What's all this ferrite bead stuff anyhow?

[gnuarm]

Working on a group design of a control board for a ventilator.  Ferrite beads are popping up in many places, mostly from external ports such as a keypad and an I/O connector.  BLA31BD121SN4D
https://www.murata.com/en-us/products/productdata/8796743204894/ENFA0008.pdf

I can't tell much about this part.  it has an impedance of 120 ohms at 100 MHz and a DC resistance of about 0.4 ohms.  Should I assume it is a simple inductor and calculate the inductance from the 100 MHz impedance?  Or is the impedance largely from the dissipative properties and so mostly resistive at 100 MHz? 

I guess my concern is figuring out how this is going to work at frequencies other than 100 MHz. 

There will be some EMI noise level from a brushed DC motor about a foot or 18" away.  I'm more worried about that over the cable from the motor control and power board.  These board I/Os are points where static can get in to damage chips.  A TVS diode is added to each line, but with values like 9.2V breakdown and 14.5V clamping voltage I'm not sure how that will protect 3.3V circuits.

I've always used a series resistor of a value that isn't obtrusive and a pair of Schottky diodes to the rails.  For the push buttons a series resistor of even 10K could be used.  For the UART a value that allows adequate baud rate operation will need to be calculated.

Is a ferrite chip more useful somehow for protecting against static than a resistor and diodes to the power rails? 

[penfold]

Ferrite beads are very often just not supplied with the correct specifications you need to do any serious design work with. Wurth Electronik tend to provide a frequency plot for most of their offerings which can aid design and selection hugely. In most circumstances, ferrite beads are intended as a bit of a safeguard in suppressing the high frequencies rather than producing a 'cleverly designed' filter, so picking one  based off its impedance at 100MHz and what values of capacitance or resistance your design can tolerate to produce 'good enough' attenuation then simply assume it will roll off a bit quicker than a second order filter and be less prone to resonating.

Ferrite beads with multiple turns, particularly with thin wire aren't particularly well suited to limiting ESD pulses since they can arc between the turns.

Like you suggested, using a TVS diode (or GDT or similar) - which I'm certain you can get them with breakdown and clamping voltages much closer to 3.3V than 9.2V and 14V - along with some Schottky diodes is a fairly common approach. Driver ICs with built in ESD protection are a good option for data lines, Wurth have a range of low capacitance ESD TVS's and fast ESD clamp diodes which are also a good option when you don't have built in protection but need to preserve some speed.

[srb1954]

I can't tell much about this part.  it has an impedance of 120 ohms at 100 MHz and a DC resistance of about 0.4 ohms.  Should I assume it is a simple inductor and calculate the inductance from the 100 MHz impedance?  Or is the impedance largely from the dissipative properties and so mostly resistive at 100 MHz? 

I guess my concern is figuring out how this is going to work at frequencies other than 100 MHz. 

Is a ferrite chip more useful somehow for protecting against static than a resistor and diodes to the power rails?

Ferrite beads are generally used for EMI suppression and it is important to know their impedance characteristics over a wide range of frequencies. However, with only a data point in the impedance specification point it is difficult to make any useful predictions regarding the wide-band performance. You may have to do some of your own measurements with a VNA to determine the impedance vs frequency curve.

Knowing the impedance vs frequency you can model a bead as a lossy inductor with a parallel resistor equal to the peak impedance versus frequency. Above the peak frequency the impedance falls off and this effect can be modelled with a small parallel capacitor. This model can be quite accurate for beads that have a sharp impedance peak but it is less accurate for those that have a broad impedance peak.

I wouldn't use a bead as primary protection for ESD or other large transients. They might take the edge off fast ESD transients but won't offer sufficient amplitude control to protect sensitive semiconductors. The old solution of clamping diodes or TVSs is usually the preferred method.

 

[Berni]

Ferrite beads are just inductors that are really really lossy on purpose.

At very low frequency they have a small DC resistance and are almost the same as a dead short.
At higher frequencies the small inductance of it starts to show up in the form of a bit of inductive reactance and affect the signal much like a small value inductor would.
At even higher frequencies due to AC losses the ESR starts to rise, causing the resistive component to rapidly overshadow the small amount of inductive reactance.

So your particular ferrite looks about the same as a 120 Ohm resistor if you hooked it up to a network analyzer and swept it at 100MHz, but at 1KHz it looks like a 0.3 Ohm resistor.

This makes ferrite beads very useful in keeping down EMI noise. It lets the low frequency signals pass trough unmodified, but restricts the flow of high frequency noise while burning it as heat on its resistance. This is better than using a regular inductor because inductors maintain a low ESR into high frequency, Yes they still have reactance that makes high frequency difficult to pass trough, but it also stores the given energy in the inductance and releases it later rather than dissipating it. This makes inductors love to ring and oscillate with capacitance around it, causing the EMI problem to potentially get even worse.

As penfold as said this 120 Ohms @ 100MHz is not always a useful number. Its preferred if the datasheet gives you a graph of impedance vs frequency, this lets you use the graph to chose a ferrite that has very little resistance at the signal frequencies you use (so that it distorts the useful signal as little as possible), but at the same time has as much impedance as possible at higher frequencies so that they get blocked.

For ESD protection it probably helps a little bit but not much. The best way to help your ESD diodes be more resilient is to add a regular film resistor between the ESD diode and the outside world connector. This resistor limits the current that can enter and helps dissipate it as heat. It also saves ESD diodes from blowing up if a large DC voltage is connected. Say you have a 3.3V input and someone mistakenly connects it to a 12V battery. This normally kills ESD diodes because it causes a large continuous current trough them, but if you have a resistor that resistor will limit the current to a few miliamps, burning most of that power on the resistor while the ESD diode sinks what is left.

[T3sl4co1l]

Fantastic datasheet... they don't even put the characteristic curve in it!
https://www.murata.com/en-us/products/productdetail?partno=BLA31BD121SN4%23
Still SOL about saturation current though, or capacitance or other coupling between adjacent elements.  (Bead arrays don't usually couple much -- even ones sold as "common mode chokes" may be more normal than common mode, YMMV.)

Assuming it's a lone element, then that's how much impedance it adds.  It's largely resistive above 100MHz, and still has a low Q below there (annoyingly, they show R going to ~zero, which is absurd..).

Put another way, yes it has inductance, but the inductance varies strongly with frequency, because its Q is low.  At low frequencies, it might measure about 1uH, but DCR dominates; at mid frequencies say 1MHz, it might measure 0.3uH, at 10MHz it might measure 0.1uH, etc.  (Well, in this case it'll be at least 0.19uH at 100MHz, and more below that, but I'm just saying as an example.)

Z ~ sqrt(F) is a better estimate for most beads, implying both the low Q (exactly Q = 1, if the condition holds), R ~ sqrt(F), L ~ 1/sqrt(F), and phi = 45°.  Such an element is described by diffusion phenomena, and an ideal one is known as a Warburg element.


Typical scenarios:
- Power filtering: best to avoid.  Use current-compensated (common mode) chokes if possible, or plain old inductors.  (Note that inductors aren't generally very lossy, so you may need to add lossy capacitor -- typically an electrolytic, or ceramic+resistor.)  Ferrite beads typically saturate (inductance/impedance drops substantially under DC bias current) at low currents, maybe 10s of mA for a part this size.  So you might get expected performance under some conditions, but peaks still pass, or EMI (or supply ripple, or..) varies with operating condition.  Sneaky.

- Unshielded multiconductor cables: adds impedance at RF, providing some termination value.  Whereas a short-terminated wire or cable can resonate strongly, the termination kills the Q of those resonant modes, potentially reducing emission / susceptibility significantly (~10dB?).

- ESD, EFT: basically no effect.  If they don't arc over in the first place, the ferrite saturates quickly (peak current of amperes!), and doesn't add much impedance even if it were linear (the source impedance is ~100s ohms typically).

(Actually, saturation can even cause pulse sharpening, maybe even making things worse -- I'd guess it's pretty rare to have the exact conditions for this to be noticeable, but it's interesting that it can happen.)

Accordingly, they make bad filters.  If you're getting EMI coming in some cable harness, and it's not at resonant frequencies, adding beads is only likely to reduce it by a few dB; even a big stack might only do 10 or 15dB.  Again, the impedance is small: compared to a ~100s ohm cable over ground plane, or in free space, a 100 ohm bead doesn't do much.  But this also shows under which conditions they can be useful: if a shielded cable is grounded in several points (say, to the enclosure at either end of a run), then the impedance near those grounds can be quite low (1s-10s ohms), and significant filtering can be had.

Tim

[Berni]

EMI Ferite beads don't saturate that easily.

They have a rated current and should work fine within it. The correct kind of ferrite works well for power applications. I often use them around switching regulators to keep the noise contained at the regulator. They don't need to even reach very large impedance to be effective, even something like a few ohms can help a power filtering capacitor do its job a lot better and make sure any weird resonances don't show up. Splits in ground planes are also good places to use ferrites since if you limit the rise time of a track going across the ground plane split this also limits the rise time of the return current trying to make its way around the split, making it behave less like an antenna and more like just DC current.

[RoGeorge]

I guess my concern is figuring out how this is going to work at frequencies other than 100 MHz. 

In one line, a ferrite bead is used to reflect noises back to the side from where the noise was coming from.

[gnuarm]

All great posts.  Thanks!

[S. Petrukhin]

Tell me if I'm wrong, but I was afraid of these beads.

Logically, they absorb the pulse of high-speed interference, but they do not convert it into heat, but accumulate it in the form of a magnetic field, therefore, at the end of the pulse, we will receive an injection of the pulse of the reverse polarity of the EMF.

[mvs]

Logically, they absorb the pulse of high-speed interference, but they do not convert it into heat

Ferrite beads are made out of high permeability ferrite, that has very high losses above 100KHz.
High speed interference will be mostly dissipated as heat.

[S. Petrukhin]

Logically, they absorb the pulse of high-speed interference, but they do not convert it into heat

Ferrite beads are made out of high permeability ferrite, that has very high losses above 100KHz.
High speed interference will be mostly dissipated as heat.

For some reason, I do not remember any characteristics in the specification, except for the DC resistance and the reactance at 100 kHz.

[mikerj]

Fantastic datasheet... they don't even put the characteristic curve in it!
https://www.murata.com/en-us/products/productdetail?partno=BLA31BD121SN4%23

Murata have their excellent SimSurfing web interface that shows all the characteristic curves and lets you download a spice model for most parts.  Unfortunately they do not appear to list ferrite chip arrays, but I think the equivalent single part would be a BLM18BD121SN1.  Max current is lower on the array as would be expected.

[gnuarm]

Logically, they absorb the pulse of high-speed interference, but they do not convert it into heat

Ferrite beads are made out of high permeability ferrite, that has very high losses above 100KHz.
High speed interference will be mostly dissipated as heat.

For some reason, I do not remember any characteristics in the specification, except for the DC resistance and the reactance at 100 kHz.

I believe you are mistaken in calling it reactance.  While it does vary with frequency like reactance, it is not storing energy so much as dissipating it as heat.  Consider it to be a resistor to magnetic field changes.  So it has no impact on DC, but higher frequency AC gets absorbed and turned into heat. 

When people explain this they often refer to a graph of the field.  it has hysteresis which results in less energy being returned to the source than if it were a pure inductor.  That energy difference is the loss turned into heat.

[TimFox]

If you integrate B H around the hysteresis loop, that is the energy per volume dissipated by one cycle around the loop.  Therefore, the dissipated power increases with frequency (same energy in less time is more power).
 The dissipated energy heats the ferromagnetic material.  Ferrites  with large hysteresis (big areas in the loop) are good for this, while “softer” ferromagnetic materials (narrow hysteresis loops) are appropriate for normal  transformers and inductors.

[Berni]

Storing the energy in the field is what the difference is between an inductor and a ferrite bead.

High frequency inductors also contain ferrite cores, but they use a different blend with various added materials. With this the core material can be widely optimized for certain tasks. In the case of an inductor it is optimized for low loss and high saturation along with a decently high permeability, since this will make the inductor return as much of the energy back out and store as much energy as possible.

The ferrite material in ferrite beads on the other hand is optimized to be high loss. Turning as much of the AC magnetic field on it into heat. Some of it might be BH magnetizing losses some might be eddy currents etc. Another easier way to imagine it is having an inductor but also with a thin wire shorted turn wound on it. At DC the shorted turn is doing nothing, but as you go up higher in frequency that shorted turn starts to have more current flowing trough it, causing it to get hot due to resistive losses so it dissipates the power.

When it comes to good data Wurth elektronik does a particularly good job. They have an online tool that helps you find the appropriate ferrite or inductor for the job:
https://redexpert.we-online.com/redexpert/#/module/1/infopanels/FZI/sel/10MHz:9.25O:/productdata/=74279270/applicationbar/System/on/systemApplication/1/noise/15dB+250MHz+0A/source/1O/sink/1O
It provides full graphs of its performance both electrically and thermally. If you give it the application data it also calculates various things for you, like for example you tell it what Buck DC/DC setup you have and it will calculate both DC and AC losses(this one is tricky to do by hand) on the inductor and tell you how warm it will get.
We use a lot of Wurth components because the datasheets have nice data and graphs and they provide ready to go CAD libraries for all the parts with proper 3D models and everything. You will also find accurate simulation models of there inductors and ferrites in LT Spice. They also happily provide free samples trough there local representative and let you buy directly in 100 quantity at prices that are 1/2 to 1/4 of what distributors like Farnell/Digikey/Mouser etc sell there stuff at.