Why a Transformer Configuration Changes What You Can Measure

Why this matters

A voltage reading is only interpretable if you know the shape of the system that produced it. On one common service, a leg reading 208 V to ground is completely normal and landing a 120 V control transformer on it destroys the control transformer in seconds. On another, a leg reading zero to ground is normal and a leg reading full voltage to ground is also normal, because nothing is grounded at all. Techs get hurt and equipment gets burned not because they measured wrong but because they interpreted a correct reading against the wrong system.

Line-to-line is the load's voltage, line-to-ground is the system's story

Every three-phase load that connects to three conductors cares about the line-to-line voltages and nothing else. Those three readings should be near equal, and a sibling article covers what an inequality between them does to a motor.

Line-to-ground readings do something different. They tell you where, if anywhere, the system is intentionally connected to earth, and therefore which single-phase loads can legally and physically be derived from it. Two systems with identical line-to-line readings can have entirely different line-to-ground readings, and the difference is the whole subject of this article. So the rule is: read line-to-line first to know the load's supply, then read all three line-to-ground to know what kind of system you are standing in front of. The second set is the one you take before you land anything single phase.

The configurations you meet, and what each one reads

Take a 240 V or 208 V class system for the comparison so the numbers are on the same basis:

The two relationships that generate that table are worth holding rather than memorising. In a grounded wye, line-to-ground is the line-to-line divided by 1.732, so 208 over 1.732 is 120 and 480 over 1.732 is 277. In a center-tapped delta the grounded point is the midpoint of one winding, so the two legs on that winding sit at half of 240, and the third corner sits at 240 times 0.866, which is 207.8 V and gets called 208 to ground even though the system is a 240 V system.

The high leg, drawn

        high leg: about 208 V to ground
                |
                B
               / \
      240 V   /   \   240 V
             /     \
            A ----- C
               |
        center tap of the A-C winding,
        bonded to ground
        A to ground 120 V, C to ground 120 V

Every line-to-line reading on that system is 240 V. Two legs give you 120 V to the grounded point and one does not, and there is no electrical marking on the conductor itself that makes the difference obvious once someone has repulled a panel. The NEC as adopted in the edition in force in your jurisdiction requires the high leg to be identified, and specifies where it lands in a panelboard busbar arrangement, in Articles 110 and 408; that identification is a code requirement on the installer and not a physical property you can rely on finding.

Readings that look wrong and are not

Corner-grounded delta. One leg reads at or near zero volts to ground and the other two read full line voltage to ground. A tech who has only worked on grounded-wye systems reads that as a phase-to-ground fault and starts hunting. It is the system working as designed: one corner of the delta is intentionally bonded, so it is at ground potential by definition. The line-to-line readings, all three near equal, are what tells you the system is healthy.

Ungrounded delta. All three legs read something to ground, often roughly similar, and the readings wander when you change meters or when a load switches. Nothing is bonded, so what your meter is reading is the capacitive coupling of the whole system to earth through a high-impedance path, and a digital meter's input impedance is high enough to be part of the circuit it is measuring. Those numbers are not a voltage in any useful sense, which is why systems like this get permanently installed ground detectors instead. On this system a first ground fault does not trip anything and does not announce itself, which is precisely the feature it exists for and precisely the reason a second fault on a different phase is a line-to-line fault through two accidental grounds.

Phantom voltage anywhere. A high-input-impedance meter on a disconnected conductor running alongside energized ones will read a voltage that no load will ever see. This is not a configuration issue but it kills the same diagnosis, so the discipline is the same: a reading that does not survive a low-impedance load applied across it was never a source.

Worked example: three minutes on an unlabelled panel

A shop is asked to add a 120 V circuit for a condensate pump in a building whose panel schedule is illegible. The panel is four wire, so a neutral exists and the assumption in the room is that it must be 208Y/120.

Before any probe goes in, this is energized work. 29 CFR 1910.333(a)(1) permits it only where the employer can demonstrate that de-energizing introduces additional or increased hazards or is infeasible due to equipment design or operational limitations, and determining an unknown system configuration is that case, because the entire purpose of the reading is the presence of voltage. Establish the shock and arc-flash boundaries and select PPE on the basis in NFPA 70E-2021, 130.5 and 130.7, a consensus standard that binds you through your employer's electrical safety program or a contract rather than on its own; use a meter and leads rated CAT III or better at or above the voltage present; keep one hand out of the enclosure and stand to the side of the hinge line while the cover comes off. Once the determination is made and the work starts, open the feeding disconnect, lock and tag it under 29 CFR 1910.333(b)(2), and prove dead with the live-dead-live sequence at NFPA 70E-2021, 120.5. 29 CFR 1910.147 does not govern this exposure, because (a)(1)(ii)(C) excludes work on conductors and equipment in electric utilization installations; the construction counterpart for this work is 29 CFR 1926.417.

Six readings, taken in one sitting:

Line to line, all three land within 1 V of each other, so the supply itself is balanced and healthy. Line to ground is where the system identifies itself: two legs at 120 V and one at 208 V is the center-tapped delta signature from the table above, and the arithmetic confirms it, because 240 times 0.866 is 207.8 V and the meter read 208 V.

That single reading changes the job. A 120 V circuit can be taken from A or from C and not from B. If the only available spaces in the panel land on the B bus, the correct answer is that the circuit cannot go there, not that a 120 V load is close enough to 208 V to survive. And the failure mode is not subtle: a 120 V coil or control transformer across 208 V is at 173 percent of its rated voltage, which is a smoke-and-fail event rather than a slow degradation, and it takes out whatever the coil was controlling on its way.

Note what would have happened without the line-to-ground readings. A four-wire panel with 240 V line to line and a neutral present is exactly what a 240 V wye system would look like from the line-to-line readings alone, and that system would have given about 139 V from any leg to neutral. Both systems have four wires, both have equal line-to-line readings, and only the line-to-ground set separates them.

What changes the answer

A neutral that is present but not connected to a source. In a four-wire panel fed from a three-wire feeder with a neutral run for a subpanel further on, line-to-neutral readings can be floating rather than derived, and they will read as something. Check that the neutral is bonded at the service and continuous back to it before you trust a line-to-neutral number.

A separately derived system inside the building. If the panel you are reading is fed from a step-down transformer in the building rather than from the service, its configuration is set by that transformer's secondary, not by the service. The readings are still the evidence, but the conversation about what can be changed goes to whoever owns that transformer.

A modification you are asked to make. Re-grounding a system, converting a corner-grounded delta, or deriving a neutral where none exists is a design change, not a repair. Whether a proposed change is permitted, and whether it affects the listing of the equipment involved, is settled by the authority having jurisdiction and the equipment manufacturer, not by a technician on site.

Verifying the determination before you land anything

Prove the ground reference is real. A line-to-ground reading is only as good as the point you called ground. Read from each line to the equipment grounding conductor and to the neutral bar separately; on a properly bonded grounded system they will agree closely, and a disagreement means the bond, not the phases, is the finding.

Load-test the point you are about to use. Apply a low-impedance load, or use a solenoid-type tester alongside the digital meter, on the specific line-to-neutral pair you intend to feed the new load from. A digital reading of 120 V that collapses under a real load was a phantom, and finding that out with a tester costs a minute rather than a callback.

Write the configuration on the panel. The single most valuable output of this exercise is a legible note inside the door recording the configuration, the six readings and the date. The next tech's first assumption is the same wrong one you started with, and the note is what breaks the cycle.

References

• 29 CFR 1910.333(a)(1) and (b)(2) for energized work and de-energizing; 29 CFR 1926.417 for lockout and tagging of circuits in construction
• NFPA 70E-2021, 120.5, 130.5 and 130.7, applied through an employer electrical safety program or contract
• NEC Articles 110 and 408 as adopted, in the edition in force in your jurisdiction, for conductor identification and panelboard busbar arrangement on a system with a high leg
• See related: What a Distribution System Is Actually Doing; What a Secondary Voltage Depends On