EMRFD Message Archive 15025

Message Date From Subject
15025 2018-08-25 11:51:04 peter_dl8ov RF Decoupling of Connections
If I want to bring in a connection through an RF tight enclosure then the normal solution is a feedthrough capacitor, easy. So, how do I handle this problem if there are thirty wires, or maybe fifty? Fitting dozens of feedthrough capacitors is not an option as the case will end up like Swiss cheese, leading to mechanical issues and RF leaks.

So, multiple decoupled connections, any suggestions?

Peter DL8OV

15026 2018-08-25 15:32:30 Bill Carver Re: RF Decoupling of Connections
Peter, I drill holes in the RF tight enclosure for a row (or two rows) of 0.025" square pins that are in strips with 0.1" spacing. Then I drill an identical row of holes in a strip of double sided copperclad PCB material, with some holes to ground the copperclad to the enclosure wall. Then I cut through the copper surface to make insulated squares, 0.1" x 0.1" around each pin. I use a little routing tool just to remove the copper, makes an insulating strip about .01-.02" wide. Then I solder SMD capacitors from each square to the screw-grounded copper around the pins. As has been posted here, even fairly large capacitance SMD caps are good bypasses over many decades of frequency. I have some old partial-reels of 0.33uF 50V 1206 capacitors I use where frequency response is not an issue, and of course you can use different value capacitors at each post and trade off ability to pass signals versus stripping off RF.

To make sure grounding is as low inductance as possible I wrap the edge of the PCB with some copper foil and for long strips use multiple mounting/grounding screws. I think the performance limitation will be resonance of the grounded strip, and the shared (common) inductance of the strip. That shared inductance will produce some coupling between all the circuits through the wall. By putting the pins in the center of the strip, and grounding the strip on both sides of the row of pins, with identical bypass capacitors from each side of the pin to adjacent foil, this limitation should be minimized. As you know, the capacitors should be the same value: if one is big, the other small, then the big one will become inductive and parallel resonant with the smaller one at some frequency. Depending on the "Q" of that resonance you could have an impedance peak. 

Having a "jig" with holes on 0.1" centers (A scrap of Vectorboard) makes drilling holes on the pin centers a little easier, then you can "open the holes up" with a larger drill: the diameter of the holes in the enclosure of course have to be large enough so the pins don't short to the enclosure wall. The holes in the PCB material, however, are best left just large enough to pass the pins: I use a #63 drill.

If the pins are long on both sides then you can also slip -43 or -73 ferrite beads on inside before the internal wire connection to make an "L" filter.  Or, even beads on both sides for a "T" filter. The diameter of most beads is more than the spacing of the row of pins, but beads are made with smaller diameter. Fair Rite 2673000501, for example, have an OD of .076". Or in -43 material the P/N is 2643000501. Those part numbers are "preferred" and available here from Mouser. Of course the small diameter means the series impedance is also small. If not ever point does not warrant having a bead, only some, then you can use larger diameter, higher impedance, beads just by skipping adjacent pins.

It is possible to carry this to an extreme if the wall of the enclosure is thick. Those two bead part numbers have a length of .06" (just over 1.5mm), and they can be sandwiched into sheet metal between inside and outside capacitor boards. And even more extreme, a bead buried inside the wall of the enclosure, SMD capacitors on both inside and outside, with more beads slipped over pins on inside and outside! This sounds a little crazy, but I did this for a DDS LO.

The pins make convenient connection point with "SIP" (straight line) connectors, making a very convenient "disconnect" capability for the RF right enclosure.


Bill  W7AAZ

15029 2018-08-27 11:38:17 kb1gmx Re: RF Decoupling of Connections
 It really depends.  Once you bring wires in and out you break the plane and 
the shielding may be greatly reduced.

Without specifics often, finding ways to reduce the number of wire is a starting point.

Once you at that point assuming no RF on the wires or high speed data some form of connector
my usual us 9pin, 25 pin, or 40pin DB series.   Then shield the external cable or wire groups.

Internally a choke and a cap to ground can be used to maintain some level of shielding.
However data leads cannot often tolerate that.

in one case I solved it as it was a bunch of select leads for relays.  Rather than 1 wire for 
each relay (10 of them) I put a decoder and relay driver inside as it was all very low 
speed data.  Then I only needed 5 wires (4 data and 1 DC power) against 11.

So a lot of "depends" unfortunately.  Hopefully some suggestions.