3
\$\begingroup\$

When using ferrite beads in spice, I often have the impression that they are not lossy enough compared to the real thing. In particular, the conventional model of RLC all in parallel is basically a short at frequencies well above GHz.

I realize that there is interwinding capacitance, and that HF currents therefore can take a shorter path through the component, but they still have to pass beside the ferrite material, which still has a large loss term at GHz (even peaking at the ferromagnetic resonance frequency in case magnetic material is used). So the resistive behavior should still matter when the part is well in its capacitive regime, right ? Not at its full strength but it shouldn't be ignored, I guess.

Below, I compare the standard bead model to a customized one, where the capacitance is divided into smaller capacitors each with a bit of series resistance.

enter image description here

The effect is that the phase stays much closer to 0° beyond the peak impedance and never really drops to -90° as it would for a pure capacitor.

enter image description here

The transient behavior is also much more benign compared to the standard bead model. Here, a 1 Volt step is applied to the bead into a 100 pF capacitor with a small series inductance of 500pH, and capacitor voltage is monitored:

enter image description here

So again, is this modified bead model reasonable or is there some good physical reason to ignore loss in the very high frequency regime?

\$\endgroup\$
6
  • \$\begingroup\$ Inter winding capacitance does not form turns around the core hence, the core losses are out of the equation hence at high enough a frequency you are left with pure ohmic wire/connection conduction loss and no losses to do with the ferrite. Have you tried looking at typical graphs in data sheets? \$\endgroup\$
    – Andy aka
    Dec 18, 2021 at 14:19
  • \$\begingroup\$ Just about any magnetic model will involve not only a lumped model for the frequency, but also one for the distributed RLC(G), if you plan on analyzing the model, itself, in which case there will likely be some FEMM analysis which will give the approximation for the SPICE. So what you have there is one section, which would be repeated, with shunt capacitors between stages (a multiplied Pi section). Something like this (V(b), a crude approximation given your values; for when you need a (R)good enough model). For simulation only, a simple model will do. \$\endgroup\$ Dec 18, 2021 at 17:20
  • \$\begingroup\$ Can this will help modelling? Google search "Behind the Magic of High Frequency SMT Chip Bead Ferrites" from Wurth elektronik, eq3 . Something as these beads? Until 10GHz ... mouser.be/datasheet/2/445/742841210-1720802.pdf \$\endgroup\$
    – Antonio51
    Dec 18, 2021 at 17:41
  • \$\begingroup\$ Here a simulation with a "serial" model i.stack.imgur.com/nAWbY.png \$\endgroup\$
    – Antonio51
    Dec 18, 2021 at 20:58
  • 1
    \$\begingroup\$ Quite simply, a clamp on core does actually form a strong container for the magnetic field created by the current. A FM isn't a closed ferrite so that loss is much less. You really do need to pick an FB that has a full characteristic so you can see the resistance drop back down to low values beyond the self-resonant frequency. \$\endgroup\$
    – Andy aka
    Dec 18, 2021 at 22:05

1 Answer 1

3
\$\begingroup\$

TL,DR:

The usual parallel RLC circuit is inadequate to model ferrite beads. Their "capacitive" frequency range is substantially resistive, due to ferrite losses.


Long story:

I found this page from TDK about a specific high loss bead. They offer both a simple (RLC in parallel) model for spice and a "precision" model. The precision model looks rather complex but when looking at the netlist, it becomes clear that there is no pure capacitance present and there is always at least 20 Ohms of resistive impedance in series at high frequencies.

Spice confirms its resistive behavior at very high frequencies and also the transient behavior is very different from the simple model at high frequencies. So it seems that the resistive behavior at very high frequencies is an essential part of a better ferrite bead model, which the simple model does not capture.

enter image description here

enter image description here

Another high loss bead I use is this part from Taiyo Yuden. Their spice model is also more detailed than the usual RLC circuit. It nicely reproduces the datasheet curve, but what is more interesting, the behavior above 3 GHz first turns again slightly inductive before becoming substantially resistive at an even higher resistance than its rated peak resistance.

enter image description here

While it can be argued that it is not accurate to expect such a high resistance, I think that my experience with beads, as well as models from these two very reputable bead manufacturers agree, that beads are not a capacitive short at those very high frequencies. They are still very lossy, because current still passes by the ferrite core and magnetizes it.

\$\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.