0
\$\begingroup\$

Can a feedthrough capacitor be considered (and used) as a capacitor with a very low equivalent series inductance?

\$\endgroup\$
  • \$\begingroup\$ Not that I'd particularly know of, no. How do you come to this suspicion? They are often used in RF filter applications, but AFAIK they're not optimized for especially good ESR \$\endgroup\$ – Marcus Müller Apr 8 '17 at 22:15
  • \$\begingroup\$ I've not writen "equivalent series resistance", I've written "equivalent series inductance". \$\endgroup\$ – MikeTeX Apr 8 '17 at 22:22
  • \$\begingroup\$ oh, sorry, yeah, just saw that. I think the following applies: there's some feedthrough caps that are sold as especially low ESL. So I guess the others aren't. \$\endgroup\$ – Marcus Müller Apr 8 '17 at 22:23
  • \$\begingroup\$ For example: product.tdk.com/info/en/catalog/datasheets/… has a low-ESL subtype (as opposed to the STD subtype) \$\endgroup\$ – Marcus Müller Apr 8 '17 at 22:24
  • \$\begingroup\$ This makes sense. \$\endgroup\$ – MikeTeX Apr 8 '17 at 22:26
1
\$\begingroup\$

No, they can't, but not for the reason you think. You can get simple feedthrough bypass capacitors, as well as the feedthrough filters that Tom Carpenter talks about in his answer.

If you wanted to take a line from one box to another, and isolate the RF by using a shunt bypass capacitor, you would need a capacitor of very low series inductance. In fact, for most applications, to do it in a single stage would require an unphysically small series inductance. If you crank up SPICE and see what you need, you might find you need well sub-nH, which you just cannot achieve with a 2 terminal shunt capacitor. 1mm of wire is, give or take, about 1nH.

But a feedthrough cap is not a 2 terminal device, it has three terminals. The problem with a 2 terminal device is the incoming RF generates a voltage across the stray inductance of both connections. This voltage is then output to the next box.

With a 3 terminal feedthrough, the outgoing voltage is picked off from the capacitor in series with only the ground connection. The output voltage therefore includes the voltage across the ground lead stray inductance but not the input lead stray inductance. As the ground lead is very wide (several mm) and very short (<1mm) in a cyclindrical feedthrough, this inductance is well sub-nH.

Although it works well in isolating RF when used as a 3 terminal device, the fact it has only one low inductance lead means it cannot be used as a low inductance 2 terminal capacitor.

The improvement over a 2 terminal shunt capacitor is much the same as you get when measuring a low resistance with a 4 terminal Kelvin connection, you do not need leads of extremely low resistance, as the voltage is picked off on the resistor itself.

You can also buy 'Filtercon' filters which look very much like feedthrough capcitors, where the input and output lead inductances are enhanced by adding a ferrite bead to each, to turn the simple shunt capacitor into a 3rd order low pass filter. But these are more expensive than a feedthrough capacitor.

In a related part, have a look at the BAR81 from Infineon. This is a PIN diode designed for shunt operation. One lead is made very low inductance, by use of the leadframe two pins in parallel, this goes to ground. The other lead is a hop-on, hop-off, connection, the same way that a feedthrough capacitor is.

|improve this answer|||||
\$\endgroup\$
  • \$\begingroup\$ +1 for explanations and links to Filtercom filters and BAR81 Infineon. I don't grasp in what the answer of Tom Carpenter differ much from yours: he did not say that there are no feedthrough capacitors and only filters exist, but only that they are essentially equivalent to the low pass filter in his schematics (this is exactly what you say no?). Anyway, your additional explanations about stray inductances etc. were much appreciated. \$\endgroup\$ – MikeTeX Apr 9 '17 at 18:53
  • \$\begingroup\$ No, his first two sentences stress the series inductance. A feedthrough cap has minimal series inductance. The essential feature is that it's a 3 terminal capacitor rather than 2 terminal capacitor, with the ground terminal very low inductance at the expense of the top terminal being a higher inductance. \$\endgroup\$ – Neil_UK Apr 9 '17 at 19:52
3
\$\begingroup\$

Simply, no. Feed-through caps intentionally have a high inductance on the signal path to allow them to work as filters. By running the current through a thin and quite inductive trace within the capacitor you end up with an L-C filter which is makes them so good at filtering and why you can replace ferrite beads with them.

Essentially a feed-through cap resembles the left hand diagram below, as compared to a regular parallel capacitor on the right.

Feed-through Cap Model Image Source

You can get low ESL feed-through caps, however this is referring to the parasitic inductance of the ground path (the circled inductor in the diagram above). You want to keep the parasitic inductance of the capacitor part of the circuit as low as possible to ensure good filtering.

For low ESL in general you want to look at capacitors specifically designed for this. Once such option is those in reverse SMD packages - 0306, 0508 etc. For these sorts of packages, the inductance is reduced by making the connection to the plates wider.

|improve this answer|||||
\$\endgroup\$
  • \$\begingroup\$ Exactly the converse to what I though. They say these capacitor is used to "bypass" RF signals, they not say "filter" these signals; this is simply misleading. \$\endgroup\$ – MikeTeX Apr 8 '17 at 22:27
  • 1
    \$\begingroup\$ @MikeTeX they bypass higher frequencies to ground, thus filtering them from the signal. Good for power lines, etc. \$\endgroup\$ – Tom Carpenter Apr 8 '17 at 22:29
  • \$\begingroup\$ It turns out that this is simply the usual electrical model for capacitors that does not apply for feedthrough capacitors. \$\endgroup\$ – MikeTeX Apr 8 '17 at 22:44
  • \$\begingroup\$ SMD feedthrough caps used as decoupling have ESL similar to MLCC of the same size... But you can also get CLC filters (like your left schematic which appropriately says "filter") with a feedthrough cap and tiny ferrite beads, all integrated inside a 0805 or smaller package these days. Example: mouser.com/ds/2/281/ENFD1032-519500.pdf \$\endgroup\$ – peufeu Apr 9 '17 at 10:15
  • \$\begingroup\$ @Tom. There seems to be inexact things in your answer (according to Neil_UK, a feedthrough capacitor may well have an as small series inductance as possible. See his answer and comments below). Could you edit your answer in order for the accepted answer to be entirely correct? (I dislike removing points after I gave them, and your answer is nicely illustrated). \$\endgroup\$ – MikeTeX Apr 10 '17 at 6:18
2
\$\begingroup\$

I tested this using a network analyzer some time ago, using Murata NFM21PS106 SMD feedthrough caps.

The gist of it is that, when you consider ESL from one terminal only, they behave more or less like a SMD capacitor of equivalent size. Most of the ESL is in the connection from your power plane to the cap anyway, so the type of cap matters little. Reverse geometry caps have lower ESL, but that's half due to the fact you can stick more via's on them.

X2Y caps, though, have very low ESL, because they form a clever low-inductance system with the vias. You cannot neglect via and mounting inductance, it is essential.

As filters, though, feedthrough caps are ridiculously good.

|improve this answer|||||
\$\endgroup\$
1
\$\begingroup\$

Here is a bulkhead-mounting filter from Tusconix, offering 70dB atten at GHz enter image description here

|improve this answer|||||
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.