3
\$\begingroup\$

So I am setting up an automated test environment for an electronic product and one of my needs is to simulate a user pulling the cable from the product. Since only 4 wires of the RJ45(T-568B) cable are used (https://www.iplocation.net/rj45-wiring), I am soldering these 4 wires (1,2,3,6) on a relay (S4-5V) and controlling them with a micro controller, and I have grounded all the other wires on both ends of the cable (the one coming from the router and the one that goes to the product being tested).

T-568B
Pin Color Pin Name 1 Orange Stripe Tx+
2 Orange Tx-
3 Green Stripe Rx+
4 Blue Not Used 5 Blue Stripe Not Used 6 Green Rx-
7 Brown Stripe Not Used 8 Brown Not Used

I have checked the connections on both ends of the cable with an multimeter (using the "noise on short circuit" function) and the wires 1,2,3 and 6 were connected as expected when the relay was closed and the other ones were grounded. So I assume my cabling is right, only problem is the device still can't connect to internet with the relay closed. Does anyone has an suggestion of why could that be?

Thanks in advance,

\$\endgroup\$
2
  • 2
    \$\begingroup\$ Welcome to the wonderful world of HF, where shape of conductors matters. \$\endgroup\$ Oct 1, 2015 at 15:58
  • \$\begingroup\$ You're changing the impedance of the twisted pair cables by adding the relay. \$\endgroup\$
    – Alexxx
    Oct 1, 2015 at 16:07

2 Answers 2

3
\$\begingroup\$

To expand on the comments, which are pretty much the answer:

The cable has a specific characteristic impedance. This impedance is created by the capacitance between wires and the inductance along them. It is in effect caused by the number of twists, the tightness of the twists, the insulator properties, etc.

Taking apart one pair for 10 to 15mm for the purpose of crimping it into a connector, which is then also designed to give the best possible conductor-to-conductor distance, is pretty much okay, no major changes there.

It is a very different thing to tear apart a wire and put an entire relay into it. You are changing the capacitances and inductances in the cable so much that the impedance changes at the point where you where you put the relay.

This change in characteristic impedance works like a mirror to part or all of the high frequency signal, where some or all of it gets reflected back to the sender. This confuses the sender (or even damages it, but modern network gear is pretty damn well protected, so it'll be unlikely), causing it to cancel the package it is sending, because it thinks it is seeing "collisions" with other data.

So every packet sent on either side comes back from the point of your relay and causes a collision detection and abort.

It is why so many HF designers are so very well paid. It's a very complicated job to make every PCB trace, every connector and every wire so that a 100MHz to 10GHz signal gets through okay. And your relay most certainly was not designed by a HF engineer.

\$\endgroup\$
3
  • \$\begingroup\$ Thanks @Asmyldof, do you know of any relay that could work on this situation? \$\endgroup\$
    – clds
    Oct 2, 2015 at 9:03
  • \$\begingroup\$ @Clds more than half of your problem is not inside the relay. But no. No traditional relays. Micro relays maybe. Thing is, nobody wants to do that with a relay, because generally it makes no sense to want to. \$\endgroup\$
    – Asmyldof
    Oct 2, 2015 at 9:58
  • \$\begingroup\$ @Clds you could try looking for solid-state analog switches designed for RF, and lay them out on the PCB with controlled-impedance traces. You may even find purpose-designed components that do something very similar to what you want. It would be better than the relays. Be mindful of what speed Ethernet you are using, too. Slower ones will let you get away with more. Honestly you could probably slap in relays and have it work, but you'd be pushing things, and for a test rig you don't want to introduce sources of instability in the first place, so try to do it as well as you can. \$\endgroup\$ Dec 24, 2016 at 18:50
1
\$\begingroup\$

There are a few ways to approach this:

  1. Make a mechanical device to actually physically disconnect and RJ45. Cut the tab off of the RJ45 plug, static mount an rj45 coupling and solenoid to a board, and use the solenoid to pull the rj45 out of the coupling. Might need some guides or rail so the rj45 goes back in readily. This replicates a connector being removed and the electrical make/break that physically occurs but is a little work to set up.
  2. Use an ethernet isolator (optocoupler based) to split the network. Drop power to the isolator using a standard relay, which will effectively break the connection. This is easy to set up but doesn't replicate the RJ45 physical contacts making/breaking.
  3. Same as #2, but use a switch instead of an isolator. A switch is adding something intelligent into the network, so beware. A switch is cheaper and more readily available than an isolator though. If you've got a managed or 'smart' unmanaged switch, you could also configure it to mirror data from ports 1/2 to a 3rd port and wireshark the loss/recovery events.
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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