Why Ferrite Beads?

[SockThief]

Hi!

From a question here, https://designhelp.fedevel.com/forum...clarifications I asked in robertferanec's series on PCB Decoupling, in the final video he "solved" the problem by removing a Ferrite bead and replacing it with a 0-ohm resistor; unfortunately the discussion on what impact this would have was missing

so, I am asking here - what role does the Ferrite play? I have looked at several power supply designs, robertferanec's imx6rex and there is no Ferrite, neither if i am correct does the Switch Mode Power Supply design course; however 28 pins does. I have looked at the Raspberry Pi's Compute Module I/O boards and they have much simpler power supplies, and then i have seen another iMX6 SOM that has a much. much more complex power supply design.

as a beginner in this area, i am rather confused as to what role the ferrite plays, how to design with them, and in the 28 pins example, what happens when we remove the Ferrite?

[qdrives]

Here https://designhelp.fedevel.com/forum...upling-the-pdn was another discussion.
It also has simulation with the results. The results show that the ferrite filters more at some frequencies and at others is worse than a resistor.

[SockThief]

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

I am beginning to realise I may be overthinking many aspects here - looking at many designs you see many different approaches to the same problem and many solutions. In some cases they are simple designs, low component counts etc. There may be excessive noise or other issues, but for the most part it probably don't cause enough of a problem that the extra design effort and component count makes it worthwhile to eliminate. On the other end of the spectrum you have systems designed for industrial use where the reduction of noise is taken to the extreme - where as much noise between circuit blocks is filtered, but also great care is taken such that noise from the outside is not coupled in, and that the device will not radiate noise. The simpler designs are probably more concerned with eliminating radiating interference (for compliance) whereas the latter designs are looking to be stable and reliable in as many installation areas as possible and so take a more cautious/over-engineer route.

it would also appear that designers start with a simple design; simulate, test, measure and then rectify the issues they find. Over time they build up a knowledge-bank based on experience and are then able to predict issues that may happen as they've seen them before and so include remedies in future designs.

Over the last few days I have spent a lot of time talking to many people on many aspects of design - from these conversations there seems to be a case to be made that "noise" is an extremely complex (and often unpredictable) part of the real world. With enough experience you can predict issues, but as robertferanec showed in his videos, you can still be caught out - you can't always predict everything. Instead you design the project to the best of your knowledge and experience and test it. You may run some simulations etc. It is then more important to be able to find the source of noise or interference in your case - there is not a one-size-fits-all-solution.. the designs that are out there are not guaranteed, the solutions work in their situation, but you need to sort out your own design for your own issues (maybe our boards are different materials, more or less layers, longer or shorter tracks etc - there are just too many variables at work). But even then you might get caught out... that's what HW revisions are for.

Now having said all this, there is a wealth of knowledge and experience out there that we can take advantage of - and there are common problems with common solutions, and so of course take advantage of that. At the end of the day, design for what you need - if you can cut a component without sacrificing performance/functionality/reliability/stability (or that sacrifice is deemed acceptable) then go for it.

[qdrives]

[ designers start with a simple design; simulate, test, measure and then rectify the issues they find.

Well, I would say that most designers - design and test functionally and only rectify issues they find.
Some measure (multimeter), less measure with an oscilloscope. Only very few run simulations.

Three years ago I would not be able to answer half the questions on filtering correctly. Hardly did simulations and never heard of power and signal integrity. The only thing I 'knew' about PDN was on voltage drop due to DC resistance.
At that time I would add ferrite beads to many circuits expecting them to filter not knowing that it may in fact make it worse. Heck, from my later experience with simulating those circuits, I would almost say that in most cases it will be worse.
But it is true - there is a wealth of information out there. I hope robertferanec does not stop making those highly educational videos.

[robertferanec]

I used to place beads everywhere, because everyone was doing it - it was in every reference schematic. And explained as - filters the power - it did make sense. However after I went to PDN I started to understand the power delivery is more complex and now I see the power delivery differently as I used to - impedance, resonances, inductance, ...

[SockThief]

oh, thank god! now i don't feel so bad i keep feeling like this is stuff i "should know"... i have just been adding things because its "best practice" and "no one ever got fired for adding a bypass". but this video from dave jones (https://www.youtube.com/watch?v=PgHQ7ES1uDk)

EEVblog 1512 - Why Bypass Your PCB Like THIS?

has taught me otherwise - I am beginning to think that maybe I am digging into things a bit deeper, and that is a good thing, but a bad thing as it tends to slow progress.. but it seems that is what everyone does!

I have noticed also, for example in the recent ESP design video from robertferanec that even data sheets are full of this - for example the section where Robert simplifies the debounce/decoupling as you had two 100nF caps in parallel!

i wonder if I have to start trusting my instinct a bit more, and having faith in my ability, rather than basing everything on "well, Espressif did it, so I must be wrong"... and Robert's videos, as you point out qdrives goes a long way in bringing these issues up to designers like me! and i am eternally greatful for Robert for making be a better design engineer... and he does it for free

[qdrives]

I know that feeling: do I trust the datasheet or the newer theories (Rick Hartley, Eric Bogatin Steve Sandler, etc.) and the way I interpret them.

Do also note that most design are able to work (reasonably) well even if you have 'terrible' power and signal integrity.
However, when you are at the limits, is when the strange things happen -- when it is on my desk it works, but when it is in the machine it fails...

[SockThief]

Too true. At that is also a knock on perhaps of modern tools that are giving us greater insights. Also we are at the (current) end-point of evolution on the subject. Our knowledge is constantly growing, but our application is a generation behind… or we learned from someone that “that’s the way it’s always been” and continued until we stop and ask “but why?” … I am laughing at myself that it took till I was 39 to finally think for a minute, instead of back in school when I had the chance.