Power Inductor Selection: Saturation Current, DCR and Core Loss

The inductor is usually the largest and most expensive component in a switching regulator. Its datasheet quotes two different current ratings, two unrelated loss mechanisms, and an inductance value that only holds at zero current. Pick from the first line of a distributor table and the converter either runs hot or stops regulating under load.

The two current ratings mean different things

Every power inductor datasheet quotes both, and they are limited by unrelated physics. You have to satisfy both.

Rated current, sometimes called heating current or Irms, is thermal. It is the DC current that raises the part's temperature by a fixed amount, typically 40°C, in still air. DCR sets it, along with how well the package sheds heat.

Saturation current, Isat, is magnetic. It is the current at which inductance has fallen by a stated percentage, usually 20% or 30%. Past saturation the core stops storing energy proportionally, ripple current climbs, and the current waveform develops the upward hook that precedes a blown high-side FET.

The catch is that manufacturers quote Isat against different drop criteria. A 20% figure from one vendor is not comparable to a 30% figure from another, so read the drop percentage in the footnote alongside the number in the table.

Sizing for a buck converter

Start from the ripple current target. For a buck, 30% of maximum output current is the conventional starting point:

L = (Vin_max - Vout) x Vout / (Vin_max x fsw x Iripple)

Then check the part against peak current, not average:

Ipeak = Iout_max + Iripple / 2

Isat must exceed Ipeak with margin, and it must do so at the highest ambient the design sees. Ferrite saturates earlier when hot, often 10-15% earlier at 85°C than at 25°C, and datasheet curves are usually plotted at 25°C.

A worked example: a 12 V to 3.3 V buck at 500 kHz delivering 3 A, targeting 30% ripple (0.9 A), needs about 5.3 µH. Choose the standard 4.7 µH value and ripple rises to roughly 1.0 A, so Ipeak is 3.5 A. Specify Isat at 4.5 A or better and Irms at 3.2 A or better.

Where the losses go

Conduction loss is Irms² × DCR, and it dominates at high load. Halving DCR halves this loss, which is why low-DCR parts earn their price premium on high-current rails.

Core loss scales with switching frequency and with the square of flux swing, so it dominates at light load and high frequency. It is invisible in a DCR comparison. Two parts with identical DCR can differ by a factor of three in core loss at 2 MHz, so above roughly 1 MHz a selection made on DCR alone is incomplete.

Ferrite cores have low core loss and a hard saturation knee. Composite and moulded-alloy cores saturate softly, with inductance rolling off gradually rather than collapsing, at the cost of higher core loss. For a converter that has to survive brief overcurrent, that soft knee is worth paying for.

Shielded, semi-shielded and unshielded

An unshielded inductor sitting next to a feedback trace couples switching noise straight into the control loop. That failure passes simulation and shows up only on hardware. If the layout puts the inductor within a few millimetres of high-impedance analog nodes, pay for shielding.

Rule of thumb: choose shielding from the layout, not from the BOM cost target.

Selection checklist

• Compute L from the ripple target, then re-check ripple at the standard value you actually buy.
• Verify Isat > Ipeak at maximum ambient, and confirm which drop percentage the figure refers to.
• Verify Irms > output current, derated for the enclosure rather than the datasheet's still air.
• Budget conduction and core loss separately above about 1 MHz.
• Feed the inductance tolerance, typically ±20%, back into the ripple calculation.

FAQ

Q: Can I use an inductor rated below my peak current if the overload is brief?

Sometimes, with a soft-saturation composite core, where inductance rolls off gradually and the converter degrades rather than failing outright. With a hard-knee ferrite, no: inductance collapses and current rises almost without limit until the switch's overcurrent protection fires. Check the L versus I curve rather than trusting the single quoted number.

Q: Why does my converter whine audibly?

Audible noise usually means the inductor is operating with large flux swing in the 2-15 kHz band, most often because the converter has dropped into a pulse-skipping light-load mode. Magnetostriction in the core turns that excitation into sound. A moulded part is generally quieter than a wound drum, and raising the light-load switching frequency moves the excitation above hearing.

Q: Does a higher inductance value always reduce ripple?

It reduces ripple current, but for the same package it costs DCR, it slows transient response, and it moves the converter closer to discontinuous conduction at light load. Ripple targets much below 20% rarely pay for themselves.

Q: What does the ±20% inductance tolerance do to my design?

Feed both extremes back through the ripple equation. At -20% the ripple current rises, and with it Ipeak, which is the case your Isat margin has to cover. At +20%, check that transient response still meets the load-step requirement.

Select power inductors with parametric filtering in Novapart's power inductors category, where you can filter by inductance, current rating and package, or start from Bourns, Coilcraft or Würth Elektronik. For matching networks and filters, see RF inductors, and for the other half of the LC filter, capacitors. Sizing a rail? Upload the BOM for consolidated sourcing with cross-references and lead times.