Harmonics and power quality field guide for electrical crews

What harmonics are, and why a modern building has them

Harmonics are currents and voltages at whole-number multiples of the 60 Hz fundamental: 180 Hz is the 3rd, 300 Hz is the 5th, 420 Hz is the 7th, and so on up the spectrum. A clean power system is a single 60 Hz sine wave. Add harmonics and that sine wave gets distorted into a flattened, peaky, or notched shape that still cycles 60 times a second but is no longer a clean sine.

The source is the load, not the utility. A resistive load like a heater or an incandescent lamp draws current as a smooth sine that follows the voltage. A nonlinear load draws current in short pulses, gulping near the voltage peak and drawing almost nothing the rest of the cycle. That pulsed current is a 60 Hz fundamental plus a stack of harmonic currents riding on top of it. The load injects those harmonic currents back into the system, and as they push through the impedance of the wiring and the transformer, they distort the voltage everyone else on that bus has to use.

This is why the problem grew up with electronics. A building full of computers, drives, LED drivers, and chargers is a building full of nonlinear loads, and the harmonic currents they make do real damage: they overheat the neutral and the transformer, trip breakers that read more current than a clamp on the phase suggests, and pull equipment below the voltage quality it needs. The harmonics were not designed in. They came in the door with the equipment.

The nonlinear loads that make the harmonics

Almost everything with a power-electronic front end is a harmonic source, because the first thing it does with your AC is rectify it to DC through a diode or transistor bridge that conducts in pulses. The list is most of a modern electrical bill.

Variable frequency drives are usually the heaviest single offender in a commercial or industrial building, because the six-pulse rectifier at the front of a bare drive can produce current distortion in the range of 80 to over 100 percent of the fundamental, and there are often a lot of them. UPS systems and battery chargers rectify the same way on their input. The switch-mode power supplies inside servers, PCs, and IT gear each draw a pulsed current, and a data hall stacks thousands of them. LED drivers and electronic ballasts are small individually but numerous, and they are heavy on the 3rd harmonic. EV chargers are the fast-growing one: they are rectifiers feeding a battery, and a bank of them on a long feeder is a harmonic source and a voltage-drop problem at the same time.

The practical read is that nonlinear load is now the majority of the connected load in many buildings, not a special case. When you walk a facility looking for the harmonic source, start at the largest drives and the UPS, then count the distributed electronic load. The big concentrated rectifier and the thousand small supplies are two different problems with two different fixes.

What is the difference between voltage THD and current TDD?

Voltage THD and current TDD are both distortion percentages, but they measure different things against different baselines, and IEEE 519 holds the customer and the utility to each one separately. Mixing them up is the most common way a power-quality conversation goes wrong.

Total harmonic distortion, THD, is the root-sum-square of the harmonic content expressed as a percent of the fundamental at the moment you measure. Voltage THD describes how distorted the voltage waveform is, and it is the number that tells you whether the bus quality is good enough for the equipment on it. Current THD has a trap: at light load the fundamental current is small, so the same harmonic amps read as a huge percentage, and a lightly loaded drive can show 80 percent current THD while doing almost nothing to the system.

Total demand distortion, TDD, fixes that trap. TDD divides the harmonic current by the maximum demand current over a window, not by the instantaneous fundamental, so it does not punish you for harmonics during light loading. IEEE 519 writes its current limits as TDD for exactly this reason. The rule of thumb: judge the voltage by THD, judge the current you inject by TDD, and never quote a current THD figure without saying what the load was when you took it.

The harmonic orders: 3rd, 5th, 7th, and the triplens

Harmonics come in a spectrum, but a handful of orders carry most of the trouble, and which orders show up tells you what is making them. The orders divide into three sequence groups, and the sequence is what decides where the harmonic does its damage.

Three-phase rectifiers, the six-pulse front end on most drives and UPS, produce the characteristic orders 5th, 7th, 11th, 13th, and up, with the amplitude falling off roughly in inverse proportion to the order. The 5th is a negative-sequence harmonic, so its rotating field turns backward against the motor and adds heat and braking torque. The 7th is positive sequence and turns forward. Both end up as extra loss and heat in motors and in the iron of transformers.

The triplens are the third group and the special case: the 3rd, 9th, 15th, and the rest of the multiples of three. They are zero-sequence, which means all three phases of the 3rd peak at the same instant and in phase with each other. Single-phase rectifiers, the switch-mode supply and the LED driver, are heavy 3rd-harmonic sources. Knowing which orders dominate is the first diagnostic: a spectrum loaded with 5th and 7th points at three-phase drives, while a spectrum loaded with the 3rd points at single-phase electronic load, and the fix is different for each.

Why is my neutral overloaded or running hot?

An overloaded neutral on a three-phase four-wire system is almost always triplen harmonics, and the third harmonic is usually the culprit. On a balanced system the three phase currents are 120 degrees apart and the fundamental cancels in the shared neutral, which is why the neutral is normally the quiet conductor. The triplens break that cancellation.

Because the 3rd harmonic is zero-sequence, the third-harmonic current on all three phases is in phase, so instead of canceling in the neutral it adds. With a panel full of single-phase electronic loads, the neutral can carry well above the phase current, and in the worst case the third-harmonic neutral current approaches 173 percent of the phase current. The shared neutral that was sized to match the phases is now the most heavily loaded conductor in the raceway, and it is carrying current the phase conductors never see.

Here is the part that hurts: there is no overcurrent device on the neutral. The breaker watches the phases, so an overloaded neutral does not trip anything. It just heats, quietly, until the insulation cooks, a termination fails, or you smell it. You find a hot neutral by clamping it under real load and comparing it to the phases, and if the neutral reads near or above the phase current with electronic load on the panel, the triplens are your problem. The fix is a full-size or oversized neutral on these circuits, often a doubled or 200 percent neutral, plus attention to the transformer feeding it.

Transformer heating and the K-rated transformer

Harmonic currents heat a transformer far harder than their amps alone suggest, because eddy-current losses in the windings rise with the square of the harmonic frequency. A standard transformer fed a heavy nonlinear load runs hot at a load well below its nameplate kVA, loses insulation life, and in a bad case becomes a fire risk. This is the same triplen current that overloads the neutral, now circulating and heating the transformer that feeds the panel.

The answer is a K-rated transformer, built and tested for harmonic load. UL recognizes the K ratings K-1, K-4, K-9, K-13, K-20, K-30, K-40, and K-50 under UL 1561, with the K-factor measuring how much harmonic heating the unit is built to take. A K-rated unit uses larger or specially formed conductors, more cross-section in the windings, and a neutral bus commonly rated at 200 percent of full-load current to carry the additive triplens. A K-13 is a common choice for general office and computer load; heavier drive-dominated load points toward K-20 and up. The project load study sets the K-factor.

The alternative to specifying a K-rated unit is derating a standard one, and the penalty is steep: a standard transformer feeding single-phase electronic load can need 30 to 50 percent derating to keep from overheating, so you buy and house a much larger unit to do the same work. For sizing the transformer itself, the secondary protection, and the separately derived grounding around a dry-type unit, see the dry-type transformer guide; this guide stays on the harmonic side of the choice.

What is IEEE 519?

IEEE 519 is the recommended practice for harmonic control in power systems, and the 2022 edition is the current one. It sets distortion limits at the point of common coupling, the PCC, which is the boundary where the utility's system meets the customer's, commonly the service entrance or the revenue meter. The whole standard turns on a split of responsibility at that point.

The utility owns the voltage and the customer owns the current. The customer is expected to hold the harmonic current they inject at the PCC, expressed as TDD, within the limit set for their short-circuit ratio. The utility is expected to hold the voltage distortion at the PCC within its limit by keeping the system impedance in check. As a common reference, voltage limits often cited are individual harmonics around 3 percent and total around 5 percent on systems above 1 kV up to 69 kV, with a higher allowance, around 8 percent total, at low voltage; the current TDD limit varies with the ratio of available short-circuit current to load current, with stiffer systems allowed more.

Two cautions before you quote a number on a submittal. First, IEEE 519 is a recommended practice, not a law on its own, so it controls when a utility tariff, an interconnection agreement, or the project specification adopts it, which is increasingly common for large or generator-paralleling services. Second, the limits and the tables are edition-specific and apply at the PCC, not at every panel inside the building, so confirm the edition the project invokes and where the PCC actually sits before you measure against it.

How do you measure and find harmonics?

You measure harmonics with a power quality analyzer or a permanent meter that samples the waveform fast enough to resolve the individual orders, not with an averaging multimeter. The analyzer captures the current and voltage waveforms, runs the spectrum, and reports THD, TDD, the individual orders, and often the waveform shape and the K-factor of the load. IEEE 1159 is the recommended practice for how to monitor and characterize power quality, and modern analyzers follow the IEC 61000-4-7 method of measuring harmonics over short windows and aggregating them, typically up to the 50th order.

Finding the source is the real skill, and it is a hunt up the tree. Distortion is worst near the offending load and dilutes toward the stiffer source, so you measure at the PCC for the compliance question, then move down into the building toward the panels and feeders where the current TDD climbs. The phase that is hottest, the neutral that is overloaded, and the spectrum that is dominated by the 5th and 7th versus the 3rd all point you at the load type. A clamp on the neutral that reads near the phase current with electronic load present is the triplen tell without even opening the spectrum.

On a permanent install, the metering already on the gear can do this watching for you. A data center or critical facility usually has an EPMS with revenue and check meters that log harmonics continuously, so the trend shows you a rising distortion problem before it becomes a hot transformer. For how that monitoring hierarchy is set up and verified, see the EPMS and power monitoring guide. The handheld analyzer finds the problem once; the permanent meter catches it as it grows.

The mitigation methods, and how to choose

There are two honest strategies for harmonics: tolerate them with gear built to take the heat, or reduce them at or near the source. Most real projects use both, and the choice is driven by where the harmonics come from and how much you have to cut.

Tolerating means sizing for the harmonics you accept: the K-rated transformer, the oversized or doubled neutral, conductors and gear rated for the extra heating. It does not lower the distortion, it survives it, and it is the right call when the load is distributed and modest and chasing the source would cost more than living with it. Reducing means treating the current: a line reactor or DC choke on each drive as the cheap first step, a passive tuned filter for a known dominant order, an active harmonic filter for a mix of orders and changing load, or a multi-pulse or phase-shifted rectifier that cancels harmonics inside the equipment.

The decision tree is short. One or a few big drives: reactors first, then a passive filter or an 18-pulse drive if you need to hit a hard limit. A whole bus of mixed nonlinear load that has to meet IEEE 519 at the PCC: an active filter sized for the total, because it adapts to whatever the load throws. Distributed small electronic load heating a neutral and transformer: K-rated transformer and oversized neutral, because there is no single source to filter. The wrong move is buying one expensive box for a problem that a few reactors and a properly sized neutral would have solved.

The line reactor and DC choke on a drive

A line reactor is the cheapest harmonic fix on a drive, and it is the one that should already be there. It is a three-phase series inductor ahead of the rectifier, commonly 3 to 5 percent impedance, that spreads the rectifier's current pulse out in time so the peaks are lower and the waveform is less spiky. A DC link choke does the same job on the DC side inside the drive. Many drives ship with one or accept one as an option.

The numbers are honest about what it buys you. A bare six-pulse drive can sit near 80 to over 100 percent current distortion; a 3 to 5 percent reactor pulls that down into the range of 35 to 40 percent. That is a large cut for a small box, and it comes with side benefits: the reactor protects the drive's input rectifier and DC capacitors from voltage spikes and notching and lengthens their life. For that reason a reactor earns its place even when harmonics are not the headline concern.

What a reactor will not do is get you to an IEEE 519 limit on its own. Thirty-five percent current distortion is far better than 100, but it is nowhere near a 5 to 8 percent TDD target at the PCC. Treat the reactor as the mandatory first step and the floor under the drive, not the whole answer. If a hard limit is in the spec, the reactor goes in and then a filter or a multi-pulse front end does the rest.

The passive harmonic filter and the resonance caution

A passive harmonic filter is a tuned circuit of inductors and capacitors built to absorb a specific harmonic order, usually the 5th, by presenting a low impedance at that frequency so the harmonic current flows into the filter instead of into the system. Sized and tuned to a drive's current rating, a good passive filter can bring current distortion down into the single digits, often the 5 to 8 percent range, which is enough to meet many limits. It is passive, durable, and has no electronics to fail.

The catch is that a passive filter is a fixed tuned circuit, and tuned circuits resonate. The same capacitance that traps the target harmonic forms a resonant pair with the system inductance at some other frequency, and if that resonant point lands on a harmonic the load is making, the filter amplifies it instead of absorbing it. Filters are designed to keep that resonance away from the active orders, but change the system, such as a new transformer, a different available fault current, or capacitors added elsewhere, and you can move the resonance onto a harmonic and make things worse.

Two practical rules follow. A passive filter is matched to a specific load and a specific system, so it is best on a steady, known load like a single large drive, and it does not adapt when the load profile changes. And any passive filter or power-factor capacitor added to a building with harmonics has to be checked for resonance against the existing spectrum, not just dropped in. The detuned reactor that solves this on a capacitor bank is the next section's subject.

The active harmonic filter

An active harmonic filter is a power-electronic device that measures the harmonic current the loads are drawing and injects an equal and opposite current in real time, canceling the harmonics so the supply upstream sees a clean fundamental. Because it synthesizes the correction cycle by cycle, it adapts to whatever orders and amplitudes the load presents, and it can hold current distortion to under 5 percent across a changing mix of loads.

That adaptability is what makes it the tool for a whole bus rather than a single machine. Put one active filter at a panel or switchboard feeding a dozen drives, a UPS, and assorted electronics, and it corrects the lot, sized to the total harmonic current rather than to any one load. It does not resonate the way a passive filter can, because it is not a fixed tuned circuit, and it can also trim displacement power factor as a bonus. For meeting IEEE 519 at the PCC on a complex, varying load, it is usually the cleanest path.

The cost is the obvious downside, both the unit price and the fact that it is active electronics that need cooling and can fail, unlike a passive reactor. The economics favor an active filter when several nonlinear loads share a bus, because one unit covers many sources; they favor reactors and passive filters when one or two large drives dominate. Size the active filter to the measured or calculated harmonic current with headroom, not to the load kVA, because it only has to supply the harmonic component.

Multi-pulse drives and phase-shift cancellation

Multi-pulse rectification cancels harmonics inside the drive by splitting the rectifier into sections fed through a phase-shifting transformer, so the harmonic currents from each section are out of phase and cancel each other before they reach the line. A 12-pulse drive uses two six-pulse bridges fed 30 degrees apart, which cancels the 5th and 7th and leaves the 11th and 13th as the lowest orders. An 18-pulse drive uses three sections at 20-degree shifts and cancels everything up through the 13th, leaving the 17th and 19th as the first significant orders.

The result is low distortion built into the equipment with no separate filter to tune or maintain. An 18-pulse drive can meet tight limits on its own, which is why it shows up on large, critical drives where a clean front end is part of the spec. The cancellation is real and reliable when the supply is balanced.

The weakness is exactly that condition. Multi-pulse cancellation depends on a balanced supply voltage, and it falls apart fast when the voltage is unbalanced. On an 18-pulse drive at part load, pushing the voltage unbalance from 0 to 3 percent can drive current distortion from around 10 percent up toward 35 percent, erasing most of the benefit. So multi-pulse is excellent on a stiff, balanced service and a poor bet on a soft or unbalanced one, and a phase-shifting transformer feeding paired standard drives is a way to get similar cancellation across two drives that would otherwise both inject the 5th and 7th.

Power factor versus harmonics, and the capacitor trap

Power factor and harmonics get tangled because the word covers two different things, and the difference decides whether a capacitor bank helps or hurts. Displacement power factor is the old, familiar one: the phase angle between the fundamental current and voltage, the lagging power factor of motors that capacitors were invented to correct. True power factor is the ratio of real power to total apparent power including the harmonics, so it folds in the distortion the nonlinear loads add.

On a building heavy with nonlinear load, true power factor is lower than displacement power factor because the harmonics carry apparent power that does no work. The trap is reaching for the classic fix. A plain capacitor bank corrects displacement power factor by adding capacitance, but in a system full of harmonics that capacitance forms a parallel resonant circuit with the supply inductance, and if the resonant frequency lands near a harmonic the loads are making, the bank amplifies that harmonic. The capacitors then carry huge harmonic current, overheat, and fail, and the voltage distortion gets worse instead of better.

The rule is blunt: do not drop a bare power-factor capacitor bank into a building with significant harmonics. If you need power-factor correction there, use a detuned bank, which puts a series reactor on each capacitor stage to tune the combination below the lowest harmonic, commonly around 227 Hz on a 60 Hz system (the 189 Hz figure is the 50 Hz equivalent), so the bank stays inductive at every harmonic frequency and cannot resonate. And remember that capacitors fix displacement, not distortion: to improve true power factor you have to cut the harmonics with a filter, not add capacitance.

What are the symptoms of a harmonics problem?

A harmonics problem rarely announces itself as harmonics. It shows up as a list of nagging complaints that get blamed on the equipment, and the common thread is heat and current that does not add up.

The classic tells: a neutral conductor running hot or reading near or above the phase current, a transformer overheating at a load below its nameplate, and breakers that nuisance-trip because the true RMS current including harmonics is higher than a clamp on the fundamental suggests. Motors run hot and lose life from the negative-sequence 5th turning their field backward, with the trouble usually starting once voltage distortion climbs into the 8 to 10 percent range. Power-factor capacitors fail early or fuse out from resonant harmonic current.

On the electronic side you get equipment misbehaving for no clear reason: drives faulting, UPS dropping to battery on clean utility, controls and sensors acting intermittently, and flicker or light dimming. The distortion can also show up at the source as voltage notching from the rectifier commutation. None of these alone proves harmonics, but a hot neutral plus a hot transformer plus nuisance trips in a building full of drives and electronics is the pattern, and a power quality analyzer turns the suspicion into a spectrum.

Harmonics in data centers and on the generator

A data center is a dense field of nonlinear load: thousands of server power supplies and the UPS systems feeding them, all rectifying. The IT load is the harmonic source, and the UPS input can present significant distortion back upstream depending on its design, so the whole power chain from the PDU to the utility has to be planned for harmonics rather than surprised by them.

The hard part is the generator. A utility source is stiff, with low impedance, so it absorbs harmonic current with little voltage distortion. An alternator is a much softer source with higher impedance, so the same nonlinear load that looked fine on utility can drive the voltage distortion on the generator high enough to trip the UPS or destabilize the bus when the facility transfers to standby. This is why generators feeding nonlinear load get oversized and specified with care.

The practice that holds up: oversize the alternator rather than the engine, because the alternator size sets the impedance and the fault and step-load behavior, and mandate features that handle nonlinear load, commonly a 2/3-pitch winding and PMG excitation for reliable excitation and fault clearing under distorted load. A common starting point is sizing the generator kW well above the UPS rating for the harder UPS types and confirming the voltage distortion the UPS will tolerate on generator. The number that matters is the ratio of nonlinear load to generator capacity: a small nonlinear load on a large alternator barely distorts, while a generator sized tight to a UPS-dominated load can chase its own tail on transfer.

Designing and monitoring for harmonics instead of chasing them

Harmonics are cheapest to handle on the drawing, and most expensive to handle after the building is hot. The retrofit, opening up to add filters, pulling a second neutral, swapping a cooked transformer, costs many times what the same decisions cost at design when they are just a line item.

The design moves are known. Size the neutral on panels feeding single-phase electronic load as full or oversized, often a 200 percent neutral, because the triplens are coming whether you plan for them or not. Specify the transformer's K-factor from a real load study of the connected nonlinear load, not a guess, so the K-13 or K-20 matches the building. Put line reactors on drives as a baseline. Where a hard limit applies at the PCC, do the harmonic study early and design in the filter, active or passive, as part of the gear rather than as a field afterthought.

The estimate-versus-reality gap is real here too. A design that ignored harmonics prices a standard transformer and a normal neutral, and the field discovers the heat later, so the fix becomes a change order or a callback nobody budgeted. Catch it in the load study, where the nonlinear fraction of the load is a number you can see, and the mitigation lands in the base scope instead of the punch list.

Design is not the end of it, because the load that makes harmonics keeps changing as the building fills up and equipment is added. Permanent metering catches that drift: a meter at the service and the major feeders that logs THD, TDD, and the individual orders turns harmonics from a periodic survey into a trend, so a rising problem shows up as a slope on a chart before it shows up as a hot transformer. On critical facilities this lives in the EPMS, where the same metering hierarchy that handles billing and capacity also watches power quality; see the EPMS and power monitoring guide. Without a full EPMS, an annual or post-change survey with a portable analyzer is cheap insurance, because the harmonic picture is a function of what is plugged in, and that never stops changing.

Field checklist

What to document

A harmonics finding that nobody wrote down is a finding that gets rediscovered the hard way when the transformer overheats again. The record ties the load to the distortion it makes and the fix that was applied, so the next person can see why the neutral is doubled and the transformer is a K-20.

Capture the load and its nonlinear fraction, the harmonic source identified, the measured voltage THD and current TDD with the load condition at the time, the dominant orders, the limit being held to and where the PCC is, the mitigation chosen and its rating, and who measured it. If a filter or K-rated transformer was specified, record the basis so a future change to the load can be checked against it.

Common mistakes

• Sharing a neutral sized for the phases on panels full of single-phase electronics, so the triplens overload it.
• Feeding heavy nonlinear load through a standard transformer that then overheats below its nameplate.
• Dropping a bare power-factor capacitor bank into a harmonic-rich system and causing resonance.
• Skipping the IEEE 519 study where a utility or spec requires it, then failing at the PCC.
• Undersizing the mitigation, like calling a line reactor the whole fix when a hard limit needs a filter.
• Quoting current THD without the load condition, so a lightly loaded drive looks like a disaster.
• Never finding the actual source and treating symptoms at the panel instead of the load.
• Specifying a multi-pulse drive on a soft or unbalanced supply where its cancellation falls apart.

Standards and references

IEEE 519, with the 2022 edition current, is the recommended practice for harmonic control, setting voltage and current distortion limits at the point of common coupling and splitting the responsibility so the customer holds the injected current and the utility holds the voltage. It is a recommended practice, so it governs when a utility tariff, an interconnection agreement, or the project specification adopts it; confirm the edition and the PCC location before measuring against it.

IEEE 1159 is the recommended practice for monitoring power quality, covering how harmonics and other disturbances are measured and characterized, and modern analyzers follow the IEC 61000-4-7 measurement method. K-rated transformers are built and tested under UL 1561, with the recognized ratings K-1 through K-50, and the harmonic loss derating math behind the K-factor traces to ANSI/IEEE C57.110. The NEC, NFPA 70, governs the conductor and neutral side, including treating the neutral as a current-carrying conductor where triplens are present and sizing it accordingly; the article and section numbers shift between code cycles, so confirm them against the adopted edition.

Equipment listings and the manufacturer's instructions for drives, filters, UPS, and transformers can set tolerances and requirements tighter than any general standard, and where they do, the listing controls. Cite the standard that actually governs the point, and let the utility agreement and the project specification set the limit when they are stricter than the recommended practice.

Units, terms, and conversions

Harmonics carry a vocabulary that reads differently across an analyzer screen, a spec, and a utility letter, so the same idea shows up under several names.

Distortion is given as a percent: THD against the fundamental, TDD against maximum demand. Harmonic order is the multiple of the 60 Hz fundamental, so the 5th is 300 Hz. The point of common coupling is the PCC. Power factor splits into displacement power factor, sometimes DPF, and true or total power factor that includes distortion. The K-factor rates a transformer for harmonic heating. Frequency is hertz; on 50 Hz systems the orders land at multiples of 50 instead, so confirm the system frequency before reading a foreign spec.

Harmonic order

Whole-number multiple of the fundamental; the 5th is 300 Hz on a 60 Hz system

THD

Total harmonic distortion, the harmonic content as a percent of the fundamental

TDD

Total demand distortion, harmonic current as a percent of maximum demand current

Triplen

A multiple of the 3rd harmonic; zero-sequence, so it adds in the neutral

PCC

Point of common coupling, the boundary between utility and customer where IEEE 519 applies

K-factor

A transformer's rating for handling harmonic heating, from K-1 to K-50 under UL 1561

True power factor

Real power over apparent power including harmonics; lower than displacement PF when distortion is present

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• How do you size a dry-type transformer?
• What is an EPMS in a data center?
• What is an isolated ground receptacle?
• What is power factor?
• How long does data center electrical commissioning take?