Common mode chokes: There is more than meets the eye

Рет қаралды 6,411

Sam Ben-Yaakov

Күн бұрын

#EMI #switchmpde #noise #PWM #coupledinductor

Пікірлер: 47

@SkyhawkSteve Ай бұрын

I spent a fair bit of time working on EMC design issues, and eventually learned to read the datasheets for the ferrites. Like all components, the real parts do have limits due to the parasitics, etc. I never really had any luck with simulations, since it was nearly impossible to characterize the circuit board and all other components. Still, it was invaluable to know at which frequencies the ferrite was effective and which frequencies that the impedance dropped off. It was also useful to learn that there is quite a bit of tolerance on the rated peak impedance, and it was worthwhile to characterize the actual ferrites that you were testing with. Sometimes you want to test with the lowest impedance parts, just to be sure that the production items would pass the EMC tests too. Thanks for the video!

@sambenyaakov Ай бұрын

Thanks for sharing

@MaicolLXVII 3 ай бұрын

Hi Sam, I found this video very interesting. This is Michele, I'm the manager of a group that releases automotive actuators and sensors. It's incredible how complex things become when you have to integrate multiple electric/electronic parts in a system. Even if everything is, apparently, carry over from past projects.. Understanding EMC is very important in this perspective. Thanks for this video, very clear and informative at the same time. Michele.

@sambenyaakov 3 ай бұрын

Hi Michele, thank for your note.

@SergiuCosminViorel 4 ай бұрын

splitting the wires, i did not do in this regard, so not even suspecting it. i won't do modelling, ever, but i learned here that even for such apparent simple circuit, measurements and some empirical approach are needed. thank you, for bringing this!

@sambenyaakov 4 ай бұрын

Thanks for comment

@sambenyaakov 3 ай бұрын

"splitting the wires" this is to explain that when measuring the impedance form input to output impedance it is like one wire. Nobody is splitting wires. What a relief that you do not do this.

@Cydget 3 ай бұрын

Good stuff. This topic wasn't taught to me in school, and even people at work were not able to articulate the concept that well to this level of depth. Most of the testing we do is kind of like the shotgun approach where we just try a bunch of things/values and see if it gives us a lower score. I get that engineering is a lot of that due to how complex these systems get, but I'm always trying to get a better understanding even if the shotgun method gives faster results. If you do a part two, maybe mention some practical issues when measuring such as noise floor( how to reduce it)/ wire length impact(why does the standard have a minimum) / grounding plate locations/orientations/ type of wire+connectors (romex/thhn/wago)

@sambenyaakov 3 ай бұрын

Thanks fo feedback.

@simoncliffy Ай бұрын

Excellent little talk and especially in progressing onto real & imaginary permiability. The imaginary component (as I recall from undergrad applied physics) is due to the work being done in the ferrite lattice to move the magnetic moments around and continually attempting to balance the energy needed to create a (newer smaller) domain wall, and trying to avoid that influence of a larger but more distant domain. As the the frequency goes up as we're moving the Bohr magnetons around more rapidly (with spins on the electrons in the outer 3d & 4s shells of the Fe & Mn atoms) by attempting to redefine domain boundaries around, there'll reach a stage when our fast megahertz frequency fields don't have enough energy to move the domains or redefine their size for any length of time and we'll just end up uselessly nudging around electrons on outer orbitals in the ferrite lattice - completely lossy. Re(mu) ends up being a barrier where it can no longer store energy from our B field, so it stops being inductive. It's interesting that exactly the same kind of thing goes on in dielectric when attempting to store high frequency energy between the displacement field and electric field in a capacitor. In this case, trying to lob energy in and out of the ions in crystals of aluminium oxide in an electrolytic stops working when the energy is insufficient to create a fast and big changes in charge more than at a few hundred kHz, so we have to look towards using lattice ions to store displacement field energy in something like a MLCC when Im[epsilon] has it's peak. This D-field/E-field interaction in dielectric has an identical form to B-field/H-field in magnetics, has a hysteresis curve just as for magnetics and is responsible for the bias voltage effects on reducing capacitance in type II dielectrics, whose annealing is probably the result of the 3%/decade temperature and bias losses you've already done an excellent little talk on. It's lovely when two systems demonstrate similar effects due to similar causes in two entirely different areas of electronics!

@simoncliffy Ай бұрын

End of para 1, I meant Im[mu], mu'' nor real mu!

@sambenyaakov Ай бұрын

Thanks for sharing . Interesting. This similarity was already recognized by Steinmetz😊

@anatolimordakhay3899 4 ай бұрын

Always a pleasure to watch your videos, interesting and very educating. Glad I got the opportunity to attend your classes in person at BGU as a student some years ago, and that I still get the chance to learn from your videos 👏

@sambenyaakov 4 ай бұрын

👍😊🙏Nice to read from you. Where are you at?

@freedomisfood6966 Ай бұрын

Thanks for your share knowledge

@sambenyaakov Ай бұрын

👍🙏

@Duracellmumus 3 ай бұрын

Thank u for show that cruves, i have a shot now for why this is not working well with high speed dc load. In my wiev this type common mode chokes are do better job before the diode bridge at AC side.

@sambenyaakov 3 ай бұрын

I am showing a DC (battery) source

@fedep2753 4 ай бұрын

very interesting, as always. thank you.

@sambenyaakov 4 ай бұрын

Thanks

@RyujiSuzuki Күн бұрын

You could go one step further to describe that the leakage inductance is related to μ''. I felt strange that the word "oscillation" was used to describe the impedance going up and down due to self-resonance of the windings (|Z| vs f plot). The inductor used as an example was chosen because of the availability of the SPICE model, but it appears inadequate to reduce the ripple harmonics by 20+ dB (though the additional capacitors should help reduce the diff mode)... so how would one build a better bisection inductor pair?

@zorz67 4 ай бұрын

Dear Professor Ben-Yaakov, thank you so much for your video. Very interesting indeed. The discrepancy in the simulations with the Wurth model is 6dB in attenuation, which, I believe, might be explained with the 2x factor that is probably not accounted for in the 50 Ohm to 50 Ohm impedance matching. You have 6dB attenuation at DC in sims and 0dB in measurement. Also, it seems to me the impedance model is not correctly taking into account the inductor DCR. The measured impedance is one order of magnitude higher at DC.

@sambenyaakov 4 ай бұрын

Thanks for input.

@esijal 4 ай бұрын

Excellent presentation professor 🙏

@sambenyaakov 4 ай бұрын

Many thanks!

@purerhodium 4 ай бұрын

For the resistive component of the impedance, it works out if you assume that each winding has a resistance of 168 mΩ. In parallel (common mode), the resistance of the choke is 84 mΩ, and in series (differential mode), it works out to 336 mΩ. I'm not sure if this is how it works in reality, but presumably it is how the LTspice model is programmed.

@sambenyaakov 4 ай бұрын

There is though a discrepancy between the data sheet and the LTspice model, isn't there?

@klauskragelund8883 4 ай бұрын

@@sambenyaakov Datasheet is max value. Simulation model uses characterized data, so typical.

@sambenyaakov 4 ай бұрын

@@klauskragelund8883 The difference is not just in the magnitude but in shape, especially at high frequency, And, since when the datasheets show max values in plots?

@SMV1972 4 ай бұрын

Спасибо Вам! Прекрасный урок!

@sambenyaakov 4 ай бұрын

Thanks

@tonyh6309 4 ай бұрын

Wouldn't the impedance decline be better explained by inductor having an SRF at approx 2MHz due to parasitic parallel capacitance? I calculate 8.8pF (arising from inter-winding plus winding-core-winding capacitances) which doesn't seem unreasonable given the size of the windings and core. If I understood you, you showed the permeability plots for 3F3 because it also, helpfully included the plot of imaginary permeability; the actual material used for this part could have a somewhat higher frequency characteristic before its permeability starts to decline. That aside I found it helpful and well presented explanation of the issues. The imaginary permeability component was definitely new to me.

@sambenyaakov 4 ай бұрын

Capacitance can indeed emulate the drop and this is how simulation models are built. But the physics is that the permeability decreases and this is the real reason for the drop in response at high frequency. On top of it there might be the resonances due to parasitic capacitances.

@tonyh6309 4 ай бұрын

@@sambenyaakov Sorry, but I'm pretty sure that your explanation for this *specific* case is wrong. The Wurth part you chose has a *nanocrystalline* core which has significantly different complex permeability characteristics to conventional ferrites. In particular it has a much broader frequency range, 200kHz to 200MHz according to a Wurth presentation. Unfortunately I can't find the permeability data on Wurth's site but I found a good paper at pubmed (NIH). I can't post a link here but searching for 29360754 will easily find it. Figures 3 and 7 are particulary interesting - the first showing that u'' is almost as high as u' from low frequencies unlike ferrites which have low u'' losses at low frequencies. (This might, in part, explain the higher than expected resistance at lower frequencies). Fig 7 shows that Z continues to rise steadily well past the frequency where XL plummets due to u' falling. So the impedance curve for the Wurth CM choke you showed is almost certainly explained by the parasitic parallel capacitor dominating above 2MHz, declining at 20dB/decade. Another data point is this presentation by Wurth themselves on their CM chokes: see from 32 : 57 kzbin.info/www/bejne/r3ymaYymZapsfJo At 33 : 20 he states "..where the impedance drops due to the capacitance effect". Ok he could have got it wrong or was avoiding the complexities of permeability variance with frequency but I doubt it. Note that the part he is showing is an MnZn ferrite, not nanocrystalline. Finally why isn't the leakage inductance impedance similarly affected by the collapse in u' after 2MHz? It continues to rise linearly, well past 2MHz. This doesn't detract from the points you were making about the properties of ferrites which encouraged me to look into far more than I expected.

@tonyh6309 4 ай бұрын

@sambenyaakov Sorry, but I'm pretty sure that your explanation for this *specific* case is wrong. The Wurth part you chose has a *nanocrystalline* core which has significantly different complex permeability characteristics to conventional ferrites. In particular it has a much broader frequency range, 200kHz to 200MHz according to a Wurth presentation. Unfortunately I can't find the permeability data on Wurth's site but I found a good paper at pubmed (NIH). I can't post a link here but searching for 29360754 will easily find it. Figures 3 and 7 are particulary interesting - the first showing that u'' is almost as high as u' from low frequencies unlike ferrites which have low u'' losses at low frequencies. (This might, in part, explain the higher than expected resistance at lower frequencies). Fig 7 shows that Z continues to rise steadily well past the frequency where XL plummets due to u' falling. So the impedance curve for the Wurth CM choke you showed is almost certainly explained by the parasitic parallel capacitor dominating above 2MHz, declining at 20dB/decade.

@tonyh6309 4 ай бұрын

@@sambenyaakov YT suppressed my comment so I split it into two parts: Another data point is this presentation by Wurth themselves on their CM chokes: see from 32:57 kzbin.info/www/bejne/r3ymaYymZapsfJo At 33:20 he states "..where the impedance drops due to the capacitance effect". Ok he could have got it wrong or was avoiding the complexities of permeability variance with frequency but I doubt it. Finally why isn't the leakage inductance impedance similarly affected by the collapse in u' after 2MHz? It continues to rise linearly, well past 2MHz. This doesn't detract from the points you were making about the properties of ferrites which encouraged me to look into far more than I expected.

@tonyh6309 4 ай бұрын

@@sambenyaakov YT make it incredibly difficult to comment - this is my third attempt splitting up my reply: @sambenyaakov Sorry, but I'm pretty sure that your explanation for this *specific* case is wrong. The Wurth part you chose has a *nanocrystalline* core which has significantly different complex permeability characteristics to conventional ferrites. In particular it has a much broader frequency range, 200kHz to 200MHz according to a Wurth presentation. Unfortunately I can't find the permeability data on Wurth's site but I found a good paper at pubmed (NIH). I can't post a link here but searching for 29360754 will easily find it. Figures 3 and 7 are particulary interesting - the first showing that u'' is almost as high as u' from low frequencies unlike ferrites which have low u'' losses at low frequencies. (This might, in part, explain the higher than expected resistance at lower frequencies).