Passive filters are cheaper and simpler and work beautifully — until the load stops being what you assumed it was.
Once a harmonic limit has been established at the point of common coupling and line reactors alone will not meet it, the choice narrows to three practical families: tuned passive filters, multi-pulse drive front ends, and active filters. At full rated load all three can hit the same compliance number. They diverge sharply everywhere else.
A passive filter is a series inductor with a shunt trap tuned near a target harmonic — most commonly the 5th, since it is the dominant harmonic of a six-pulse rectifier. The trap presents low impedance at the tuned frequency, diverting that harmonic current away from the supply.
Strengths. No control electronics, nothing to fail in software, no cooling fans in many designs, low lifetime cost, well-understood behaviour, and typically the lowest first cost per ampere. A well-applied passive filter regularly delivers below 5% TDD at rated load.
The part-load problem. This is the issue that catches people. The filter's shunt capacitance is essentially fixed, while the harmonic current it is correcting falls with load. At light load the capacitive reactive current dominates and the installation presents a leading power factor. Consequences range from a utility power factor penalty, through voltage rise at the bus, to genuine instability.
Tuning and detuning. A passive filter is tuned to the system it was designed for. Changing the supply transformer, adding power factor correction capacitors, or substantially changing the connected load can detune it. In the worst case the filter and the system source form a resonant circuit that amplifies a harmonic rather than trapping it.
An 18-pulse drive uses a phase-shifting autotransformer to feed three six-pulse rectifier sections at 0°, +20° and −20°. The phase shift causes the 5th, 7th, 11th and 13th harmonics to cancel between sections, leaving the 17th and 19th as the lowest significant orders.
Strengths. Excellent distortion performance — often 5% TDD or better — achieved by topology rather than by added correction. No capacitors, so no leading power factor and no generator interaction. Performance holds reasonably across the load range.
Costs. Substantially higher first cost, larger footprint and weight, and additional losses in the phase-shifting transformer. Performance also degrades with input voltage unbalance: a few percent of unbalance can noticeably erode the cancellation the topology depends on.
An active filter measures load current, computes the harmonic content, and injects an equal and opposite current through an IGBT inverter stage.
Strengths. Performance holds across the load range — this is the decisive advantage. It corrects whatever spectrum is actually present rather than a spectrum assumed at design time, so it handles mixed and changing loads. Many units also provide reactive compensation and load balancing, and capacity can be expanded by adding units.
Costs. Highest first cost per ampere. It is an electronic product with a finite service life, requires cooling and airflow, and has its own losses. Rating is in harmonic amperes, not kVA, and undersizing is a common error: an active filter asked to supply more corrective current than it is rated for simply clips and leaves you non-compliant.
| Tuned passive | 18-pulse | Active | |
|---|---|---|---|
| TDD at full load | <5% | <5% | <5% |
| Performance at 25% load | Degrades; leading PF | Good | Good |
| Generator compatible | Needs design attention | Yes | Yes |
| Sensitive to system changes | Yes — can detune | Voltage unbalance | No |
| Relative first cost | Low | High | Highest |
| Footprint | Moderate | Large | Moderate |
| Electronics to maintain | None | None | Yes |
| Expandable later | No | No | Yes |
The filter shunt capacitance is essentially fixed while the harmonic current it corrects falls with load. At light load the capacitive reactive current dominates, so the installation presents leading power factor. That can bring a utility penalty, bus voltage rise, or instability on a generator.
Only with design attention. Generator voltage regulators are generally designed for lagging load and can become unstable on a leading one. Options include contactor-switched filter stages that disconnect capacitance at light load, or moving to an 18-pulse or active solution.
It achieves low distortion by topology rather than added correction, so there are no capacitors, no leading power factor and no generator interaction, and performance holds across the load range. The trade-offs are substantially higher first cost, larger footprint, and sensitivity to input voltage unbalance.
In harmonic amperes of corrective current, not in kVA. Undersizing is a common error — an active filter asked to supply more corrective current than its rating simply clips and leaves the installation non-compliant. Rating should come from measured harmonic current, not from connected load.
Yes, and the interaction can be severe. Capacitors and system inductance can form a resonant circuit that amplifies a harmonic rather than reducing it. Any installation with existing power factor correction needs a resonance study before a passive filter is added.
Data center distribution and grid-scale battery storage both present harmonic and thermal problems that traditional plant experience handles badly. The differences are specific and they change the component selection.
Most harmonic specifications are written against the wrong measurement point. Understanding where IEEE 519 limits apply, and how the Isc/IL ratio sets them, is what separates a $4,000 line reactor from a $40,000 filter.
NEMA 1, 3R, 4X and 12 are not a simple ladder of increasing protection. Each addresses different hazards, and the wrong choice either wastes money or fails in service — often through heat rather than ingress.