3
\$\begingroup\$

I have a hardware device which includes Ethernet connectivity. I have a surface-mount Ethernet jack, with top entry (as opposed to right-angle). Part number is RIA Connect AJS03B8813, if that helps. No integrated magnetics, this is fine, I have a discrete part. All is well and good.

Except that I'm having trouble finding that part, so I need to find a replacement. However, it seems that no one makes a jack that has all of the following features:

  • top-entry (i.e., part sits flat on board and cable comes "down" into it, perpendicular to the board -- in other words, NOT right-angle),
  • surface-mount, and
  • shielded.

I can find shielded right-angles aplenty, and shielded top-mounts, but they're through-hole (not an option, the other side of the board is densely populated.) All I can find are unshielded parts, such as

So, my question is: are there any potential problems using this unshielded jack in a 10Mbps Ethernet application? I can't see the shielding on the jack mattering that much, since most clients don't use shielded cat5 anyway.

Any input would be appreciated. Thanks!

EDIT: thanks for all the replies. I'll just go ahead with the unshielded part. I know that most cables aren't shielded, so I figured it was probably pointless, just wanted to get some other opinions.

\$\endgroup\$
  • 1
    \$\begingroup\$ I've used an unshielded 20m cable for 10Mbit/s for years, where wires in addition where connected the wrong way (wires where not paired correctly so it was no good for 100Mbit/s, got it for free because of that). \$\endgroup\$ – starblue Jul 14 '11 at 20:21
  • \$\begingroup\$ @starblue, your SNR just died inside. Think of the BER man, the BER! \$\endgroup\$ – Kortuk Jul 14 '11 at 22:40
  • \$\begingroup\$ @andersop and everybody - I added a link to the part's datasheet. We're trying to make users aware of the importance of this, especially for less common parts, so that others don't have to go searching for it and that everybody is sure to be talking about the same thing. In this case it's just a mechanical drawing, but it's a good time to cultivate good habits. \$\endgroup\$ – stevenvh Jul 15 '11 at 5:16
  • \$\begingroup\$ @Kortuk As far as I could tell the BER was good enough, so it just shows that 10MBit/s is not that critical. And relax, I threw that cable away 7 years ago. \$\endgroup\$ – starblue Jul 15 '11 at 7:14
  • \$\begingroup\$ @Kortuk - I recently had to redesign a 100Mb/s ethernet board I designed a few years ago. It was only at this point that I discovered I had got my TX+ and TX- connected backwards, but it never seemed to be a problem. Always worked fine. \$\endgroup\$ – Rocketmagnet Apr 19 '12 at 0:06
6
\$\begingroup\$

Unless the application is in a very noisy environment shielding is not required, or really even recommended.

Shielding may be needed in very noisy environments such as some industrial settings but it comes with its own set of problems.

Special attention needs to be paid to ensure that the shield is properly grounded, preferably only on one end of the cable. If both ends are grounded then both systems need to be able to deal with the ground potential offset that will almost certainly exist. In more complicated network topologies this can be a serious issue.

In short, if you don't REALLY need shielding, don't use it.

\$\endgroup\$
  • 1
    \$\begingroup\$ ethernet is differential signaling, I think, which makes it even less susceptible then other technologies. \$\endgroup\$ – Kortuk Jul 14 '11 at 19:46
  • 1
    \$\begingroup\$ @Kortuk: Yes, ethernet uses transformer-coupled twisted pair differential signals. \$\endgroup\$ – Olin Lathrop Jul 14 '11 at 20:20
6
\$\begingroup\$

Shielding just the jack without the cable also being shielded is rather pointless. About the only purpose of a shielded ethernet jack would be if it mates up with the chassis in such a way to minimize the electrical hole thru the chassis for the ethernet cable to pass thru. If your connector shield is only connected to the board ground and not to anything else it touches off the board, then it's just a waste. I haven't used a shielded ethernet connector yet.

The much bigger issue is common mode radiation from the cable. This is where I've seen naive designs fail FCC testing. Some think this is what shielded connectors are for, but they are pretty useless for that.

Fortunately the solution is usually pretty easy. Take a good look at the ethernet transformer datasheet, and you will probably see extra windings in series with the outboard leads. These form a balun, or common mode choke. This is inductance in series with the common mode signals both wires of a pair pick up from your board, but little or no inductance in series with the differential mode signal (the actual ethernet data). You don't need much capacitance (20-50 pF) to ground on each line to attenutate the common mode high frequencies, well less than the ethernet capacitance spec or a relatively small length of cable. One thing to watch out for though is what that does to the isolation voltage. Ethernet is transformer coupled, in part to avoid ground loops. I forget the spec, but it's 100s of Volts at least. If you do put capacitors there, you either give up some of that isolation voltage, or the capacitors will need to be rated for high voltage and therefore large.

\$\endgroup\$
  • 1
    \$\begingroup\$ Baluns formed the favorite part of my Antenna theory class. Quite an interesting field and some very interesting designs were use when going to odd antenna designs. Love the answer. \$\endgroup\$ – Kortuk Jul 14 '11 at 22:39
2
\$\begingroup\$

Even for 1Gb it's not a necessity (for Ethernet itself), for unshielded 10Mbit I can bet 1 million $ :-)

The only slight worry is If you have sensitive millivolt-range unshielded signaling(like sensors reading or audio inputs) on the board, you may get slight interference.

\$\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.