Troubleshooting
Diagnosing current imbalance across every motor on a service
Short answer
Compute unbalance per NEMA MG-1 as maximum deviation from the three-phase average, not the min-to-max spread: a 13% spread is 6.5% unbalance. Identical A-mid, B-low, C-high readings on ten different motors mean the asymmetry is upstream of the machines, so measure voltage unbalance and negative sequence at the motor terminals and current sharing in the parallel feeder sets, not ten motor circuits.
On this page
- Is a 13% spread between highest and lowest phase current really 13% unbalance?
- How much current unbalance does a small voltage unbalance cause?
- Why does the same phase pattern on every motor point upstream?
- Where do I measure voltage to see what the motors actually receive?
- Do equal voltage magnitudes mean the three phases are balanced?
- Can five parallel 500 kcmil sets in steel conduit cause this?
- How do I rule out high resistance in the individual motor circuits?
- Does 6.5% current unbalance force motor derating?
- Questions asked
Ten refrigeration compressors from 75 to 300 hp on one 2000 A, 480 V wye service, and every one reports the same picture: A-phase current in the middle, B lowest, C highest, with about 13% between the low and high readings. That repetition is the diagnosis. A winding fault, a loose lug, or a worn contactor pole shows up on one machine; ten machines of different frames and loads sharing an identical phase ordering are all seeing one asymmetry imposed from upstream. Stop testing motors and work the service: get the arithmetic onto the NEMA basis, then measure what the motors actually receive.
Is a 13% spread between highest and lowest phase current really 13% unbalance? #
No. ANSI/NEMA MG 1-2016 (Revised 2018) defines unbalance as the largest deviation from the average of the three readings, and on that basis a 13% low-to-high spread is 6.5% current unbalance, because the spread runs roughly double the MG-1 number whenever the middle phase sits near the average.
percent unbalance = 100 × largest deviation from the three-phase average ÷ that average
Run the service voltages through it first. Readings of 481 V, 484 V and 482 V average to (481 + 484 + 482) ÷ 3 = 1447 ÷ 3 = 482.33 V. The deviations are 1.33 V, 1.67 V and 0.33 V; the largest, 1.67 V, sits on the 484 V phase. That gives 1.67 ÷ 482.33 × 100 = 0.35% voltage unbalance at the entrance.
On the current side, as an illustrative reconstruction: with a 100 A phase average, 93.5 A on B and 106.5 A on C puts 13 A between low and high, which is 13% of average, while the largest deviation is 6.5 A, or 6.5%. That is the figure derating curves and relay comparisons use, and the one carried through the rest of this diagnosis. Log all three phases at the same instant on the same instrument before computing anything; a relay display subtracted from a handheld clamp reading is two measurements, not one. These percentages are spot estimates that move with load and instrument, so re-check them against the motor and relay documentation before acting.
How much current unbalance does a small voltage unbalance cause? #
About six to ten times the voltage unbalance, so 0.35% at the service entrance predicts roughly 2.1% to 3.5% current unbalance, and the measured 6.5% is about double the top of that band.
The multiplier exists because unbalanced voltage contains a negative-sequence component that spins a reverse-rotating air-gap field, and the rotor answers it with close to locked-rotor impedance rather than running impedance: a fraction of a percent of voltage becomes several percent of current, and that current heats the rotor instead of turning the shaft. A walkthrough of calculating motor current asymmetry from voltage unbalance carries the sequence-component detail behind the ratio.
Two assumptions are hidden in the six-to-ten rule, and both are open here: the voltage is supposed to be measured where the motor sees it, and the machine is supposed to be near rated load. The next measurements close those gaps.
Why does the same phase pattern on every motor point upstream? #
Because motor-local faults do not repeat across machines: ten different horsepowers all reporting A in the middle, B low and C high means one asymmetry is being fed to all of them in parallel, upstream of every branch circuit and starter.
A machine with its own problem stands out against the fleet instead of matching it. A motor running over nameplate amps after a soft-starter bypass is a one-circuit investigation, not a service one. The corroborating facts lean the same way: the service metering, the compressor microprocessor protectors and a true-RMS clamp meter agree on the pattern, the neutral current is zero, and the current is pure 60 Hz with no harmonic content. None of that alone separates a supply cause from a feeder cause from a load-side cause; the measurements below do.
Where do I measure voltage to see what the motors actually receive? #
At the load side of the starters with the motors running: service-entrance readings show what the utility delivers, not what survives the parallel feeder, the switchboard and MCC bus, the branch breaker and the contactor poles.
These are energized readings on running equipment, so use metering rated for the location and take them only if you are qualified for energized measurement. Then work the chain:
- Read A-B, B-C and C-A at the service entrance with all compressors running and compute the MG-1 unbalance.
- Read the same three at the MCC horizontal bus; the difference between the two sets is the unbalance added across your feeder.
- Read the same three at the load side of the starter on the largest and the smallest compressor, which is as close to the motor terminals as most plants can reach.
- Repeat the entrance reading with the compressors staged off. Unbalance that shrinks with load is generated downstream of the transformer by unequal impedance in your own conductors; unbalance that holds constant is a source characteristic.
If the starter-load unbalance reaches 1% while the entrance holds its low reading, the asymmetry is manufactured between those two points and the parallel feeder is the prime suspect. If it stays under 0.5%, magnitudes are not the mechanism and the phase angles are next.
Do equal voltage magnitudes mean the three phases are balanced? #
No: balance requires equal magnitudes and 120-degree spacing, and a few degrees of angular displacement produce negative-sequence voltage that no magnitude reading reveals, which drives exactly this current pattern.
Use a power quality analyzer or a meter that reports symmetrical components, and record the negative-to-positive-sequence voltage ratio, V2/V1, at the service entrance. The per-phase power factors already measured there, 0.72 on A, 0.61 on B and 0.68 on C, are the supporting tell. A three-phase induction machine is a symmetrical device and does not genuinely present three different load power factors; a 0.11 spread across phases is angular asymmetry appearing as an apparent displacement difference. The same readings say the compressors run light, and light load makes current unbalance read worse for a given voltage unbalance. The harmonic sweep that showed pure 60 Hz ruled out harmonic currents and nothing else; phase-angle asymmetry is invisible to it.
A V2/V1 above roughly 0.5% with balanced magnitudes points at the supply: an unbalanced single-phase load elsewhere on the utility circuit, unequal transformer winding impedance, or unequal line impedance upstream of you. That becomes an engineering conversation with the utility, with your sequence data on the table, not a plant repair.
Can five parallel 500 kcmil sets in steel conduit cause this? #
Yes, in two ways: asymmetric geometry among the sets changes each conductor's mutual reactance so the five sets per phase share unequally, and unequal lengths put unequal impedances in parallel, so the feeder stays on the suspect list until you clamp it.
NFPA 70, the National Electrical Code, 2023 edition, requires paralleled conductors to be the same length, the same conductor material, the same size and the same insulation type, terminated in the same manner, with the installation arranged so the sets share current equally. Confirm the requirement with the local authority having jurisdiction and the equipment documentation. Five sets per phase, 40 ft, in underground steel conduit, with lengths that vary by routing, is a mismatch on the length requirement before any clamp comes out.
Fifteen conductors decide it:
- Clamp each one individually and log it by set and phase.
- Sum the five readings per phase and compare with the metered phase current; agreement validates the clamping before you interpret anything.
- Within each phase, compute each set's deviation from the five-set average. More than about 10% spread means the sets are not sharing.
- Compare the pattern across phases: B low in one set and C high in another is the geometry problem showing itself.
- Verify every conduit carries a complete set of phases; a lone conductor in steel conduit picks up reactance and induced heating that no other phase sees.
The fix is geometric: regroup so each raceway holds a complete A-B-C set, bring the lengths to match, and clean and re-terminate while the conductors are open.
How do I rule out high resistance in the individual motor circuits? #
Lock the MCC out, test for absence of voltage, and only then measure line-to-line resistance from the load side of each contactor through the motor windings with a low-resistance ohmmeter; under 5% deviation from the average of the three readings is normal winding and cable scatter.
Above 5%, chase the outlier back through the terminations until it moves. This is why the clean infrared scan settled so little: a camera sees the heat of a joint carrying more than its share, not a cool joint carrying less, and a resistance measurement sees both. Then restore the service and start the compressors one at a time, recomputing each machine's deviation-from-average unbalance. If the identical A-middle, B-low, C-high ordering appears on every machine no matter which ones run, the cause is imposed. If one machine's pattern changes when it runs alone, that circuit carries its own contribution on top of the service's.
Does 6.5% current unbalance force motor derating? #
Not by itself: MG-1 ties required derating to voltage unbalance at the motor terminals, where successful operation at rated load is assured up to 1% unbalance, with derating above that and roughly a 0.95 factor at 2%, so the derating decision waits on the starter-load voltage readings, not on the current spread.
The decisions by finding:
- Starter-load voltage unbalance at or under 1%: run at rated load and monitor, and re-check at full load, because the light loading visible in the power factor inflates the current reading.
- Starter-load unbalance above 1%: derate per the MG-1 curve and fix the plant-side asymmetry, meaning feeder grouping, lengths and terminations, then re-measure.
- Balanced magnitudes with V2/V1 above 0.5%: take the sequence data to the utility as a supply-asymmetry case.
- Parallel-set spread past the 10% line: regroup and equalize the feeder before anything else is touched.
- One machine deviating well beyond the fleet: that circuit gets the milliohm follow-up on its own.
Check the motor nameplate and documentation for the machine's own unbalance capability, since applied standards and machine ratings vary.
Questions asked
Why is the neutral current zero if the phase currents are unbalanced?
Because the motors are three-wire loads: current that leaves on one phase returns on the other two, so unbalanced phase currents produce no neutral current at all. Zero amps on the neutral neither clears the supply nor implicates it; it only confirms there is essentially no line-to-neutral load on the service.
Could power factor correction capacitors produce the same pattern?
Yes, if any are installed: a blown fuse on one phase of a fixed capacitor bank leaves the surviving units unevenly distributed by phase and imposes a fixed phase ordering on every downstream motor. That makes capacitor fuses the first check whenever correction capacitors exist. This service has none, so that path is closed here.
Why does current unbalance read worse at light load?
Positive-sequence current falls with load while negative-sequence current, set by the supply asymmetry, barely changes, so their ratio grows as the machine unloads. Per-phase power factors of 0.61 to 0.72 on motor-only load indicate lightly loaded compressors, so part of the 6.5% reading is load level rather than defect. Trust the reading taken at full load.
Can I trust the unbalance percentage my meter displays?
Not until you know its formula. Metering and relay displays compute unbalance by different methods, including deviation from the average, deviation from a reference phase, and ratios of sequence components, and two methods can differ by a factor of two on the same readings. Check the meter documentation for the definition it uses before comparing its percentage to any MG-1 limit.