Some years back, in one of my more inspired moments, I bought a lifetime supply of 1.5 nF Soviet feedthrough capacitors from eBay. I have about
1,500 of them left, which should be good for building protos till about
2175 AD.
Feedthroughs are great because they let you get supply and control wiring in and out of RF shields without trashing the shield effectiveness. They solder into a hole in the can, so whatever RF gets onto the wire, gets shunted to ground before leaving the shield. Magic.
There are two problems, though: (i) they require hand wiring, and (ii) they now cost $15 each!
SMT devices that claim to be feedthrough caps exist, but on their face they don't solve the wire pickup/re-radiation problem.
I hate those RF boxes with feedthrus. You have to solder wires to them, on odd surfaces of the boxes. Soldering to the can and then jumpering to the board is klutzy too.
A connector on the PCB, with some filtering parts, should usually be enough. And there are filtered D-subs too.
An RJ45, with or without the magnetics, could be a well shielded power or slow-signal connector too. PoE version maybe. They shield well to a cutout in a box without being captive. Cheap.
I got a bunch of nickel-plated steel tins (the size and shape of the smaller Altoids package) off AliExpress. The lids slide on and off rather than having hinges, so the contact is better, and a few drops of solder improves matters further. With LC filters inside the box and feedthrough caps to the outside, you'd never know what was in there--my patent-pending shielded coax loop doesn't pick up anything.
If you make a little PCB with a ground plane on one side, and put a plated-through hole in the middle for a signal wire to pass through, not connected to the ground plane, and put surface mount capacitors radially between the through-hole and the ground plane, that would be a pretty good approximation to the feedthrough capacitors that you bought.
I used 4 capacitors in a + arrangement, you could use more if your pick-and-place permits. You'd have to solder the groundplane of that little PCB to the tinplate of the box around the hole you mount it on, either by sweating the tinplate to the groundplane itself from the other side of the tinplate, or with castellated pads on the edges of the PCB, or by replacing one entire face of the tinplate box by the PCB with a groundplane oriented outwards. I've made and used these, and they were good enough for what I needed. It would be interesting to measure some with a VNA.
Obviously if your product's main PCB doesn't need parts on both sides, you can integrate the feedthrough-replacement-PCBs with the main PCB, reducing assembly costs.
If you have a lot of layers, you could also add additional capacitance inside the feedthrough PCB, by connecting the signal pad to a circle or other shape on an inner layer adjacent to groundplane layers. Putting multiple capacitors in parallel is not always an improvement of course, if resonances occur, but with some testing it may be possible to improve over whatever parasitic inductance the surface mount caps exhibit.
Not a bad idea, thanks. If we can keep the hole in the tinplate small enough, it might be OK just with regular mounting screws--the featureless bottom-side ground should look after the electrostatic shielding, leaving the box to do the last bit of magnetic shielding.
The screw machine folk still make, I hope, turret terminals which press into a PTFE collar, which would work for a feedthrough. Similar to these
formatting link
except the PTFE collar expands when the terminal is pressed in, so it locks into the hole.
There's no current supplier, though, for the cylindrical ceramic capacitors that were the core of those old feedthroughs, and you can't turn those out on a screw machine.
The SMT equivalents I have seen are (e.g.) Murata's NFL series of 3- terminal caps. The idea is that the centre pin straddles a trace on your board, with multiple vias to the ground plane, and the shield can is soldered to this trace, with a mouse-bite to go over the body of the SMT part.
It's never going to be as effective as a traditional leaded feed-through, as you lose the 360 shield connection and the RF currents in the shield are not symmetrically distributed. I've never seen this quantified, though.
The part we used was NFL21SP106X1C3D. But the design (long since retired) was from 20 years ago when we had no RF test equipment and not much understanding of EMI issues. It would be interesting to go back and run some proper tests on them.
Thanks. That's OK for electrostatic shielding, but won't fix the magnetic problem because there's no simple way to get the steel to be continuous. I guess one could stick little PCB tabs through the shield as long as it was on only one side of the board.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.