Tip "Use ferrite beads or 0R resistors on powers Robert Feranec"

[gyuunyuu1989]

The video from guru Robert Feranec titled "TIP #035:Use ferrite beads or 0R resistors on powers" is quite helpful. It is short and to the point. This however leaves certain questions unanswered.

When it says use ferrite bead or 0R resistor, it does not describe any attributes on what criteria to use to choose a ferrite bead.

It also seems to imply that ferrite bead or 0R resistor both do the same function. Maybe I am wrong about this.

Also, if we are going to use a ferrite bead instead of 0R resistor, is it still possible to measure the current on the supply rail?

Finally, when it says 0R resistor, it means that it is a value very smaller than 1ohm but not 0. But resistors have tolerance. Also, the resistance changes with temperature rise and possibly other factors. How exactly does a person measure accurate value of current through the 0R resistor when its value is not exactly fixed for every single part?

[qdrives]

They both serve the same function -- the ability to isolate parts of the circuit for diagnostics. When there is a short, you can take out the resistor and measure on which side the short is.

Measuring the current on the supply rail has no direct influence on these 2 components as neither of them can be used for this. If you want to use such for current measurement, you should add a known low resistance resistor (like 10mOhm) and measure the voltage across that.

Ferrite beads can cause problems for power supply stability. robertferanec has made a video (with Florian Hammerle, Omicron Labs?) on replacing the used ferrite bead with 0R. Problem is that Robert has too many videos for me to find it.

[gyuunyuu1989]

I thought ferrite bead is specifically used to filter high speed noise. I am not sure why it would be grouped with 0R resistor.

Ferrite bead making power supply unstable, I did not know such a thing was possible.

[binayak]

Ferrite beads are not doing the same function as a zero-ohm resistor.

Ferrite bead (along with a capacitor) forms a series-PDN element to filter out the noise coming from a noisy power rail towards a sensitive IC. It is NOT to be used to filter out the self-generated noise of the IC when the IC switches.

You can check articles and videos of Dr. Istvan Novak for how to design such series PDN filter (ferrite+capacitor)

[qdrives]

I found the video of Robert: https://www.youtube.com/watch?v=hZSOhVdzqZk

[gyuunyuu1989]

I have seen all three videos of which this was part 3. It is about PDN and shows how the PDN in this specific PCB was fixed by removing the ferrite bead that was creating a large peak in the PDN impedance for the microcontroller VCC at lower frequencies.

The fundamental question remains unanswered: When it says use ferrite bead, no method is given to decide on the attributes of these ferrite beads. Basically when I was looking through several FPGA board schematics, I found that some of them were using ferrite bead on some of the clock generating IC's power pin. However, all clock source IC power was not connected via a ferrite bead and all the FPGA schematics were not using ferrite beads. This made me very confused. It was not at all obvious why only specific ICs were chosen for the ferrite bead and also how the decision was made to use a specific value and package for the ferrite bead. From what I know, the ferrite bead is used to filter high frequency noise. What is not at all clear is, why only certain places were chosen to use the ferrite bead, how the ferrite bead's value and package were chosen and also how does this differ from using an inductor (ferrite bead vs inductor difference). Finally, we want to minimize the inductance between VRM and power pins on ICs so using something like a ferrite bead seems to be a major contradiction. This contradicton is strenghtened after watching that video you mention in your post.

One schematic of interest is attached to this post as "C10_GX_DK_Rev_A1_Release.pdf". The ferrite bead's name starts with FB. Please search using string "MHz" to find the clock sources. It has been used in most places but not on component Y7.

[robertferanec]

I am planning to re-do that video. I have learned a lot since and using beads in power rails may not be the best idea. I think, for small currents maybe it could be useful (and if you know the exact frequencies in the power), but for higher current power rails (and unknow frequencies) it could be a problem. (high current = when possible voltage drop will by higher than input voltage tolerance)

PS: I will be doing a video about bead selection

[qdrives]

Ferrite beads are used to filter (high) frequency noise......

To start with the question on what the difference is between a ferrite bead and inductor -- A ferrite bead is more lossy. It turns the noise into heat, whereas an inductor is used to store the energy and release it later. However, the reality is not perfect, so the ferrite bead also acts like an inductor and the inductor has losses.


The most important thing to think about is -- what noise do you want to filter?

In the past I used ferrite bead regularly to filter the switching noise from the microcontroller to the rest of the system. Or from the RS232 driver to the rest of the system.
I also hoped that it would filter noise from the switching regulator a bit more. This last bit is (often) wrong as the anti-resonance and reduced regulator effect are more problematic than the noise itself.

In your example for FPGA I can imagine that a ferrite bead to the power input of a clock generator will prevent a lot of noise on the power rail. A clock generator does not draw a large current - on average - but does have high peaks. At a known frequency, so things can be simulated, measured and tuned.

As for selection:
- Current capability (do note that ferrite beads do not like overloading.)
- DC resistance (voltage drop + heat)
- Size
- Frequency band
- Price
- Availability
But lets hope robertferanec video will make it a bit more clear.


And yes, it is a contradiction. Like the three value capacitor myth, one that will remain for a long time. Like most engineers, I was guilty of doing it too.

[gyuunyuu1989]

It seems that stuff like the 3 capacitor myth and other things related to PDN are only properly understood by very few people. I mean most people will just follow what they have seen engineers do in the past or their university teacher. However, since the digital circuit operating frequency (going up) and voltage (going low) and currents (going up) have changed, a lot of things that were true in the past and sufficient are not sufficient anymore. But the university and most engineers are not fully aware of this since they did their basic knowledge building a long time ago. The true answer to these can be obtained by anyone but only by using very extremely expensive tools (software and hardware) and a lot of tedious measurements.

The subject of PDN seems to have a whole lot of contradictions in it simply because the nature of the device we are focussing on, its operating frequency, its operating voltage and acceptable ripply on it, and the currents it needs and a lot of other stuff on the PCB can make each case unique so there is no one size fits all if we want to get the best efficiency. What works in one case to improve performance will actually cause reduced performance in another case. It really is confusing and there are few people like Eric Bogatin who really know how to tackle all this with the current generation of devices.

I will wait for Rober's video which will explain how to choose ferrite bead and how to know if I should even use one in the first place.

[qdrives]

To start of: https://www.youtube.com/watch?v=y4REmZlE7Jg

The true answer to these can be obtained by anyone but only by using very extremely expensive tools (software and hardware) and a lot of tedious measurements.

No I do not think so. The first step is being made aware.
Then for some you can use simple/free tools to check. Take LTspice for instance and ferrite beads. In my comment here https://designhelp.fedevel.com/forum...upling-the-pdn I showed multiple 'solutions' to filter and you can see the effects.

can make each case unique so there is no one size fits all if we want to get the best efficiency

There are multiple issue:
- Cost of the best solution. Take the flat PDN of Steven Sandler. You may need to add low resistance resistors which cost money and take board space.
- Simulate the solution -- takes time and may require advanced simulators
The first question to ask: is it worth it and is it required?