Ferrite Beads vs. Inductors for High-Frequency Filtering on PCBs

Unwanted high-frequency energy is present on every PCB, and the first instinct of many designers is to place an inductor in the path and block it. That instinct works at low frequency and fails at high frequency. The reason lies in how inductive reactance scales with frequency, where the parasitic capacitance of a wound component takes over the behavior, and why a ferrite bead, which appears on a schematic as an inductive part, is actually used as a lossy element rather than a reactive one.

This article explains the electrical difference between an inductor and a ferrite bead, the frequency ranges in which each is useful, and how a bead converts RF energy into heat instead of simply opposing current. It then works through a set of practical questions taken from real boards and package designs: the ground relationship between an AC adapter and a notebook chassis, the leakage voltage that appears on a floating enclosure through Y-capacitors, the choice between power balls and ground balls in a PBGA package carrying DDR signals, and the many roles a zero-ohm resistor plays in a finished design.

Highlights:

• Inductive reactance rises in proportion to frequency according to XL = 2πfL. An ideal 10 mH inductor measures 628Ω at 10 kHz and reaches 6.2MΩ at 100 MHz, at which point it behaves as an open circuit but still degrades any signal routed through it.
• The parasitic capacitance between the windings of a real inductor limits that component to operation below 1 MHz, so high-frequency suppression cannot be solved by adding more inductance.
• A ferrite bead behaves as a resistor in parallel with an inductor: at low frequency the inductance shorts out the resistance, and at high frequency the rising reactance forces current into the resistive path, where the energy is dissipated as heat.

Inductive Reactance and the Frequency Limit of a Real Inductor

Inductors are used to control EMI inside the PCB. For any inductor, the inductive reactance is directly proportional to frequency, which is expressed by the relationship XL = 2πfL, where XL is the reactance in ohms. The behavior is easy to see with an ideal component. A 10 mH inductor has a reactance of 628Ω at 10 kHz. At 100 MHz the same component presents 6.2MΩ. At 100MHz, therefore, the inductor can be treated as an open circuit.

Being an open circuit is not the same as being harmless. If a signal is routed through such an inductor at 100 MHz, the quality of that signal will be degraded when observed in the time domain. The component has not removed the disturbance; it has changed the shape of the waveform that carries it.

There is a second, more fundamental limit. Just as with a capacitor, the electrical parameters of the inductor itself, in particular the parasitic capacitance between the turns of the coil, restrict the component to operation below 1 MHz. Above that point the part no longer behaves as the inductance printed on its datasheet, and the model shown in the schematic stops describing the physical device.

This raises an obvious question. If an inductor cannot be used at high frequency, what should be used instead? The answer is the ferrite bead.

Ferrite Beads: A Material Selected for High Frequency

A ferrite bead is built from ferrite material, an iron-magnesium or iron-nickel alloy. These materials have a high permeability, and at high frequency and high impedance the capacitance between the turns of the internal coil is minimized. The construction is chosen so that the parasitic effect that ruins a conventional inductor at high frequency is kept small.

Ferrite beads are normally intended only for high-frequency circuits. At low frequency they still retain the full characteristics of an inductor, including both a resistive and a reactive component, and therefore introduce a small amount of loss on the line. At high frequency the bead presents essentially only the reactive component, jωL, and that reactive component increases as frequency rises. In practical terms, a ferrite bead is a high-frequency attenuator for RF energy.

For analysis, the bead is best modeled as a resistor in parallel with an inductor. That simple equivalent circuit explains everything the component does, and it explains why a bead cannot be evaluated by looking at an inductance value alone.

A Dissipative Device Rather Than a Reactive One

At low frequency, the resistance in the equivalent circuit is shorted by the inductance, and current flows through the inductive path. At high frequency, the high reactance of the inductance forces the current to flow into the resistance instead. The transition between these two behaviors is what makes the bead useful.

In essence, a ferrite bead is a dissipative device. It converts high-frequency energy into heat. For that reason, its performance must be explained in terms of resistance, not inductance: the part is doing lossy work on the disturbance, absorbing the energy rather than returning it to the circuit. Designers who treat a bead as a small inductor and expect it to behave like one at high frequency will misread both its purpose and its effect on the line.

Practical Questions on Grounding, Packaging, and Zero-Ohm Resistors

The behavior described above appears repeatedly in day-to-day design reviews, usually in the form of questions about grounds, return paths, and components that seem to do nothing. The following four cases cover the ground relationship between an adapter and a notebook, the leakage voltage on an unearthed enclosure, the assignment of power and ground balls in a package, and the many functions of a zero-ohm resistor.

Q1: Does the ground at the AC end of a notebook adapter and the ground inside the computer, that is the chassis, maintain a very low potential difference? What is the ground relationship between them? How is the inside of the adapter designed? When measuring supply harmonics, are the harmonics generated mainly by the adapter or by the notebook itself?

A1: In theory, both the computer chassis and the adapter ground should be protective earth, with no potential difference between them, connected directly to the ground of the building. The input voltage of an AC adapter is referenced to neutral, while the output is DC and isolated from the input. The output voltage is referenced to the negative terminal of the DC supply. Inside a notebook adapter there is normally an isolated AC/DC supply built around a flyback or forward converter, and switching power supply references describe that topology in detail. A switch-mode adapter will certainly produce harmonics. Whether the notebook itself contributes is not something that can be asserted without investigation, but the adapter is likely to account for a large proportion of what is measured. A useful experiment is to power the notebook from a linear supply and compare the harmonic spectrum; the result should be markedly different.

Q2: The ground of general electrical and electronic equipment is not earth, for example on an electronic load. To prevent static damage, a wrist strap is worn during work, but once the strap is connected to earth the operator is frequently shocked during work. I measured an AC voltage between the electronic load and earth, unstable and more than 100 V. What causes this?

A2: The AC voltage generally comes from the Y-capacitors of the power supply filter, which couple it to the chassis. Once the chassis is connected to earth, the voltage disappears. The value of the Y-capacitor is limited by a requirement: if the earth connection is poor, the leakage current flowing out of the chassis must not become large enough to injure a person.

Q3: EMC problems are currently addressed through peripheral circuits, the PCB, and structural shielding, but EMC behavior is also related to the interconnect routing inside the chip. The following case is an example. I am doing the package design of a SOC in PBGA form, aimed at a four-layer PCB with the stackup signal-ground-power-signal. When arranging the package balls, the usual practice for providing a good return current path for signals and reducing power/ground bounce is to insert power and ground balls into high-speed signal areas in a fixed ratio. I have looked at examples of ball distribution for Intel north bridge and memory controller hub packages. In DDR (high-speed) signal areas, some examples insert both power and ground balls, while others insert only ground balls. In my view, because the DDR interface follows the SSTL_2 specification and uses CMOS output circuits, both power and ground should be present. Yet the literature says: "most return current for a transmission line travels on the nearest reference plane regardless of the direction of current on the trace. It matters not whether the signal transitions from high-to-low or low-to-high, the return current travels on the nearest reference plane." According to that statement, the noise current does not seem to care whether it flows away through the power plane or the ground plane. Why is that?

A3 (first answer): High-frequency signal current ultimately has to return to ground, so once the power pins of the chip already handle the supply adequately, placing ground balls takes priority. Current publications generally accept that the ground plane and the power plane are equivalent for high-speed signals, provided the impedance between the power plane and the ground plane is low enough. On most boards this is not yet achieved, which is precisely the subject of power integrity research, and the impedance of the power plane itself is sometimes relatively large. For that reason, routing should still give priority to return through the ground plane.

A3 (second answer): High-frequency signal current always returns to the source along the loop with the lowest inductance, and the higher the frequency, the tighter the coupling of that loop. Above roughly 50 to 100 kHz a signal begins to show this property. The GND or POWER layers present a series transient impedance relative to the signal trace, and the farther the copper layer is from the trace, the larger that transient impedance becomes. High-frequency return current therefore selects the nearest copper layer, the image plane, as its path back to the driver. If the signal changes layers, the return current crosses at the signal via through the capacitance between the GND and POWER copper planes, and spreads over the inner surfaces of both layers. The signal return voltage drop caused by that impedance is called ground bounce.

Q4: Some resistors are marked 0 ohm. What purpose do they serve?

A4: A zero-ohm resistor has several distinct uses on a real board. It can carry no electrical function at all and simply be placed for convenience during debugging or for design compatibility. It can serve as a jumper: if a section of the circuit is not needed, the resistor is left unpopulated without affecting the appearance of the assembly. It can stand in for a matching component while the parameters of a matching network are still unknown, then be replaced with a part of a specific value once the circuit has been tuned. It can be removed so that a current meter can be connected to measure the consumption of a particular block. If routing simply cannot be completed, a zero-ohm resistor can be added to cross a trace. At high frequency it acts as an inductance or a capacitance depending on the characteristics of the surrounding circuit; the inductive use mainly addresses EMC problems, for example between ground and ground, or between the supply and an IC pin. It provides single-point grounding, meaning protective ground, working ground, and DC ground are separated on the equipment and each becomes an independent system. It can also act as a fuse.

The single-point grounding of analog and digital ground deserves separate treatment. Any ground must eventually be connected together and then taken to earth. If the grounds are not connected, the result is a floating ground, which carries a potential difference, accumulates charge, and causes static problems. Ground is the 0 V reference; all voltages are derived with reference to ground, so the ground standard must be consistent, and the various grounds should be shorted together. Earth is regarded as able to absorb all charge and remain stable, which makes it the final reference point. Even on boards that are not connected to earth, the generating plant is earthed, so the supply on the board eventually returns to earth through the plant. Connecting analog ground and digital ground directly over a large area, however, causes mutual interference, and not connecting them is also unsatisfactory. Four methods resolve the conflict: connect them with a ferrite bead, with a capacitor, with an inductor, or with a zero-ohm resistor.

Each method has a limitation. The equivalent circuit of a ferrite bead is a band-stop filter: it significantly suppresses noise only at a particular frequency point, so the noise frequency must be estimated in advance to select the appropriate part, and a bead is unsuitable when the frequency is uncertain or unpredictable. A capacitor blocks DC and passes AC, which creates a floating ground. An inductor is bulky, has many stray parameters, and is unstable. A zero-ohm resistor is a very narrow current path that effectively limits loop current and suppresses noise, and resistance attenuates across all frequency bands, since a zero-ohm resistor still has impedance. In this respect it is stronger than a ferrite bead.

A zero-ohm resistor is also used as a bridge for the current return path. Once the ground plane has been split, the shortest return path for signals is broken, and the signal loop has to detour, forming a large loop area. The influence of the electric and magnetic fields becomes stronger, and the circuit is more likely to interfere or be interfered with. Bridging the split region with a zero-ohm resistor provides a shorter return path and reduces interference.

The component is also used in configuration circuits. As a rule, products should not contain jumpers or DIP switches, because users sometimes change the settings and create confusion; to reduce maintenance cost, zero-ohm resistors are soldered onto the board in place of jumpers. An unpopulated jumper behaves as an antenna at high frequency, so an SMD resistor gives better results. Other uses include crossing traces during layout, debugging and test access, temporary substitution for another SMD component, and temperature compensation. More often than not, the reason is an EMC countermeasure. Finally, a zero-ohm resistor has lower parasitic inductance than a via, and a via also disturbs the ground plane, because a hole has to be drilled through it.