Ferrites as a Filtering Element - Are They Effective?

[@TheDutchGuyOnYT]

Can somebody put a ferrite between Altium and the stupid idea to cancel perpetual maintenance? I’m very angry.

[@TheDigital19]

in a few years when they have lost a ton of customers they hopefully realize that this was their biggest mistake in history

[@sanjaybatra6593]

means ?

[@rfrisbee1]

The main difference between a ferrite bead and an inductor when used as a filter is that the ferrite bead, if chosen correctly, will convert the noise energy into heat, whereas an inductor will just cause it to slosh around until some other lossy element(s) in the circuit convert it to heat. The complicated part in the above is the "if chosen correctly". The noise energy must fall within the lossy frequency range of the bead and the bead must be modelled at the DC current it is going to run at, as the parameters change with current! As for datasheets, well we can't even get manufacturers to stop recommending parallel 10nF + 1uF (or similar) surface mount capacitors for IC decoupling, when using a single 1uF ceramic surface mount capacitor would be better (cost, size, performance) in the vast majority of cases. Good luck getting them to stop recommending the use of ferrite beads on power rails to ICs!

[@sanjaybatra6593]

can you please share more information in sample language.

[@radsdau]

Zach, you mention the video links in the description, for info about how ferrites add noise when you're trying to remove noise... did I miss that link? Seems like something I need to learn more about. Thanks.

[@AltiumAcademy]

Check these out: Three Frequency Ranges For Ferrite Bead Usage: resources.altium.com/p/three-frequency-ranges-ferrite-bead-usage and Ferrites in PCB Design: What the Experts Say: kzbin.info/www/bejne/bqK5lpivYpaifJIsi=Nu9JzSLsTXEy_6bK

[@FriedemannSchmidt-te9kq]

Can you please carry out a simulation with higher capacitor values, with at least 10µF?

[@Rubacava_]

You also have to account for series resistance and the load current. Ferrite is basically an inductor + resistor. It doens't have that resonance but you can only power so much with it. It is usually cheaper and smaller too so it still has its uses.

[@myetis1990]

Thank you Zach, for the great video. While I was trying to take a look at ferrite beads, I learned something nice about the PI filter, thank you.

[@Fisch4678]

Thank you for the nice video :) I agree that the damping of an LC filter is generally better than that of a ferrite bead filter. But a real-world inductor will also have some parasitic capacitance, which will dominate at frequencies above the self-resonance-frequency (SRF). So the AC transfer characteristic of a real-world LC filter will also be messed up at high frequencies... just like the ferrite bead. Furthermore, a real-world capacitor will also have some parasitic inductance, which will further degrade the filter's performance at high frequencies in each of the three designs. I sometimes add a ferrite bead pi-filter on the output of a DC/DC converter to provide some damping for high-frequency noise (frequencies at which the regular LC filter of the converter doesn't work due to the parasitics). All the ferrite beads I've come across have a specified bandwidth of 100MHz, not 10MHz. My intuition always was, that an inductor with a similar SRF must necessarily have a very low inductance, which results in an increased cut-off frequency of the filter compared to the ferrite bead. Thus, if I can trade off the worse damping for a lower cut-off frequency, the ferrite will do better. Your video really made me think about my past design choices and I will definitely do some more simulations in the future :)

[@Zachariah-Peterson]

If you look at some wirewound ferrite beads you can find ratings closer to 10 MHz, most standard SMD package beads target 100 MHz range as you mention. The inductor option can also be used at these higher frequencies if you use specifically high frequency inductors as these have parasitics controlled and guaranteed operation up to higher frequencies. Coilcraft has both types of parts.

[@schmitzbeats6102]

I think the parameters of the "model" of that bead is flawed. Isn't the parallel resistor a tad small and your parallel capacitance a tad high?! Looking at the LTSpice Model parameters for parts that have an inductance in the 150nH range (Würth 742792097) gives me 35 ohms and 800f as in FEMPTO farads. Nowhere near 1.5n... Also you are unlikely encounter a scenario where there is no additional capacitance like 100n decoupling near the load or somewhere towards the power source.

[@Zachariah-Peterson]

The point was not to look at a specific part number, but to put some example circuits side by side and look at the frequency responses so they can be analyzed without additional components such as decaps. If we wanted to simulate a specific LTSpice model for a specific part number, then we would put a part number. Contrived examples are just fine for learning concepts, now you can go create an LTSpice simulation with that Wurth part number and see how it works in your system. 😀

[@schmitzbeats6102]

@@Zachariah-Peterson I would pick realistic values to draw realistic conclusions. You can do better.

[@ChrisMayfield51291]

your analysis ignores the fact that the damped LC circuit has a considerable loss at lower frequencies and lower loads due to the series resistance. you're attempting to power the load with 5V, but only 1.67V is making it to the load (in the 5ohm case) due to most of the voltage being dropped across the 10 ohm resistor. I would say the ferrite bead is still a better alternative in the examples provided as the 5V DC actually makes it fully to the load and removing the series resistance on the LC circuit re-introduces the resonant peaking.

[@Zachariah-Peterson]

Well you're pointing out that at this bead resistance value it's similar to having high intentionally placed series resistance, kind of like the damping placement that Steve Sandler mentions in one of his posts (I showed it in another video). The difference is the band in which you notice that resistance, does it need to happen in DC or in AC near a specific frequency? I think this explains why once you dampen the bead and you replace it with an equivalent pi filter, there is little impact due to the fact you have intentionally placed damping resistance.

[@sjwpcbdesign]

So, the big question is this, why continue to manufacturer ferrite beads if they provide no benefit according to simulations. What actual use case does a ferrite bead serve?

[@Zachariah-Peterson]

Well in a way they do provide a benefit in a very specific situation, which I outlined in another video. That specific situation is when you need to limit high frequency noise in a very specific band from reaching a load when that load only operates at DC. Not a lot of loads will operate like this however.

[@sjwpcbdesign]

@@Zachariah-Peterson So basically they can be equivalent to a notch filter except they prepackage for a specific frequency based on bead parameters. Other than being a single SMD would it not be better just to create a notch with a few discrete components and be done with it. Just stop producing the beads and let people produce the notch in a traditional manner. This would then make clear what to do in DC systems.

[@Zachariah-Peterson]

​@@sjwpcbdesign I think the bead still lives on because it provides the same function in fewer components and when you look at peak attenuation you get greater bandwidth at the peak frequency for a given attenuation compared to a notch filter.

[@samratroy1487]

I was using a ferrite bead pi circuit filter for a converter but was getting some issues with conductive emissions, so increased the capacitance and replaced the ferrite bead with inductor but still getting noise around 500kHz. Anyone any suggestions?

[@filipnevezi]

Tighter layout...?

[@sanjaybatra6593]

@@filipnevezi means ? place decoupling capacitor or component close to each other ????

[@filipnevezi]

​@@sanjaybatra6593reduce loops especially in fast switching parts of the DCDC converter. Inductor as close to the controller chip and short fat traces. Same for the output caps, etc..

[@sanjaybatra6593]

conductive emission are high spike voltage ? how does it creating problem to your project ? where you are using pi filter and what is the value of inductor and capacitor ? ceramic capacitor ?

[@haldanecollins3735]

I guess this is a body blow to for example the Mil-DTL 38999 and ARINC 600 filtered connector markets.

[@dukeibzusa]

What about if it is recommended by the IC manufacturer

[@myetis1990]

Manufacturers recommend placing a resistive load on evaluation boards, but this load type may not match your specific requirements.

[@dukeibzusa]

I am talking about for example Analog Device's ADM3057E or microchip's SAMV7Q21 series

[@petersage5157]

tl;dw: If noise rejection band is at least an order of magnitude higher than signal frequency, single pole RC FTW.

[@87Spectr]

Thanks for video! 1. Inductor has parasitic capacity, capacitor has parasitic inductance. You draw only equivalent scheme of FB... Is this not important? 2. LC-filter will be effective in high frequency when his cut off freq = 100kHz, and DC 2..3A ? it means that inductor and caps are big sizes, big parasitic.

[@sanjaybatra6593]

what will be the values of LC, if suppose I want to filter an SPMS 5V DC , which is given to microcontroller and other component like RTC, LCD etc?

[@87Spectr]

@@sanjaybatra6593 i think you need to read specification for your SPMS... Understanding current consumption your 5V_rail... You can use big L - power will be smooth, but only load constant...

[@sanjaybatra6593]

@@87Spectr SMPS specifications are 5V/3A. Approximate current drawn is 1A to 1.25A . I am using atmega128A + RTC+LCD. so load almost is constant. Roughly what should be value of L C1 C2 for pi-filter.

[@АлексейБулдаков-п7з]

It seems to me that the bead replacement scheme is not correct. On your circuit, the impedance is no more than 10 ohms. If you look at the data on the impedance graphs of the data sheets of beads, there will be large values. What about this?

[@Zachariah-Peterson]

There are beads with 10 Ohms resistance or 10 MHz frequency. Sometimes the values are much larger, it all depends on the component. What matters for DC range is the ratio of the resistance to the load impedance because at the lower end of the load impedance range the bead (or resistor in an RC circuit) will determine the current limiting.

[@АлексейБулдаков-п7з]

@@Zachariah-Peterson I use LTSPICE for work. The built-in models differ from the ones you use in this video. For example mfg="Wurth Elektronik" pn="74269242161 WE-TKSPB 0402" SYMATTR Value 433n SYMATTR SpiceLine Apk=1 Rser=0.12 Rpar=309 Cpar=460f (I hope the essence of the parameters is clear) It may be necessary to check your example in the AD simulator and compare the practically measured suppression values. The adequacy of the model is very important for decision-making.

[@АлексейБулдаков-п7з]

@@Zachariah-Peterson LTSPICE ferrite bead models are very different from the ones you use. For example: SYMATTR Value 433n SYMATTR SpiceLine Ipk=1 Rser=0.12 Rpar=309 Cpar=460f mfg="Wurth Elektronik" pn="74269242161 WE-TKSPB 0402" I hope that the meaning of the parameters is clear. Could you compare the models in your videos?

[@danstiurca7963]

This dude is on a mission to spread ferrite bead misinformation. Build a real circuit and prove your point. You forgot the parallel capacitance of the inductor, which allows high frequency switching noise to shoot right through your LC filter. This is what the ferrite bead is for.

[@ats89117]

Ignorance is bliss. Ferrites are useful for adding high frequency RESISTANCE and they are appropriately used in the frequency range above where their inductive reactance has fallen well below their resistance. Perhaps Zach should listen to read some of the material that was put out by Jim Williams, Henry Ott and many others decades ago...

[@joaopaulocoelho5401]

Apparently courtesy and good manners should be a blessing to others.

[@Zachariah-Peterson]

@@joaopaulocoelho5401 I think this guy just pulled a tl;dw

[@DerIchBinDa]

Hello pot calling the cattle black. 🤦

[@soheil_az]

The 1.58nF C3 that you put in the ferrite model is a bit exaggerated to me. And it’s the reason you get so bad HF roll off from your ferrite model. I’ve had better HF roll off experience in real world than what you see in your simulation. Also the fact that ferrite resistance increases with frequency is not considered in your model. Overall I don’t think you are using a good enough model for a ferrite. (Maybe its the reason you don’t like them 😉)