Insights from power electronics development practice

Anyone developing a grid-connected inverter will eventually face a topic rarely mentioned in marketing materials: the LCL filter. More specifically – the inductors inside it. They are inconspicuous, passive components, and they determine whether a device passes VDE-AR-N 4105 or not.

Let me explain why.

What the LCL Filter Actually Has to Do

The LCL filter sits between the inverter output and the grid. Its job: attenuate the switching-frequency harmonics of the PWM signal to a level where the injected current meets the limits of VDE-AR-N 4105 and EN 61000-3-12.

A simple L filter could theoretically do the same – but only with very large inductance values, which are expensive, heavy, and thermally problematic. The LCL filter, with its three-stage topology (L1 – C – L2), achieves significantly higher attenuation with much smaller components. The trade-off: it is capable of resonance. And that turns component selection into an engineering task, not a procurement task.

The Problem Starts with the Core

Inductors for power filters are not off-the-shelf commodity parts. The core material is decisive – and this is where a trade-off begins that cannot be engineered away.

Ferrite materials offer high permeability and low core losses at lower frequencies. At switching frequencies in the 16–32 kHz range typical of modern SiC- or IGBT-based topologies, core losses and saturation behavior begin to dominate. Ferrite saturates hard – meaning: once the peak current exceeds a threshold, inductance collapses abruptly. The filter loses its effectiveness exactly when it is needed most: during transient load changes.

Kool Mµ and MPP cores (metal powder cores) exhibit soft saturation behavior. Inductance rolls off gradually under load, not suddenly. This gives the control loop time to respond. The downside: higher core losses under continuous operation, which must be managed thermally.

Amorphous and nanocrystalline materials offer excellent loss profiles but are more demanding to manufacture and significantly more expensive.

The choice of core material is therefore not a question of the permeability figure in the datasheet – it is a system-level decision that simultaneously affects thermal management, control strategy, and cost structure.

Tolerances: What the Datasheet Doesn’t Show

Another topic that is underestimated in practice: inductance tolerances.

Typical datasheet figures: ±8% for metal powder cores (e.g. Kool Mµ), ±10–20% for iron powder cores. That sounds manageable. In the LCL filter, however, it means the resonant frequency

f_res = 1 / (2π · √(L1·L2·C / (L1+L2)))

shifts significantly across production spread. With a poorly chosen design, the resonant frequency can, in an unfavorable case, coincide with harmonics from the grid or the PWM frequency – with predictable consequences for the EMC measurement.

An LCL filter that passed VDE-AR-N 4105 in the lab with the prototype inductor can fail with production parts of the same type from a different manufacturing batch. I have experienced this firsthand.

The solution is not recalibration, but designing with a tolerance band: the filter must be dimensioned for the worst-case parameter combination, not the nominal case.

Thermal Behavior as a Hidden Compliance Factor

VDE-AR-N 4105 is measured at room temperature. In practice, inverters operate at ambient temperatures of 40–70 °C – in poorly ventilated enclosures, even higher.

Ferromagnetic core materials change their magnetic properties with temperature. In Kool Mµ, permeability decreases as temperature rises – inductance drops, the filter corner frequency shifts. What was compliant at 25 °C can fall outside the limit curve at 70 °C.

A production-ready design must therefore certify not just the nominal case, but the thermal worst case. This means: inductance measurement across temperature, not just at room temperature. And a filter design that still has sufficient attenuation margin even with degraded inductance.

What This Means for Certification

VDE-AR-N 4105 is not a formality. Grid operators in Germany require it. Wholesalers and distributors require it as a prerequisite for product listing. Installers are liable if a non-compliant device is operated on the grid.

Treating the LCL filter as a “should be fine” component group risks expensive redesigns shortly before market approval. I have seen projects lose three to six months because of this.

The opposite is equally true: engineers who treat the filter as a system component from the start – with consistent tolerance, thermal, and resonance analysis – can approach certification testing with a high probability of first-pass success. That saves time, cost, and frustration.

Conclusion

Grid compliance is not created in the certification lab. It is determined on the schematic – specifically, when every passive component in the signal path is selected.

Inductors in the LCL filter are not a commodity. They are critical system components whose characteristics across temperature, load, and manufacturing spread must be understood and accounted for in the design.

Engineers who do this early avoid unpleasant surprises later.

Werner Böhme is a Dipl.-Ing. in Electrical Engineering and Managing Director of awb-it GmbH. He and his team develop the ampareq Gen3 – a three-phase hybrid battery storage inverter with AI-based energy management. The company holds BSFZ certification as a recognized research organization.