full line across the contacts concludes the contacts failed and swaps the relay, then finds the new one does exactly the same thing.

The case the binary misses. A third unit passes the gate at 23.9 volts on the coil, the contacts are clearly making because the motor runs, and the customer complains of weak output and a warm panel. Contact drop across one closed pole reads 3.5 volts with a clamped 20.7 amps. That is 3.5 divided by 20.7, or 0.169 ohms of contact resistance, and 20.7 squared times 0.169 is about 72 watts dissipated inside one contact pair. That is 17.5 times the 0.2 volt threshold, and it explains the heat, the reduced output and, eventually, the welded contact that will follow. "It closed" is not the test. The drop is the test.

Ratings are two ratings, and most people read one

A relay carries a coil rating and a contact rating, and they are unrelated numbers that fail independently.

That last row is the one worth stating carefully. For the same contact set, the printed direct-current rating is typically far below the alternating-current rating at the same voltage, because an alternating arc extinguishes at every current zero and a direct-current arc has none to use. How far below is a property of that specific contact set and its arc chute, so read the printed direct-current rating rather than derating by a rule of thumb. Substituting a relay on the strength of its alternating-current number alone, for a direct-current load, produces contacts that weld on the first hard break.

The control-supply half of this arithmetic, including how to tally coil volt-amperes against a transformer's rating and what a pull-in sag looks like on a meter, is worked through in the control-voltage article rather than repeated here.

What changes the answer

Solid state relays and boards with integrated outputs. There is no armature, no mechanical contact and no meaningful contact drop to measure, so the gate above does not transfer. A solid state output has a small forward voltage drop by design and fails shorted at least as often as it fails open, which means the load can run continuously with no command present at all. Read the manufacturer's service data for what the output should measure and treat an unexpectedly running load as a plausible failure rather than a wiring error.

Latching and mechanically held contactors. Their contacts hold without continuous coil power, so a coil voltage reading during a call proves nothing about state and the gate's first step has to be replaced by a check of the holding mechanism.

A relay that tests good on the bench and fails installed. Vibration, an out-of-level mounting or a magnetic field from an adjacent conductor can prevent a healthy armature from seating. Test it in place, in its own orientation, before you condemn the part.

How to verify you got this right

After the repair, run the load through a full cycle and take the same two readings, in the same places, at the same load current. Coil voltage should sit above the listed pull-in with the whole sequence running, and contact drop on every pole should be back to tens of millivolts, not merely lower than it was. Then let the equipment reach operating temperature and repeat the contact drop, because contact resistance climbs with heat and a joint that reads acceptable cold can be well past the threshold twenty minutes in. Check every pole, not the one you measured first: a multi-pole contactor that failed on one pole has usually run all its poles the same number of cycles.

References

• 29 CFR 1910.333(a)(1) and (b)(2), the energized-work troubleshooting exception and the electrical lockout requirement for general industry
• 29 CFR 1910.334(c)(2), visual inspection of test instruments and leads for external damage before use
• NFPA 70E-2021, 120.5, live-dead-live verification
• Manufacturer documentation for coil pull-in voltage, contact current, horsepower and direct-current contact ratings
• See related: How Control Voltage Differs From Line Voltage; How Switches and Relays Fail