Ferrite beads are not resistors, and the 100 Ω is not 100 Ω

A ferrite bead is the component most often specified by picking the one that says a big number, dropping it in series with a supply pin, and moving on. It is also the component most likely to make things worse when used that way — not ineffective, actively worse.

There are four things worth knowing, and all of them follow from one fact: a bead is a lossy, non-linear, frequency-dependent impedance, and the datasheet describes it with a single number.

The rated impedance is one point on a curve

“600 Ω” or “100 Ω” on a bead is |Z| at 100 MHz, by convention. It says almost nothing about the impedance at 10 MHz or 1 GHz, and it does not distinguish between impedance that dissipates and impedance that merely stores.

Würth model the part as an inductance in parallel with a loss resistance, all in parallel with the bead’s own stray capacitance (ANP045), which gives the three regions every bead has:

Ferrite beads can have different characteristics at different frequencies. These can be roughly separated into three regions: inductive, resistive and capacitive characteristics. At its self-resonating frequency (SRF), the ferrite bead performs as a resistor, impeding high frequency signals and dissipating the power as heat.

And the design rule, stated as plainly as it ever gets:

The intended use of ferrite beads for EMI applications is, that the component must be in the resistive region over the frequency range where the suppression is needed.

That is the whole selection criterion. Below the resistive region the bead is an inductor: it stores energy and hands it back, which is how you get ringing rather than attenuation. Above it, the stray capacitance takes over and the impedance falls away — the bead becomes a small capacitor and your high-frequency noise walks straight through it.

So the question is never “how many ohms”. It is where this part’s resistive region sits, and whether it covers the frequencies that need suppressing. Two beads both marked 600 Ω can have resistive regions an octave apart.

DC bias collapses it, exactly like a ceramic capacitor

Beads are almost always in a supply line, which means they carry DC. Würth are explicit that multilayer ferrites are always operated under current bias, and the whole reason they developed a new high-frequency DC-bias measurement method is that the effect is large and badly documented.

The mechanism is core saturation: DC current magnetises the core, permeability falls, and the impedance goes with it. A bead rated 600 Ω at 100 MHz with zero bias might be a small fraction of that at its rated current. The rated current on a bead is a thermal limit — the current at which it will not overheat — not the current at which it still works.

This is the same shape of mistake as the DC bias loss in class 2 ceramics: the headline number is measured under conditions your circuit does not reproduce, the manufacturer publishes the derating curve, and almost nobody looks it up. Look it up. Pick the part for its impedance at your operating current.

Temperature does the same thing in the same direction. Würth measured a bead from −55 °C to 160 °C and found that the higher the temperature of the component, the more the impedance is shifted down, becoming more saturated. A bead carrying real current in a warm enclosure is doing less than the bench measurement suggested, twice over.

A bead and a capacitor is an LC filter, and LC filters peak

This is the failure that actually damages designs, because it turns a filter into an amplifier.

The standard arrangement is a bead in series followed by a capacitor to ground. In its inductive region — below the resistive region, which is where the switching fundamental of your converter usually sits — the bead is an inductor. An inductor and a capacitor with very little resistance between them is a high-Q resonant circuit, and at its resonant frequency it does not attenuate: it gains.

Würth describe this in the context of an anti-aliasing filter:

The common LC formation of a low pass filter (LPF) may have effects of under damping which creates a resonant peak at a frequency band around the switching frequency of the converter, consequently results in the amplification of unwanted switching noises.

Read that again, because it is the whole point of the article. If the LC formed by your bead and your decoupling capacitor happens to resonate near the converter’s switching frequency, you have built a circuit that makes the supply noise larger at the pin than it was at the input. The measurement then shows more ripple after adding the filter, which is the sort of result that gets blamed on the probe.

The fixes are all about adding loss:

• Choose the bead so the resistive region covers the resonance. Then the bead is a resistor exactly where the peak would be, and it damps it itself.
• Add explicit damping — a small series resistor, or a second capacitor with deliberate ESR in parallel with the ceramic. An electrolytic or polymer part is useful here precisely because it is lossy.
• Move the resonance by changing the capacitance, so the peak lands somewhere with nothing to amplify.

The RC filter calculator covers the first-order case where none of this arises; the moment there is an inductance in the filter, damping stops being optional.

Where a bead is the right answer

None of this means beads are bad. They are the correct part when:

• You want to isolate a noisy digital supply from a quiet analogue one on the same rail, and the noise you are removing lies in the bead’s resistive region. This is the classic AVDD/DVDD split, and TI’s SCAA048 works it through with the ferrite-plus-capacitor pi filter for PLL supplies.
• You want to dissipate interference rather than reflect it. A bead in its resistive region converts the energy to heat; an LC filter with no loss just moves it somewhere else, often back down the cable as common-mode current.
• You need attenuation above where a capacitor has stopped working, since a decoupling capacitor is inductive above its own resonance — the bead and the capacitor cover different parts of the spectrum, which is why the pi filter uses both.

And where they are the wrong answer: as a general-purpose “noise remover” dropped into a supply without checking the frequency, the bias current, or what the bead is about to resonate with.

The four questions

Before you put a bead on a schematic:

1. What frequency needs attenuating? Without an answer to that, the bead is decoration.
2. Is the part resistive there? Read the R and X curves, not the headline impedance.
3. What is the DC current, and what is the impedance at that current? Find the bias derating curve.
4. What does it resonate with, and what damps that? There is always a capacitor on the other side.

Sources

• Würth ANP045 — Behind the Magic of High Frequency SMT Chip Bead Ferrites — the three-region model, the resistive-region design rule, DC-bias measurement at high frequency, temperature saturation, and the underdamped-LC anti-aliasing example.
• TI SCAA048 — Isolating Analog and Digital Supplies in PLL-Based Clock Devices — ferrite-and-capacitor pi filters done properly.
• Würth ANP109 — Impedance Spectra of Different Capacitor Technologies — the capacitor half of the same filter, and where its impedance stops falling.
• TI SNVA871 — Output Noise Filtering for DC/DC Power Modules — post-filter design, and the damping question in a power context.