How to Measure and Interpret Phase Imbalance

Why this matters

An imbalance number with no context attached is worth nothing, and it gets quoted anyway. Two techs can measure the same machine an hour apart and report 2 percent and 14 percent, both correctly, because one measured volts at the panel and the other measured amps at an unloaded motor. Worse, the number by itself never says whose problem it is. The reading that ends the job is not the percentage, it is the pair of readings that shows whether the imbalance is arriving from the supply or being made by the motor.

Four things a reading has to name

Write these down before the number, every time, because the number is meaningless without all four:

• The point. Utility service, panel bus, starter line side, or motor terminals. Imbalance can be created between any two of those by a loose connection or an uneven single-phase load, so a percentage without a location cannot be compared with anyone else's.
• The load state. Running under its normal load, or idling. Percent current unbalance rises as load falls, because the balanced component of current shrinks while the unbalanced component barely moves. A reading on an idling motor overstates the problem.
• The quantity, and its base. Volts or amps, and the base is always the average of the three readings taken at that point. The two quantities differ by roughly a factor of six or more on the same machine, so a bare "2.6 percent" tells the next reader nothing.
• Simultaneity. Three readings taken over three minutes on a building whose load is moving are not a set. They have to be simultaneous, or taken close enough together that you can prove the supply did not move between them.

Unbalance itself is one formula for both quantities: the largest deviation of any one reading from the average of the three, divided by that average, as a percentage.

The reading sheet

Take these fields on every imbalance call, in this order. The sheet is the deliverable, not the percentage:

The ratio line is the one people leave out and it is the one that catches bad data. A sibling article establishes that current unbalance typically runs about 6 to 10 times voltage unbalance at or near rated load on a general-purpose induction motor, and that the ratio climbs well above 10 as load falls. A computed ratio far outside that band means either the motor is lightly loaded or one of your readings is wrong, and finding that out on site costs nothing.

Taking readings that are actually simultaneous

The best tool is a three-phase power analyser or a meter that samples all three legs at once, and where the shop has one this problem disappears. With a single clamp meter and a single voltmeter, do it this way:

Take the three voltages in sequence as fast as the leads move, then retake the first one. If the first reading has not moved between the start and the end of the set, the supply held still and the set is valid. If it has moved, the building is doing something and the set has to be repeated at a quieter moment or logged instead.

Do the same on the currents. Retaking the first leg last is the entire quality control on a single-instrument set and it takes fifteen seconds.

All of this is measurement on energized conductors with the machine running. 29 CFR 1910.333(a)(1) permits energized work 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 an imbalance measurement is that case because the energized, loaded condition is the thing being measured. Establish shock and arc-flash boundaries and select PPE on the basis in NFPA 70E-2021, 130.5 and 130.7, which binds through your employer's electrical safety program or a contract rather than on its own; use a meter, leads and clamp rated CAT III or better at or above the voltage present; and keep the clamp jaws, the leads and your hands out of the plane of any coupling, belt or shaft that is turning.

The roll test: supply side or motor side

The percentage tells you how bad. This tells you whose it is.

Set it up de-energized. Open the disconnect, lock and tag it under 29 CFR 1910.333(b)(2) for the electrical exposure and under 29 CFR 1910.147 for the unexpected-startup hazard of the driven machine, and prove the terminals dead with the live-dead-live sequence in NFPA 70E-2021, 120.5. 1910.147 does not cover the electrical exposure itself, which it excludes at (a)(1)(ii)(C); 29 CFR 1926.417 is the construction counterpart for lockout and tagging of circuits.

Roll all three leads by one position, never swap two. L1 goes to the terminal L2 was on, L2 goes to the terminal L3 was on, L3 goes to the terminal L1 was on. A cyclic roll of all three preserves rotation direction; swapping any two reverses it, and reverse rotation destroys scroll and screw compressors, damages some pumps and can wreck a driven gearbox. Before restarting, confirm rotation with a phase rotation meter or by watching the machine's own rotation indicator, on any equipment where reverse rotation is destructive, and do not restart it to "see which way it goes."

Re-run under the same load and re-clamp all three. Then read the result:

• The high current stayed on the same line conductor, meaning it is now feeding a different motor terminal: the imbalance is upstream. The supply, the conductors and the connections between the source and the motor own it.
• The high current moved to the line conductor that now feeds the original hot terminal: the imbalance is in the motor or its terminations. The winding, its connections, or the lugs at the motor own it.

That single test is the difference between quoting a motor and quoting a distribution correction, and it costs one outage window.

A filled-in sheet

A 240 V three-phase motor on a process pump, second complaint of overheating this year.

The arithmetic, so it can be checked: voltages average (240 + 233 + 238) / 3 = 237.0 V, deviations +3.0, -4.0 and +1.0, largest 4.0, so 4.0 / 237.0 = 1.7 percent. Currents average (18.6 + 23.0 + 20.0) / 3 = 20.5 A, deviations -1.9, +2.5 and -0.5, largest 2.5, so 2.5 / 20.5 = 12.0 percent. The ratio 12.0 / 1.7 = 7.1 sits inside the 6 to 10 band, which says the readings are consistent with each other and the machine is loaded near its rating.

Now read the two locations against each other. The starter line side is at 1.4 percent and the motor terminals are at 1.7 percent, so 0.3 percentage points of imbalance are being added between the starter and the motor. That is small and it is not nothing: it is the loaded legs dropping unevenly across the run and its terminations, and it points at the lugs and the conductors between those two points as worth a torque check and an inspection during the next outage.

The other 1.4 percentage points arrived at the starter already formed, and the roll test confirmed the direction: the high current stayed with the line conductor rather than following the motor terminal, so the motor is not making it. That takes the motor off the quote entirely, which is the point of the test given that this machine was going to be replaced for the second time.

What the shop reports is therefore two findings with two owners, not one percentage: a small correctable loss between the starter and the motor, and a 1.4 percent imbalance arriving at the starter that has to be chased upstream into the panel's single-phase load distribution and then, if it is not there, to the service.

The next reading to take, and the cheapest one available, is the three line currents at the panel main with the three-phase equipment switched off. What remains is the building's single-phase load, and if it is spread unevenly across the three legs then the imbalance is being manufactured on the customer's own side and moving circuits fixes it.

Reading the number once you trust it

The interpretation belongs to the machine, not to a universal threshold. A sibling article covers the derating curve in NEMA MG 1, in the edition the motor manufacturer built to and reaching you through the motor's literature, which begins derating at 1 percent voltage unbalance and does not recommend continuous operation above 5 percent. Use it there rather than carrying a remembered percentage between jobs.

Two things are worth stating alongside whatever the curve gives you. First, the comparison that matters is the derated ceiling against the average of the three line currents, not against the highest leg or the nameplate. Second, an imbalance you correct should be re-measured at the same point and the same load state, and the two sheets filed together, because "we torqued the lugs" with no second reading is an assertion rather than a repair.

What makes a reading invalid

A drive between your meter and the motor. The output of a variable frequency drive is not a supply in the sense this method assumes, and an unbalance percentage computed on drive output terminals does not mean what this article means by it. Measure imbalance on the drive's input, and take any question about output-side symmetry to the drive manufacturer's own diagnostics.

An unloaded or lightly loaded machine. Current unbalance measured at idle can be several times the figure the same machine gives under load, and quoting it as the finding overstates the problem badly. If the process cannot be loaded, say the reading was taken unloaded on the sheet.

A set taken during a load change. This is what the retake-the-first-reading habit exists to catch. Two of three legs read before a large load started and one read after is not an imbalance, it is a timeline.

Two legs. There is no unbalance percentage from two readings, because the base is the average of three. A tech who reports imbalance from a two-leg reading has computed a difference and called it something else.

References

• 29 CFR 1910.333(a)(1) and (b)(2) for energized work and de-energizing; 29 CFR 1910.147 for the unexpected-startup hazard of driven machinery, with the electrical carve-out at (a)(1)(ii)(C); 29 CFR 1926.417 for the construction counterpart
• NFPA 70E-2021, 120.5, 130.5 and 130.7, applied through an employer electrical safety program or contract
• NEMA MG 1 for the unbalanced-voltage derating curve, in the edition the motor manufacturer built to, reaching you through the motor's literature
• See related: What a Phase Imbalance Does to a Motor; What Happens When a Phase Is Lost