Why an Air Gap Matters
Almost every energy-storage inductor and every flyback, forward or push-pull transformer uses a deliberate air gap in the magnetic path. The gap is not a manufacturing defect: it is the single parameter that decides how much energy the component can store before the core saturates, how stable the inductance stays over current and temperature, and how much fringing flux leaks into the windings.
Engineers new to magnetics often start from the core datasheet: pick an AL value, multiply by N², and assume the inductance is fixed. Real gapped cores behave quite differently. Tolerance, fringing flux, temperature and DC bias all move the operating point. This guide explains how the gap works, how to calculate it, and where practical designs usually go wrong.
How a Gap Stores Energy
Magnetic cores concentrate flux, but they also concentrate the energy density. A high-permeability ferrite can carry only a small amount of energy per unit volume before B reaches saturation. By inserting a gap, you raise the reluctance of the magnetic path and shift most of the stored energy into the gap volume — a region that can never saturate.
For a gapped core the inductance is:
L = N² / (Rcore + Rgap)
where Rgap = lg / (μ0 · Ae) and Rcore = le / (μ0 · μr · Ae).
Because μr of a ferrite is in the range of 2,000–3,000 while air is 1, the gap dominates the reluctance as soon as it exceeds a few tenths of a millimetre. That is exactly the behaviour we want: the inductance becomes defined by mechanical dimensions rather than by the core material, so it is far more stable with temperature and with DC bias.
Calculating the Gap for a Target Inductance
Working backwards from a required inductance and a peak current, the design sequence is:
- Choose the core size from the energy requirement: E = ½ · L · Ipk², then compare with the core’s area–product capability.
- Compute the number of turns needed at the saturation limit: N = Bsat · lg / (μ0 · Ipk), using Bsat at the maximum operating temperature (ferrite loses roughly 20–30 % of its room-temperature Bsat at 100 °C).
- Solve for the gap that gives the target inductance with that N.
- Iterate: an increased gap needs more turns, which increases copper loss; a smaller gap needs fewer turns but a larger core.
A practical shortcut for ferrite cores with a dominant gap is L ≈ μ0 · N² · Ae / lg. It is accurate to within 5–10 % when the gap is larger than about 0.3 mm on an EE or ETD core.
Design Example: 100 µH, 10 A Peak, ETD49 Ferrite
| Parameter | Value | Note |
|---|---|---|
| Target inductance | 100 µH | at 20 °C, no bias |
| Peak current | 10 A | worst case, hot |
| Bsat at 100 °C | 0.30 T | Mn-Zn power ferrite |
| Effective area Ae | 211 mm² | ETD49 datasheet |
| Gap lg | 1.0 mm | centre-leg, one side |
| Turns N | 62 | from Bsat equation |
| Resulting L | ~102 µH | before fringing correction |
With a 1 mm gap on an ETD49 the fringing factor is typically 1.1–1.2, so the measured inductance will read 12–20 % higher than the simple formula predicts. Design the turns count with the fringing-corrected value, not the ideal one.
Fringing Flux: The Hidden Cost of a Gap
Flux leaving the gap does not travel in straight lines; it bulges outward and intersects the winding window. Three consequences follow:
- Extra copper loss. Fringing flux cuts the winding at right angles to the main flux, adding eddy currents in the conductor. The effect grows with gap length and with the number of layers, and is worst in solid round wire at high frequency.
- Local hot spots. Windings placed directly over the gap see the highest loss density. Moving the winding away from the gap, or splitting the gap into several smaller gaps, is usually more effective than reducing the number of layers.
- Higher measured inductance. The fringing path provides a parallel flux path, which raises L above the ideal calculation. Some designers treat this as free inductance; it also means the saturation current is reached earlier than expected.
Mitigations that work in production: distribute the gap as 2–4 smaller gaps around the core, use a wider bobbin to move the winding off the gap, use Litz or multiple parallel strands, or select a powder core with an intrinsically distributed gap.
Gap Types and When to Use Each
| Gap type | Typical use | Advantages | Watch-outs |
|---|---|---|---|
| Single centre-leg gap (ferrite) | Flyback transformers, PFC chokes | Lowest cost, easy to grind | Strong fringing, gap tolerance ±0.05 mm shifts L by 5–10 % |
| Distributed gap (powder core) | Buck/boost inductors, PFC | Soft saturation, low fringing, stable L | Higher core loss, μ limited to 14–125 |
| Multiple shimmed gaps | High-current inductors, resonant chokes | Lower fringing loss, better thermal spread | More assembly steps, mechanical clamping |
| No gap (toroid, nanocrystalline) | Common mode chokes, CTs, low-loss transformers | Highest permeability, lowest magnetising current | Cannot store DC energy, saturates easily |
Practical Design Rules
- Specify the gap as a mechanical dimension with tolerance, not just as an inductance. ±0.05 mm on a 0.5 mm gap is a 10 % inductance spread; on a 2 mm gap it is only 2.5 %.
- Evaluate Bsat at the hot-spot temperature, and derate by 15–20 % for transient current. Saturation is a thermal runaway: more current → lower inductance → faster current rise.
- Keep the winding off the gap or add a copper shield turn (a shorted turn is not acceptable — use an open copper foil tied to a quiet node if EMI demands it).
- Check the gap under DC bias, not only with an LCR meter at zero current. A gapped ferrite can lose 20–40 % of its small-signal inductance at rated current.
- Consider the acoustic effect. Gaps in magnetic paths generate magnetostrictive noise at twice the switching frequency; a potted or clamped core is measurably quieter.
Magnetic Cores for Gapped and Distributed-Gap Designs
Shaanxi SiRui Technology supplies both routes: gapped Mn-Zn power ferrite cores ground to your inductance target, and distributed-gap powder cores (Sendust, MPP, High Flux, Kool Mu) with permeability from 14 to 125 for soft-saturation inductor designs. Every lot is tested for AL value, saturation flux and core loss, and gapped cores are supplied with the measured gap and inductance on the label so your incoming inspection can be a single LCR check.
FAQ
Does a larger gap always mean a lower inductance? Yes, for the same turns count. To keep L constant you must add turns, which increases copper loss — this is the fundamental trade-off in gapped-core design.
Can I make the gap by stacking two ungapped cores? Only if the mating faces are deliberately finished to a controlled flatness. Ordinary core faces have surface roughness that behaves as an uncontrolled 20–50 µm gap, giving unpredictable inductance.
Why does my gapped inductor buzz? Magnetostriction and gap fringing force at the switching frequency create mechanical vibration. Potting, varnishing or clamping the core halves reduces it; changing the gap will not.
Related reading: Inductor saturation and DC bias · Steinmetz core loss calculation · Magnetic core selection guide
About Shaanxi SiRui Technology Co., Ltd.
Magnetic cores, power inductors, current transformers and high-frequency transformers for EV charging, solar inverters, industrial power and smart metering. Custom designs and samples available.
Siri Yuan, Sales Manager · Email: sales@sirui-tech.com · Web: www.sirui-tech.com
No. 2555, Chanba Avenue, Chanba Ecological Zone, Xi’an, Shaanxi Province