I’m designing a small switching circuit for 1000Base-T Ethernet and have a concern regarding impedance. The board design consists of 3 shielded Ethernet jacks and 4 DPDT relays (here) made for high frequency. 1 Ethernet jack will be the input and the relays will alternate between the 2 outputs.

I’m having the board laid out with 100 ohms of impedance but the relays I’ve selected have an impedance of 50 Ohms. After contacting the manufacturer, I was told that I should match the impedance between the board and relays.

If matching impedance is critical, what options do I have? One option I’ve found is using pulse transformers with a 1:1.414 turn ratio before the relay to drop to 50 Ohms and after the relay bump back to 100 Ohms. Alternatively, I found (here) that the impedance of a trace in RF circuits can be changed by simply changing the width of the trace with an abrupt step. Is it really that simple and can I do the same without any issue?

The traces for each differential pair will be striplines and I believe the entire length of each trace will be less than 2 inches from jack to jack if it makes a difference.

  • 4
    \$\begingroup\$ Why are you switching Ethernet with relays in the first place? \$\endgroup\$ – Dave Tweed Mar 31 '15 at 20:44
  • 1
    \$\begingroup\$ I need to make something like an A/B switch. (not networking switch). Should I use something other than relays? \$\endgroup\$ – Dazzyman Mar 31 '15 at 22:03
  • \$\begingroup\$ I've seen this done for failover on nic cards that need high availability. Or for bypass like this en.m.wikipedia.org/wiki/Bypass_switch \$\endgroup\$ – Some Hardware Guy Apr 1 '15 at 0:00
  • \$\begingroup\$ I posted some things before but got scolded for putting questions in the answer :). First you say you have 100ohm traces but do you mean 100ohm differential which I would expect for gige? In that case your relays are probably 50ohm single ended so two per lane would match up nicely. \$\endgroup\$ – Some Hardware Guy Apr 1 '15 at 0:01
  • \$\begingroup\$ Next if you are really trying to go from 100 to 50 with just the pcb consider a taper instead of an abrupt width change. \$\endgroup\$ – Some Hardware Guy Apr 1 '15 at 0:02

I designed a two gigabit Ethernet ports with bypass board almost two years ago.

You have miss just one thing with impedance:

  • Ethernet lines are differential pairs with a differential impedance of 100 Ohms
  • Your relay is single line with characteristic impedance of 50 Ohms

But in your differential pair, each line will have its own characteristic impedance, lower than 100 Ohms but higher than 50 Ohms (for instance a quick computing in Saturn PCB give me 77 Ohms for line characteristic impedance of a 100 Ohms differential pair).

So yes you will have an impedance mismatch but so have you when you have a via, a pin, a connector, and also when you have no ground reference on your Ethernet pairs between transformer and connector.

A little impedance mismatch is not a big deal, this will distort your signal but really I don't think you will be able to find it.

I will be more worried by the maximum cable length for your bypass than by this impedance issue.

Here is an extract of the schematic of the board I designed: enter image description here

We used more general use relays (insertion loss = 0.33dB @ 900MHz), but cheaper.

| improve this answer | |
  • 1
    \$\begingroup\$ Hi ZeqL, Looks interesting. The circuit that you have posted looks for LAN Bypass function. Can I get more information about it?? \$\endgroup\$ – Oshi Mar 30 '19 at 16:27
  • 1
    \$\begingroup\$ Hi Oshi. This is effectively a bypass with relay. The board have 2 ethernet ports with MDI connected through relays. When power is off relays connect the two RJ45 avoiding the Ethernet link to be "cut". When power is on relays are activated and MDI are connected to transformer. \$\endgroup\$ – zeqL Apr 1 '19 at 8:24
  • \$\begingroup\$ Thank you ZeqL. One doubt, Did you used isolated power supply for relays? If yes, How was the isolation done? \$\endgroup\$ – Oshi Apr 3 '19 at 7:29
  • 1
    \$\begingroup\$ I don't understand. Relay command is galvanically isolated from signal. So normally you don't have to add a second galvanic isolation for the board. Relay used has a 1.5kV-2.5kV insulation. Relay commands power are quite noisy so it better to filter them with a ferrite. \$\endgroup\$ – zeqL Apr 3 '19 at 22:06

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.