6
\$\begingroup\$

I have seen a couple of ethernet PHYs whose manufacturer datasheet recommends placing a ferrite bead on the power supply input of the PHY device whose value is, "Generally, a 100-220Ω (at 100 MHz) ferrite bead is used." Mostly it is on the analog domain of the PHY.

You can find it in the first point under Power Section on page 1 of this checklist.

Microchip KSZ8041MLL Datasheet Schematic

[Microchip KSZ8041MLL Datasheet Schematic DS00002856A-page 2]

Can someone tell me why that value c at that particular frequency of 100MHz was chosen? Any specific reason with respect to the operating frequencies of Ethernet?

\$\endgroup\$
3
  • 2
    \$\begingroup\$ Please show the schematic that has the FB. \$\endgroup\$
    – Andy aka
    May 22, 2023 at 8:08
  • \$\begingroup\$ @Andyaka, it is there in the document figures, page 2 \$\endgroup\$
    – Freshman
    May 22, 2023 at 8:20
  • 1
    \$\begingroup\$ @Freshman FYI this site quality standard require for a post to "stand on its own feet", i.e. include all the information needed to understand it even if the links expire. So you should have included the relevant excerpts from that datasheet (crediting the source). Now user rdtsc has already edited your post for you, but please keep it in mind for the future. \$\endgroup\$ May 23, 2023 at 8:53

4 Answers 4

7
\$\begingroup\$

Almost all ferrite beads are specified with their impedance at 100 MHz. This does not imply that they behave the same at other frequencies:

comparison of Murata BLM A and B type

(source: Murata SMD/BLOCK Type EMI Suppression Filters catalog, p. 19)

Microchip's recommendation does not tell you what frequencies they care about, so they probably want broad-band protection, and any cheap ferrite bead will suffice.

If your circuit generates noise or is sensitive to noise at certain frequencies (which you might know only after testing), then you can choose a ferrite bead optimized for that.

\$\endgroup\$
6
  • \$\begingroup\$ Thank you for the answer. Can you tell me the reason for the specific resistance range at that particular frequency? Any reason for that specific resistance range of 100-200 ohms? Does it have any relation to the operating frequency of ethernet? \$\endgroup\$
    – Freshman
    May 22, 2023 at 7:51
  • 1
    \$\begingroup\$ Microchip does not know the frequency curve of the beads; also note the use of passive voice. They just tell you that circuit designers generally use such beads with this PHY. Such beads are likely to cover the Ethernet frequency range, but the characteristics of your circuit also are important. \$\endgroup\$
    – CL.
    May 22, 2023 at 8:23
  • \$\begingroup\$ I do not get it. can you please explain a little more? \$\endgroup\$
    – Freshman
    May 22, 2023 at 8:37
  • \$\begingroup\$ A typical 100 Ω … 220 Ω ferrite bead on a typical board will work (empirically). Nobody (not even Microchip) knows if yours are typical. \$\endgroup\$
    – CL.
    May 22, 2023 at 8:53
  • \$\begingroup\$ Thank you. Can you tell me what's the bandwidth around the 100MHz frequency? What range around the 100MHz should the ferrite bead cover? \$\endgroup\$
    – Freshman
    May 22, 2023 at 17:29
4
\$\begingroup\$

The value given for the value of ferrite bead is just as inexact as for example why the datasheet given value of specifically 220 ohms for the LED, or why the reset pull-up is given specifically as 10k, or any other value that is not explicitly said why it must have some specific value.

Same applies to why caps are specfically 100nF and 1uF instead of 120nF or 1.2 uF.

As long as the ferrite bead value is within a reasonable range, it will work.

The reason why the ferrite bead impedance is given at 100 MHz is that that's how ferrite beads are rated and sold. Every manufacturer that sells ferrite beads have their impedance listed at 100 MHz.

Not much different how LEDs have their voltage rated at some rated current. Or crystals have their initial accuracy listed at some rated temperature.

So likely this range of 100 to 200 ohms at 100 MHz is good enough range, and most typical ferrite beads will have suitable impedances at lower and higher frequencies too.

It may also mean impedances higher than 200 will be just useless as it does not need to filter so much and it might be poor for normal operation if there is too much impedance at frequencies the chip works as. High impedance ferrite beads also have higher DC resistance so voltage may drop too much through them.

It may also mean that impedances lower than 100 don't filter enough and there is too much noise left unfiltered.

So no different than having a LED with series resistance out of useful range - too dim or too bright for useful operation, but ferrite bead just works for AC currents instead of DC current.

The example given is just typical range what you would typically use.

\$\endgroup\$
5
  • \$\begingroup\$ Thank you for your answer. So, since this specification is mostly generic, can the same apply to other interfaces as well? Like for example, can the same ferrite bead used in the power supply for an USB hub or transceiver? Any idea on that? \$\endgroup\$
    – Freshman
    May 22, 2023 at 17:39
  • \$\begingroup\$ That's not what it means and it can't be implied. Using a range of 100 to 200 ohms for one chip does not mean it is acceptable to use the same range for a different chip, not even a different Ethernet chip, unless found suitable. Just like you can't use a 220 ohm resistor for all LEDs if you have a 5V supply, or for all LEDs and all supply voltages. But, a 220 ohm ferrite bead can be used for many purposes, just like many LEDs can be used with a 220 ohm resistor. \$\endgroup\$
    – Justme
    May 22, 2023 at 17:56
  • \$\begingroup\$ @Justme "High impedance ferrite beads also have higher DC resistance so voltage may drop too much through them." How can this be? Beads are just a piece of ferrimagnetic material that can be inserted over an already present conductor. How can the material resistance influence the voltage drop at DC across the wire that pass through them? \$\endgroup\$ May 23, 2023 at 10:21
  • \$\begingroup\$ @LorenzoDonatisupportUkraine An SMD chip ferrite bead is a piece of wire surrounded by the material. There's more/longer/thinner wire on a SMD chip bead when it has higher impedance. This isn't about pure beads you apply to piece of wire yourself, but ready made components. \$\endgroup\$
    – Justme
    May 23, 2023 at 10:31
  • \$\begingroup\$ @Justme OK, thanks for the clarification. So essentially they are pre-built series RL circuit. \$\endgroup\$ May 23, 2023 at 10:52
3
\$\begingroup\$

Ferrite beads make crude filters. The intent is to have some attenuation between main and local supplies, potentially to reduce EMI output and improve signal-to-noise ratio (SNR) (i.e. maximizing range).

When it comes to filtering, there are much better options than FBs to choose from, and one might consider improved filtering to many parts of such a circuit. To determine this, one needs to do an EMI and signal quality analysis with a complete assembled unit.

Most likely, basic operational testing will be satisfactory, no further testing will be performed, and the ferrite beads will be left in as dummy elements -- whether they're actually needed or not.

In other words: it's a business decision, where the complete testing and justification for these components (ferrite beads, bypass capacitors, or other filtering elements if used) is extremely expensive, while the expected value is extremely low. Perhaps a few users will be annoyed that the unit does not meet strict IEEE 802.3-part-whatever performance levels. Such users can be directed to alternatives: use better quality cable, add repeaters/switch(es) along the route, accept degraded bandwidth (maybe it only enumerates at 10Mbps instead of the 100Mbps it's supposed to), etc.

Real engineering is full of compromises from a position of ignorance. When the cost to investigate and clear away that ignorance exceeds its expected value, we fall back on assumptions and tradition. I would say ferrite beads in these locations are traditional in this sense.

\$\endgroup\$
2
  • \$\begingroup\$ Thank you for the answer. Can you tell me why that value at that particular frequency of 100MHz was chosen? Any specific reason with respect to the operating frequencies of Ethernet? Or can this value of the ferrite bead can be used for say other interfaces, if applicable, like say, USB \$\endgroup\$
    – Freshman
    May 22, 2023 at 11:40
  • \$\begingroup\$ The purpose of my answer was to illustrate specifically how the value is nearly irrelevant. Any typical value might work, including zero (short circuit). Smaller values offer less filtering value; larger values may be prone to saturation at the operating (DC) current. \$\endgroup\$ May 22, 2023 at 11:46
3
\$\begingroup\$

Can someone tell me why that value c at that particular frequency of 100MHz was chosen? Any specific reason with respect to the operating frequencies of Ethernet?

Because of conducted emissions from the digital VCC's. Most of the switching of transistors in the KSZ8041 happens in the 100MHz range. They want to isolate the noise from that from getting into the rails of the analog section. It's costly to provide another Vreg so the cheaper solution is to filter an existing digital rail.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.