How to Measure Harmonic Distortion: Practical Steps, Sensors, and Limits

Most harmonic problems are misdiagnosed long before they are corrected. Picture a plant where one transformer runs hotter than its identical twin, a feeder shows 120 A in the neutral while each phase carries only 60 A, and a capacitor bank has failed twice in 18 months. The maintenance team measures 4.2 percent total harmonic distortion, sees that it sits under the 8 percent limit used for low-voltage buses, and files the case as normal.

The conclusion first: harmonic distortion is not a single number you read off a display. Voltage distortion is measured at a busbar with a Class A analyzer over defined windows, and harmonic current is measured per feeder with a sensor that stays accurate at the harmonic orders you care about. Miss either half and the data will send you toward the wrong fix.

Two Separate Measurements Hide Behind One Keyword

Voltage THD is the rms sum of harmonic components from the second order up to a defined cut-off, divided by the fundamental component and expressed as a percentage. The cut-off is part of the definition, so a figure quoted to the 25th order cannot be compared directly with one quoted to the 50th. Current distortion is normally reported as a spectrum rather than one figure, because acceptance criteria are set order by order and judged against the demand load current, which is TDD rather than THD.

Two practical consequences follow. Voltage distortion describes the supply; harmonic current describes the load that is polluting it. And on a four-wire system, triplen orders such as the 3rd, 9th and 15th add arithmetically in the neutral instead of cancelling, so a feeder can pass every phase-current check and still overheat its neutral conductor.

A 4 percent voltage THD paired with a 60 percent current TDD on a single feeder is a resonance signature, not a compliance pass. Always report the two numbers together.

Class, Window, and Sampling Rate Decide Accuracy

Standards define accuracy for distorted waveforms, not for clean sine waves. IEC 61000-4-30 Class A is the level to specify when results may be disputed or used for contractual compliance; Class S is a lower-cost option for statistical surveys. IEC 61000-4-7 defines how harmonics are grouped, and the two documents are meant to work together.

Time synchronization belongs on the same list. Without a common clock, you cannot prove that a distortion peak at 14:20 came from a drive starting up rather than from a capacitor bank switching two rooms away. A multifunction meter that prints one THD figure per second is a useful trend indicator, but grouping, windowing and filtering are what make a reading defensible.

APView400 Class A Certificated Power Quality AnalyzerThis Class A analyzer records disturbance waveforms, direction, and event sequences, making it suitable for detailed harmonic and power-quality investigations.View Product →

An instrument built for this job, such as a Class A power quality analyzer, records the spectrum, the window and the waveform snapshot together, which is exactly what a harmonic investigation needs.

Current Sensors Are Usually the Weak Link

General-purpose current transformers are designed to reproduce the fundamental accurately and cheaply. Ratio error stays small, but phase error grows with frequency, and by the 25th harmonic a standard CT can be several degrees out. That is tolerable for magnitude readings and dangerous for harmonic power direction: when the true harmonic power is close to zero, a few degrees of channel phase error is enough to flip the sign, and you end up blaming the wrong side of the transformer.

Watch the secondary burden and lead length as well. A long run into a high-burden input pushes the phase response further away from the datasheet curve, and an undersized split-core CT with an air gap behaves differently from a solid-core unit. Sensors based on an air-core coil avoid saturation and hysteresis and hold a much flatter response across the harmonic range, at the cost of a matching integrator and a higher price per channel.

BR Series Rogowski Coil Flexible Current TransformerFlexible split-core AC sensor for wide-range current measurement, with no CT saturation and high bandwidth for power-quality, harmonic, and transient capture.View Product →

Whatever you select, verify three things before the survey starts: the specified bandwidth of the sensor, the actual burden of the measuring circuit, and the position of the conductor inside the measuring window. A conductor pushed against one side of the window is a common source of unexplained phase shift.

A Repeatable Measurement Procedure

The steps below keep a harmonic survey comparable between visits, which matters more than any single reading.

1. Write down the question and the limit the result will be compared against before any instrument is connected.
2. Mark the measurement points on the single-line diagram: the coupling busbar for voltage, each suspect feeder for current, and every neutral on four-wire circuits.
3. Record CT ratio, polarity and sensor orientation, and note the exact ratio used in the report so future surveys can be compared.
4. Cross-check phase angle with a known resistive load, such as a heater bank, before starting a long log.
5. Set the window and aggregation described above, then log for at least seven days to cover the production cycle and the weekend load pattern.
6. Trigger waveform captures on the highest short-time distortion values rather than on the highest rms current, because the events that stress equipment are often brief.
7. Compare the 95th percentile against the limit, not the average, and check whether peaks coincide with a specific drive, inverter or capacitor bank.

Settings to record alongside the data

• Harmonic grouping method and fundamental frequency
• Aggregation interval and total observation time
• Cut-off order or frequency of the spectrum
• CT ratio, class, and sensor bandwidth
• Time source used for synchronization

Six Mistakes That Produce the Wrong Number

• Reporting one THD value for an entire site, which hides the feeder that is actually causing the problem.
• Trusting a clamp meter that shows an rms value with no phase or order information.
• Using a current transformer far beyond the frequency range for which it was characterised.
• Measuring for a few minutes at the lightest load of the day, when voltage distortion is typically at its mildest.
• Ignoring the neutral on four-wire systems, where triplen harmonics can exceed the phase current.
• Leaving the analyzer clock unsynchronized, which makes every correlation with plant events guesswork.

From a One-Week Survey to Continuous Monitoring

Harmonic distortion is not a fixed property of a building. It changes with the production schedule, the output of the solar inverter, and how many EV chargers happen to be drawing at the same moment. That is why a one-week survey is usually followed by permanent monitoring on the two or three feeders that carried the highest distortion.

Permanent monitoring needs meters that publish harmonic data over Modbus RTU or TCP, a gateway that can forward it to a platform without losing timestamps, and threshold alarms set on the 95th percentile rather than on instantaneous peaks. Once the data is continuous, the useful comparison is no longer against a limit but against the same feeder last month, which is how you catch a filter degrading or a new drive being commissioned.

Acrel AWT100 Smart Gateway Data Transfer UnitCE-certified wireless gateway with 4G, WiFi, Ethernet, LoRa, and RS485 downlink, connecting up to 25 devices for IoT energy-management data transmission.View Product →

Solutions of this kind are already used in IoT-based energy management systems, where harmonic trends sit next to consumption data and can be reviewed by the same team.

Questions Engineers Ask Before Buying

Harmonic distortion becomes manageable the moment you stop treating it as one number. Measure voltage distortion at the busbar, harmonic current per feeder, and the neutral separately, document the window and the sensor behind every figure, and the eventual fix, whether it is a detuned filter bank, a different drive setting or a transformer replacement, will be justified by data instead of by guesswork.

Class A analysisPer-feeder current sensingNeutral triplensContinuous trending

If you are specifying the hardware for a harmonic survey or building permanent monitoring into a new project, share your single-line diagram and load profile with our engineers. We will help you match the analyzer, sensors and gateway to the points you actually need to watch, so the first report you produce is one you can act on.