What Non-Linear Loads Do to a System

Why this matters

The same rack of electronic equipment can be installed in two buildings and cause a problem in exactly one of them. That is not luck and it is not a bad batch of gear. Every effect a nonlinear load has on a system is harmonic current flowing through upstream impedance, so the harm is a property of the PAIR - the load and the source it is bolted to - and never of the load alone. A shop that understands this stops asking "how much nonlinear load is too much" and starts asking "too much for what source", which is the only version of the question that has an answer.

The practical payoff is that you can predict which buildings will have trouble before anyone buys a meter, and you can tell a customer why the identical installation at their other site is fine.

Before anyone opens a capacitor enclosure

Power-factor correction capacitors appear throughout this card, and they are a stored-energy hazard that outlives the disconnect. A capacitor bank holds a charge after the supply is removed, and the discharge resistors that bleed it take time to do it.

De-energize, lock and tag under 29 CFR 1910.333(b)(2), which owns electrical work because 29 CFR 1910.147 excludes exposure to electrical hazards from work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C), with 29 CFR 1926.417 as the construction counterpart. Then WAIT the discharge time the capacitor manufacturer states on the equipment or in its literature before opening the enclosure, because that number is theirs and no rule of thumb substitutes for it. Prove each capacitor terminal dead live-dead-live per NFPA 70E-2021 120.5, in the edition your employer's electrical safety program or your authority having jurisdiction has adopted, and then apply a shorting and grounding device rated for the equipment before your hands go in - a discharge resistor that has failed open leaves a fully charged bank behind a disconnect that reads open. Every reading in this card that must be taken with the system running falls under 29 CFR 1910.333(a)(1), which requires de-energizing unless the employer can demonstrate that it introduces additional or increased hazards or is infeasible, with boundaries and PPE selected under NFPA 70E-2021 130.5 and 130.7 as adopted.

The single relationship everything hangs off

Voltage distortion at a bus is harmonic current multiplied by the source impedance at that harmonic's frequency. That is the mechanism, and two consequences follow from it directly.

Source impedance is mostly inductive, so it rises with frequency. An inductance presents impedance proportional to frequency, so the fifth harmonic sees roughly five times the source impedance the fundamental sees, and the seventh roughly seven times. A harmonic current that is a modest fraction of the fundamental therefore produces a voltage distortion out of proportion to its size.

A stiff source hides the same current that a weak source displays. "Stiff" here means low source impedance, which is the same thing as high available short-circuit capacity. For a transformer, available short-circuit capacity at its secondary is approximately the transformer's kVA divided by its per-unit impedance. That approximation treats everything upstream of the transformer as an infinite bus with no impedance of its own, which is never true, so the real available capacity is somewhat lower and the utility owns the authoritative number. Use it to reason about ratios, not to select gear.

The effects, each traced back to that relationship

Transformer heating. The winding eddy-current component of a transformer's losses rises roughly with the square of frequency times the square of the harmonic current, so a fifth-harmonic current that is a fifth of the fundamental contributes eddy loss comparable to the fundamental's own. That is the winding eddy component specifically, not the total loss, and it is why a transformer carrying harmonic-rich load can run at its thermal limit well below its nameplate kVA. Transformers built to tolerate this carry a K-factor rating, which describes harmonic-carrying capability and binds through the product listing rather than on its own.

Circulating triplens. On a delta-wye transformer, the zero-sequence third-harmonic current arriving at the wye neutral circulates in the delta winding instead of passing upstream. That protects the utility side and heats the winding with no matching load on the secondary, so the transformer runs hot while the metered load looks unremarkable. The neutral card owns why the triplens arrive there in the first place.

Voltage flat-topping. Rectifier loads draw their pulse of current at the peak of the voltage wave, which is precisely where the resulting drop is subtracted, so the top of the voltage sinusoid gets clipped while the region near the zero crossings is barely affected. The self-inflicted part is the part worth knowing: the equipment doing the flat-topping charges its own DC bus to that reduced peak, so a room full of drives and switch-mode supplies erodes its own ride-through margin and then faults on the next external sag.

Motor heating. The fifth harmonic is negative-sequence, so it produces a field rotating against the direction of the motor, and the rotor sees it at a high slip frequency. That is loss in the rotor with no useful torque, on a machine whose current reading may look ordinary. The harmonic card owns the sequence pattern that makes this predictable from the harmonic number alone.

Resonance with power-factor capacitors. This is the one that turns an unremarkable installation into equipment failures. Source inductance and a capacitor bank form a parallel resonant circuit, and the harmonic order at which it resonates is approximately the square root of the available short-circuit capacity divided by the capacitor bank's kvar. That estimate assumes a single lumped source inductance, a single capacitor bank and no damping, so it predicts the ORDER at which trouble sits, not how bad it will be. If that order lands on a harmonic the loads actually produce, the parallel combination presents a high impedance at that frequency and a modest harmonic current generates a large voltage at it.

One gate, two buildings

The same customer installs the same 200 kVA block of six-pulse drive and electronic load at two of their sites.

Site one. A 1000 kVA transformer at 5.75 percent impedance, no power-factor capacitors. Available short-circuit capacity at the secondary is approximately 1000 divided by 0.0575, or 17,390 kVA. The nonlinear block is 200 kVA, so it is 20 percent of transformer capacity, and the ratio of short-circuit capacity to distorting load is 17,390 divided by 200, or about 87 to 1.

That ratio is what a stiff source looks like. The same harmonic currents flow, because the loads are identical and current distortion belongs to the load, but they cross a small impedance, so the voltage distortion they create at the bus stays low. The transformer carries harmonic current and runs warmer than a linear load of the same kVA would make it, which is a loading question to check against its rating, and nothing else on the site changes. There is no complaint and no work order.

Site two. A 300 kVA transformer at 5.0 percent impedance, with a 100 kvar power-factor capacitor bank that has been in service for years. Available short-circuit capacity is approximately 300 divided by 0.05, or 6,000 kVA. The same 200 kVA nonlinear block is now 67 percent of transformer capacity, and the ratio of short-circuit capacity to distorting load is 6,000 divided by 200, or 30 to 1, less than half the stiffness of site one before anything else is considered.

Then the capacitors. The resonant order is the square root of 6,000 divided by 100, which is the square root of 60, or 7.75. That sits close to the seventh harmonic, and a six-pulse rectifier front end is a reliable source of seventh-harmonic current. So the site now has a circuit that presents high impedance at a frequency its own loads generate.

What that looks like on the ground: capacitor fuses clearing for no apparent reason, cans running hot or swelling, bus voltage distortion far above what the load fraction alone would suggest, and drives faulting in ways that do not track occupancy. None of it points at the drives, and all of it started when the drives were commissioned.

Read the two against each other. Identical load, identical current distortion at the load terminals, opposite outcomes. Site one had roughly three times the stiffness and no resonant circuit; site two had a weaker source and a capacitance that put its resonance on top of a harmonic the load produces. Nothing about the equipment specification distinguishes them, which is exactly why the specification is the wrong place to look.

What flips the answer

• A line reactor or DC choke on the drives. These raise the load's own series impedance, which reduces the harmonic current it draws in the first place. It is the only item on this list that acts on the current rather than on the system, which makes it the one that helps regardless of what the source turns out to be.
• Detuning the capacitor bank. A reactor in series with the capacitors moves the resonant frequency below the lowest significant harmonic, which is a design change and an engineering decision, not a field adjustment. Do not attempt to chase a resonance by switching stages in and out: every switching combination has its own resonant order, and you can move it from one harmonic onto another.
• Separating the nonlinear load onto its own transformer. This does not reduce the harmonic current, it changes which equipment shares a bus with it, which is often the practical answer in a building with sensitive loads.

How to verify you got this right

Establish stiffness before you measure distortion. Get the transformer kVA and per-unit impedance off the nameplate, and get the available fault current from the utility or the study, because that number is theirs. A site with a high ratio of short-circuit capacity to distorting load is not going to reward a harmonic investigation, and knowing that in advance is worth more than the measurement.

Inventory the capacitance before you blame anything. Power-factor capacitors, and the capacitance inside drives and supplies, are what turn a distortion into a resonance. Compute the approximate resonant order from the relationship above and compare it against the harmonic orders the equipment on site actually produces. A resonance that lands between harmonics is a non-event; one that lands on the fifth or seventh with six-pulse rectifiers present is a finding.

Measure voltage distortion at the bus and current distortion at the branch, and label each. The same word describes both and they are not interchangeable. The bus figure is the one that says whether the SYSTEM has a problem, and it is the one to trend rather than sample once.

Check the transformer's temperature against its loading, not against its kVA. A transformer serving harmonic-rich load can be at its thermal limit while the metered kVA says it has headroom, and the temperature is the measurement telling the truth.

References

• IEEE 519, harmonic current and voltage distortion limits and the point of common coupling concept, as it binds through a utility service agreement or project specification, in the edition that agreement names
• 29 CFR 1910.333(a)(1) and 1910.333(b)(2), with the 1910.147(a)(1)(ii)(C) exclusion; 29 CFR 1926.417 for construction
• NFPA 70E-2021, 120.5, 130.5 and 130.7, as adopted by your employer's electrical safety program or your authority having jurisdiction
• Transformer and capacitor manufacturer literature for K-factor rating, per-unit impedance and stored-charge discharge time
• See related: What a Harmonic Actually Is; Why a Neutral Can Carry More Than You Expect; How to Tell Whether You Have a Power Quality Problem