Non-linear loads have changed the way transformer capacity must be evaluated. Variable-frequency drives, UPS systems, rectifiers, EV chargers, and switch-mode power supplies draw harmonic currents that increase transformer heating beyond what the RMS current alone suggests — a transformer operating below its nameplate kVA can still exceed its intended thermal limits.
IEEE C57.110 provides the framework for evaluating this effect. Rather than relying on total harmonic distortion, the method combines the individual harmonic spectrum with the transformer’s loss characteristics to estimate the maximum permissible non-sinusoidal load current. This article explains the harmonic loss factor, walks through the derating process, and works a numerical example.
Why Non-Linear Loads Overheat Transformers
Nameplate kVA does not describe how current is distributed by frequency. A drive or charger draws a fundamental component accompanied by substantial 5th, 7th, 11th, and 13th harmonic currents, and the higher-frequency components produce disproportionately high winding and structural losses. IEEE C57.110-2018 provides conservative methods for evaluating two-winding liquid-immersed and dry-type transformers supplying such loads, both for assessing existing installations and for specifying new equipment. Purpose-designed rectifier transformers are a separate class outside its published scope, and the method is an application calculation, not a substitute for the manufacturer’s thermal design analysis.
The evaluation also differs from harmonic compliance at the point of common coupling: IEEE 519 addresses system-level distortion limits, while C57.110 addresses the effect of harmonic currents on transformer losses and loading capability. A facility can comply with its harmonic limits and still need to evaluate transformer heating.
Where the Extra Heat Comes From
Transformer load loss is more than conductor I²R loss. Leakage flux crossing the winding conductors induces eddy currents within the conductor cross-section, and in the conservative C57.110 method the winding eddy-current loss at each harmonic is treated as proportional to the square of that harmonic’s current and the square of its order — a 5th-harmonic ampere has roughly twenty-five times the eddy-loss effect of a fundamental ampere. Other stray losses in leads, clamps, and structural parts follow a milder frequency relationship. The standard therefore separates load loss into I²R loss, rated winding eddy-current loss PEC-R, and rated other stray loss POSL-R, and applies a different harmonic weighting to each. This is also why current THD alone is insufficient: the frequency distribution, not the distortion total, sets the heating.
A standard nameplate rating closes the manufacturer’s thermal budget around substantially sinusoidal current. Harmonic current changes how that budget is consumed, and derating reduces total current until calculated load loss returns to the rated value. The result is not a new nameplate — it is a calculated maximum for a defined spectrum, and it changes if the spectrum, ambient, cooling, or transformer condition changes.
The Harmonic Loss Factor
The harmonic loss factor converts a measured spectrum into a multiplier for winding eddy-current loss:
The summation runs over all measured harmonic orders including the fundamental; Ih is the RMS current at order h, and because the same reference appears in numerator and denominator, any consistent per-unit base works. Each squared ratio is a harmonic’s RMS-squared contribution, and multiplying by h² applies the eddy-loss frequency weighting, so the ratio answers one question: relative to a sinusoidal current of the same total RMS value, how much larger is winding eddy-current loss for this spectrum? For a pure sine wave FHL equals 1, and it rises as higher-order content enters. The square-of-order term makes spectral detail decisive — two loads with identical THD can have different FHL values if one concentrates distortion in the 5th and 7th while the other carries more 11th and 13th. And FHL applies only to the winding eddy-current component: I²R loss still scales with total RMS current squared, and other stray loss is weighted separately.
Estimating Permissible Loading
Step 1: Measure the Harmonic Current Spectrum
Measure current by harmonic order at the transformer terminals under representative loading, capturing all phases and the neutral where applicable. For cyclic loads, identify the spectrum associated with sustained thermal duty rather than a convenient snapshot, and record voltage distortion, balance, ambient temperature, and cooling status as context.
Step 2: Calculate the Harmonic Loss Factor
Normalize each harmonic current to a consistent reference, square it, weight it by h², sum the results, and divide by the unweighted sum of squares.
Step 3: Apply Transformer Loss Data
On an I²R base, let PEC-R be rated winding eddy-current loss and POSL-R rated other stray loss, both per unit. Rated load loss is:
For a spectrum assumed to scale proportionally with load, calculated load loss at per-unit current Ipu is:
FHL-STR is the milder factor the standard applies to other stray loss — it comes from the C57.110 procedure and is not assumed equal to FHL. Where a simplified dry-type evaluation legitimately excludes other stray loss, that is an engineering decision supported by the method and available test data, not a shortcut.
Step 4: Determine Maximum Permissible Current
Multiplying Imax by rated current gives the maximum permissible RMS load current for the evaluated spectrum. The result must still be checked against the applicable temperature-rise method, hot-spot limits, neutral loading, terminal capability, cooling conditions, and manufacturer restrictions — it is one input to an engineering assessment, not project loading authorization.
A Worked Example
Consider a dry-type transformer feeding a drive-dominated load with a measured spectrum, per unit of fundamental, of I5 = 0.20, I7 = 0.14, I11 = 0.09, and I13 = 0.07. The unweighted sum of squares is 1.073 and the h²-weighted sum is 4.77, so FHL = 4.77 ÷ 1.073 ≈ 4.45. For a design with PEC-R = 0.10 per unit, using the standard’s simplified dry-type treatment:
The transformer can carry about 87% of its rated current on this spectrum — a 75 kVA unit becomes, in effect, a 65 kVA unit. The figures are illustrative; a real study uses the full measured spectrum and the manufacturer’s loss data.
K-Factor or C57.110 Derating?
K-factor and FHL use similar h² weighting and can be numerically equal when calculated on the same total-RMS-current basis, but they serve different purposes. K-factor describes, on the specification side, the harmonic severity a K-rated transformer is built to withstand thermally — see Understanding the K-Factor of Transformers and Harmonics. C57.110 evaluation is more granular: it combines the actual spectrum with transformer-specific loss data to estimate permissible current, which makes it the right framework when the question is whether an existing conventional transformer can carry a measured non-linear load. The system-level context is covered in Understanding Harmonics in Power Systems: IEEE 519 Explained.
When to Derate vs. Specify a K-Rated Unit
Derating suits an existing standard transformer with spare capacity, a measured and stable spectrum, and operational demand that fits the reduced capability — or an interim condition while a replacement strategy is developed. The cost is that part of the nameplate becomes unavailable and future growth is constrained. For a new installation with known, sustained harmonic loading, a K-rated transformer carries the heating at rated kVA, and a harmonic mitigating transformer goes further by cancelling selected harmonics through winding configuration. Selection should follow the measured or specified spectrum, neutral requirements, and growth forecast, not a K-factor label alone.
Common Calculation Mistakes
Using THD-I instead of the full spectrum. THD compresses all distortion into one number and discards harmonic order; because FHL weights order, equal THD does not imply equal heating.
Confusing FHL with K-factor. The values can coincide numerically, but one is a loss multiplier in a capability calculation and the other a transformer withstand rating — confirm the normalization convention and purpose before using either.
Applying FHL to every loss component. FHL belongs to winding eddy-current loss; multiplying total load loss by it misstates heating, while ignoring POSL-R understates it.
Using generic loss percentages. The result is sensitive to PEC-R and POSL-R. Where the certified breakdown is unavailable, the standard’s conservative procedure and manufacturer input govern — a percentage copied from another transformer is not design data.
Assuming the spectrum is constant. Drive input current changes with loading, front-end topology, reactors, and source impedance; the study must capture the sustained worst thermal condition.
Stopping at the loss equation. Ambient temperature, ventilation, altitude, hot-spot rise, neutral current, imbalance, terminals, and protective devices can govern permissible loading; the calculation belongs inside a coordinated application review.
Rex Power Magnetics Perspective
For an existing installation, Rex Power Magnetics can review transformer construction and available test information against the measured load profile with the project engineer. For a new installation, the same analysis informs whether a conventional unit with margin, a K-factor-rated transformer, or a harmonic mitigating transformer is the appropriate design response — the objective being a defensible thermal margin over the full operating profile, not just today’s load.
Conclusion
IEEE C57.110 derating turns a harmonic current spectrum into a transformer capability assessment: FHL captures the frequency-weighted effect on winding eddy-current loss, and transformer-specific loss data determines how strongly it bites. A sound study uses representative measurements, preserves the separation between loss components, and evaluates the result within the complete thermal and application conditions — with conservative assumptions and manufacturer review where inputs are missing. That discipline is what makes the calculation a defensible engineering basis for operation or specification.