Use of ferrite beads in decoupling the PDN

[Tom Yunghans]

Hi Robert,

I just watched part 1,2 and 3 of your "decoupling capacitor" videos. Very informative. Thank you.

It appeared that the ferrite bead in the path to the VCC pin was causing undesirable ringing on the VCC at the processor pin. We usually place beads like that to keep noise on the PDN from getting to a sensitive analog VCC, but I would think it might also be useful at keeping noise from a noisy internal rail from getting back onto the PDN. I kept waiting for your conclusion explaining why it was a bad idea to place that ferrite bead, I assume that would have been your conclusion. From some of the You-tube responses, it sounded like you may have addressed that in a later video. Can you identify which video that was?

I believe that you also mentioned the fact (or I have heard it from other sources) that you have to be careful about what value of capacitors you place in parallel as the combination of parallel resonances can also cause problems. Did you discuss that in other videos?

Thank you for these videos, I find them very informative.

[robertferanec]

These topics are still on my video todo list.

But, very simply to say, the BEAD basically added an extra inductor into the circuit and together with the capacitors it created a resonance circuit - and that was the problem.

[qdrives]

With a simple simulation done in LTSpice (using spice models for the capacitors) I got these results:

1) The first "bump" (red line, eft blue circle) if a resonance with a higher gain. So if you SMPS is switching in that frequency range, the bead would create more ripple and not less.
The solution to this is to add a snubber. This results in the green responce.
2) Adding two different capacitance capacitors in parallel (violet line), causes an anti-resonance(?) bump (middle blue circle).
3) The ferrite bead does cause less filtering at higher frequency (right blue circle), but do note that the resistors, traces and VRM itself are ideal in this simulation.

The bead + snubber is the most expensive solution and take the most board space, but perhaps the higher filtering is preferred/required.
Rule of thumb on the snubber: C must be 10x higher.

Capacitors are 100nF 0805 100V X7R (C0805C104K1RAC)
and 1uF 1206 100V X7R (C1206C105K1RAC).
3.3V DC bias voltage in simulation, but 48V in the spice models.

Hope this helps.

[robertferanec]

qdrives very nice!

[gommer]

qdrives , this is a very nice reply and I've come to similar conclusions while simulating with LTSpice myself.
Can I note one small thing about semantics? Used like this, the RC networks are called damping (damper, dampening) circuits. They serve indeed a similar purpose as snubbers, but the term snubber seems to be used exclusively in damping parasitic properties of switching semiconductors (usually output capacitance of FET's in combination with parasitic inductor PCB traces).
Adding this because it might save some readers confusion when googling this topic.

[qdrives]

gommer Well, I use it also to "dampen" the ringing of switching FETs ;-)

[binayak]

Hi Tom,

Ferrite beads can be effectively used as a part of "PDN filter" in very specific cases. It is a scalpel - not a blunt sword.

There is a very nice video by power integrity master Dr. Istvan Novak in Samtec's gEEkspEEk webinar series titled "Using Ferrites and Inductors in Power Distribution Networks (PDN)". Please go through it and clear all your doubts on when they are needed, and how to use them in the design.

URL = https://www.samtec.com/support/video...pdn-487288222/

Hope it helps.

Regards.