Three-Terminal Capacitors vs Two-Terminal Capacitors: The Ultimate Comparison

In this video, we conduct a simple yet effective test to fairly compare a three-terminal capacitor with a two-terminal capacitor. Using a Vector Network Analyzer (VNA), we analyze the S21 results to highlight their performance differences.
We believe this is the most comprehensive and possibly the only direct comparison you’ll find on KZbin!
Enjoy the video and let us know your thoughts in the comments!

[@MachOneDesignEMC]

Join our community discussion page, let us dive deep and benefit from this discussion!

[@WR3slo]

Comparison of this 3-terminal 1206 to two 0603 or 0805 would be interesting too.

[@ValuedTeamMember]

3 terminal caps. What will they think of next? You've really scratched my Geek itch ¯\_(ツ)_/¯ Thank you.

[@ovalwingnut]

[@MachOneDesignEMC]

Next we will do the so-called X2Y caps, stay tuned. We also have feedthrough caps, and more!

[@ovalwingnut]

@@MachOneDesignEMC

[@carolmartin7042]

3-contact caps! Interesting. Chief

[@k.c.sunshine1934]

I like those EM shields in the background.

[@MachOneDesignEMC]

They are from Tekbox, very good product at affordable price. Fun fact, I think the shielding material has either silver or some expensive material there.

[@RemiDavidKoster]

I don't believe this says a whole lot about how it would perform in a real life circuit, where the limiting factor would probably be trace, via and plane inductance of the circuitry connecting them, which is now simply calibrated away. This mainly measures the ESL already specified in the datasheet. I wonder how much it would really matter in an actual circuit, and like @WR3slo says, if two parallel 0603's or 0402's placed close together wouldn't simply achieve the same? Also, why the 10mm distance between the two normal capacitors? Wouldn't the extra trace inductance introduce an LC resonance between them?

[@MachOneDesignEMC]

I agree that this cannot be a real life scenario, and yes, some other parasitic inductance was simply calibrated away. This really shows the low ESL, which is already stated in the datasheet. 0603 0402 would definitely perform better, especially if one place them close to the circuit. As for the distance between the two, it was demonstrated and published by researchers in the university, I also did a video just on this specific subject. The point in the papers they made was that this essentially avoid mutual coupling between the two caps and form a C-L-C, but in reality, how many circuits have a luxury of leaving big space? Definitely a lot more to explore on this subject, Remi! Thanks for the valuable feedback!

[@RemiDavidKoster]

@@MachOneDesignEMC Thanks for the video and the in depth response! Also the other post you made. Really looking forward to the other videos you mention. I see how spacing the capacitors like that can improve filtering, and I appreciate the reasoning behind it. However, in this case, I believe the typical application for these capacitors will be decoupling ICs, where the goal is minimizing PDN impedance (so there won't be any voltage ripple in the first place). Adding additional traces there would simply increase the impedance again, right? Curious to learn about the mutual coupling you mention, I've never come across that before.

[@juliatruchsess1019]

I suggest a separate discussion of X2Y™ capacitors - while they look similar to NFM-type 3-terminal caps and are often thought to be the same, they are in fact quite different, in some ways superior, and have more applications, e.g. as replacements for CM chokes.

[@MachOneDesignEMC]

Hi Julia, thanks for the great suggestions, we have this planned, and I am also intrigued to see the performance.

[@gabrielnilo6101]

would love to see the comparison in a complete product... for example, nowadays we have CPUs that achive 4GHz+, would I see any difference? Maybe the voltage spikes are minimized ?

[@MachOneDesignEMC]

In fact, iPhone has many three terminal caps. I am sure many other high speed circuits use these caps .

[@franzliszt3195]

I guess I now understand what some designs use 4 small capacitors instead of one.

[@MachOneDesignEMC]

I think 4 is a lot, but you are right, 2 or 3 beat one for sure.

[@franzliszt3195]

Could you put the two capacitors closer together if they are an right angles to each other?

[@franzliszt3195]

I mean, the further they are from each other, the more loop inductance introduced?

[@MachOneDesignEMC]

Great suggestion. We have one episode scheduled for two weeks time looking at this, don't miss out.

[@tomasbergh]

Good again. Why didnt you compare with 2pcs of 4n7 0603 in parallel? Approx same physical size as one 1206 10n.

[@MachOneDesignEMC]

Hi Tomas, I should have used 2 * 4.7nF caps, but I think as a fair comparison, I should still use 1206, as I am pretty sure the 0603 one will have lower ESL.

[@tomasbergh]

@@MachOneDesignEMC Yes, ofcourse you should compare 1206 to 1206 but then also show 2*0603 as it is approximately the same total size, and thereby also some learning? 😀 From one nerd to another nerd: Thanks for nice videos!

[@analoghardwaretops3976]

Paralleling multiple caps of same value will always show the min .value @ around a narrow band of freq. better low min. value freq. spread is achieved when a mix of different value caps &/or types are paralled...... then too it's really different behaviour in an actual layout as compared to the one in a test jig.😅

[@MachOneDesignEMC]

I agree. I did another video, which I changed the capacitor layout and measured the conducted emissions, I will share the findings soon. Thanks!

[@franzliszt3195]

Toy Huben?

[@MachOneDesignEMC]

Todd Hubing

[@curtisnewton895]

"3 terminals" ??? whut ?

[@DeadQuirky]

I always cringe when I see engineers handling RF terminals like BNC connectors. I was taught never to touch them because the oils from your fingers can attract and trap dust and other contaminants, potentially increasing resistance and degrading signal integrity. Is this still considered best practice? I also make a habit of cleaning terminals with isopropyl alcohol before taking any serious measurements.

[@MachOneDesignEMC]

I think this is true, especially for very high frequency RF stuff.