SiC MOSFET Active Harmonic Filter for THDi Correction

A 1,000 kVA transformer feeding six-pulse VFDs, spot welders and single-phase loads is a textbook harmonic source. Field measurements on such busbars typically show current THD of 25–40 % at the PCC, with the 5th and 7th orders dominant. Third-order currents add up in the neutral conductor and overheat it.

Fixed capacitor banks rarely survive this environment. When the natural resonance between the bank and transformer leakage inductance falls near the 5th harmonic, the capacitors carry 1.5–2 times rated current and fail within months. A SiC MOSFET active harmonic filter changes the outcome because it does not rely on passive tuning: it measures the load current and injects an opposite-phase harmonic current in real time, cancelling distortion at the source.

Measure the Harmonics Before You Touch the Switchgear

Start with a power-quality survey, not a nameplate calculation. Use a Class A meter at the PCC for at least one full production week, and record the spectrum at the transformer secondary and each major feeder.

Read the spectrum the way you read a vibration report. Six-pulse VFDs produce 6n±1 orders — 5th, 7th, 11th, 13th — with magnitude falling roughly as 1/h. Single-phase welders, UPS rectifiers and LED lighting produce 3rd-order currents that do not cancel in the neutral; electrode and spot welders add flicker and interharmonics that no passive trap can track.

Compare the measured TDD against IEEE 519-2014 Table 2, which limits current distortion to 5–15 % depending on the short-circuit ratio at the PCC. If you are above the limit — or if capacitor banks, transformers or neutrals are already failing — you need dynamic compensation. If reactive power is the only problem, an SVG is the right tool; if harmonic current is the problem, an active harmonic filter is the right tool. Many sites need both.

Passive Traps Correct Yesterday's Load Profile

A tuned passive bank is a fixed-impedance device. It compensates the one or two orders it was tuned for, at the load level that existed when it was ordered.

Industrial load profiles drift. Lines add VFDs, welders age, and the spectrum shifts; the trap detunes, its components age under thermal stress, and the bank starts absorbing harmonic current instead of blocking it. That is exactly the mechanism that destroys capacitors. Passive banks also cannot compensate zero-sequence currents, because they have no neutral-current path.

An active harmonic filter measures the load continuously, extracts the harmonic components digitally, and generates compensating current through SPWM-modulated switching. It tracks changing loads cycle by cycle and can be configured to compensate selected orders from 2nd to 50th. A passive trap is a filter tuned once; an active filter is a controller that follows the load.

Why a SiC MOSFET Active Harmonic Filter Outperforms IGBT Designs

The switching device sets the ceiling for bandwidth, efficiency and power density. This is where a SiC MOSFET active harmonic filter separates from conventional IGBT designs.

An IGBT carries a tail current at turn-off, so its switching losses climb steeply with frequency. IGBT-based filters typically switch at 10–16 kHz, which caps the highest compensable harmonic order and forces larger output reactors to smooth the ripple. A SiC MOSFET has no minority-carrier tail; its switching losses stay low well above 20 kHz, so the filter compensates high orders with a smaller output inductor and lower current ripple.

The thermal budget matters as much as the electrical one. SiC devices run at higher junction temperatures with lower losses, which shrinks the derating curve. A well-designed SiC MOSFET active harmonic filter delivers full rated current at 45 °C ambient; many IGBT designs derate above 40 °C. In a hot Indian or Gulf electrical room, that difference decides whether the filter delivers its nameplate rating in summer.

Table A — Si IGBT vs SiC MOSFET active harmonic filter (typical figures; confirm per datasheet)

Three-Level NPC Output: Halve the dv/dt Before It Reaches the Bus

The output stage topology determines how cleanly the compensating current is injected. A 3-level NPC inverter divides the DC-link voltage, so each switching step is Vdc/2 instead of Vdc.

The results are measurable. Output dv/dt is roughly halved, which reduces conducted and radiated EMI into adjacent control wiring. The PWM waveform contains less harmonic content at the same switching frequency, so the output filter is smaller and residual high-frequency current on the bus is lower. Each device also sees half the DC-link voltage, which improves reliability under the switching transients that kill power modules. SPWM with phase-shifted carriers spreads the switching spectrum further. None of this shows in a brochure photo; it shows in the commissioning waveform and in the EMI levels measured around the cabinet.

Sizing from Field Data, Not from the Transformer Plate

Size the filter from the measured spectrum. The harmonic current to cancel is the RMS sum of the individual orders:

I_h = sqrt( I5² + I7² + I11² + ... )

A quick field rule works when you only have a THD reading: harmonic current ≈ load current × THDi. A 400 A feeder at 35 % THDi carries roughly 140 A of harmonic current, so one 150 A module covers it with margin. Add 20–30 % headroom for load growth — modules parallel easily, but upsizing a cabinet later costs more than ordering it right the first time.

Table B — Sizing matrix from typical field measurements (verify each site with a survey)

Do not ignore the zero-sequence channel. In welding and single-phase-heavy plants the neutral carries the summed 3rd-order current while the phases look acceptable; a 4-wire AHF compensates it, a 3-wire unit cannot. Confirm that the CTs sample the load without including the filter's own output, or the control loop can oscillate — CT polarity errors are the most common cause of a filter that "makes harmonics worse".

Installation and Commissioning Realities

An active filter is a parallel device: it connects to the bus through a breaker and never sits in the load current path. Retrofitting does not require redesigning the distribution, but it does require a short planned outage for CT installation and cabling — normally a few hours, not a plant shutdown.

Keep the filter close to the distorting load. Long cable runs between the filter output and the PCC add impedance and weaken high-order compensation. In dusty or hot rooms, respect the clearance and airflow figures in the datasheet; a filter in a 50 °C enclosure with blocked ventilation derates exactly like the IGBT design you tried to avoid. Commissioning is a measurement exercise: verify phase sequence and CT polarity, close the breaker, and watch the THD meter step down within one to two mains cycles of a load step.

The Payback Appears on the Utility Invoice

The economics rest on three measurable items. First, the reactive-energy and distortion penalties that many utilities bill at the PCC. Second, the transformer and cable capacity released when distortion current is removed — a transformer at 34 % THD carries roughly 10 % more RMS current for the same real power, and that margin returns as usable capacity. Third, the avoided failures: capacitor banks, breaker trips and production stops cost more than the filter in a single incident.

Run the calculation on your own meter data. If the annual penalty plus one capacitor bank replacement exceeds 20–25 % of the installed filter cost, payback is inside three years — before counting the transformer capacity gain. That number comes from their invoices, not from our claims.

FAQ

1. How do I size an active harmonic filter?

Measure THD and the harmonic spectrum at the PCC over a full production cycle with a Class A meter. Rate the filter for the RMS sum of the harmonic currents to cancel, plus 20–30 % headroom. If only THD is known, use load current × THDi as a first estimate, then confirm with a spectrum; modular units parallel easily if the load grows.

2. Does installation require a plant shutdown?

The filter connects in parallel and never interrupts the load, but CT installation and cabling need a short planned outage — normally a few hours on a weekend or maintenance window. Energizing and commissioning the filter does not disturb production.

3. What maintenance does an active filter need?

Less than a capacitor bank, but not zero. Cooling fans and air filters are the wear items — inspect annually in normal rooms, quarterly in dusty sites. DC-link film capacitors have long service lives; check busbar torque and terminal temperatures during the annual inspection. The filter reports its own thermal and DC-link data, so most faults are caught before they stop production.

4. Can it run alongside my capacitor bank or generator?

Yes, with one engineering check. If the capacitor bank stays in service for reactive correction, verify its resonance point does not coincide with the orders the filter compensates; if it does, add detuning reactors or let an SVG take over the reactive duty. With a generator, place the CTs so the generator current is inside the measurement loop, or the generator keeps supplying harmonics the filter cannot see. Both cases are resolved during the site survey, before any equipment is ordered.

Conclusion: Specify from Physics, Not from Marketing

A SiC MOSFET active harmonic filter earns its place in a VFD- or welder-heavy plant because its physics — low switching loss at high frequency, wide junction-temperature margin — and its 3-level NPC output deliver wider bandwidth, lower dv/dt and full rated output at 45 °C ambient. Size it from measured harmonic data, keep the CT installation correct, and the result is a TDD below the IEEE 519 limit, a cooler transformer, and a utility invoice without distortion penalties. Those are engineering outcomes, and they are the only arguments that survive a technical evaluation.