Author Topic: routing ethernet signals between PHY and MAGJACK  (Read 823 times)

0 Members and 1 Guest are viewing this topic.

Offline phil from seattleTopic starter

  • Super Contributor
  • ***
  • Posts: 1063
  • Country: us
routing ethernet signals between PHY and MAGJACK
« on: May 19, 2020, 03:42:23 am »
One of the rules is to make sure the two differential signal traces have the same length. I'm using Eagle and it's hard to get them precisely the same length using meander. I can get very close - 0.002% for the transmit pair (37.722068044 mm and 37.722818122 mm) and 0.0004% for the receive pair (15.941376467 mm and 15.941444677 mm). I think receive is ok - I doubt my board vendor has that level of accuracy.  But for transmit, I wonder if this is ok?  I suspect it's fine but don't want to create a reliability problem.

Also, for Kicad users, it is easier/better than Eagle for routing differential pairs like this? I've been telling myself I should switch over but after 20 years, Eagle is pretty ingrained. I've not found anything I can't do with it.
« Last Edit: May 19, 2020, 03:45:39 am by phil from seattle »
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: routing ethernet signals between PHY and MAGJACK
« Reply #1 on: May 19, 2020, 06:23:38 am »
Mind this is PHY side, which is driving a CT winding.  The two traces act as stubs against each other, so their lengths should be kept as short as practical, and lengths matched.

What level Ethernet?  At 100BASE-T, a few nanoseconds of imbalance would be big deal.  That's almost nothing to worry about on a PCB.  1000BASE-T needs to be tighter, but even simply within a few mm is more than good enough!

Signal quality really only matters in the extreme, near maximum length, with poorly matched or corroded cables and connectors, poorly made splices, etc.  For casual runs say in the 10s of m long, the built in signal quality enhancements of 1000BASE-T more than make up for any of that.

Ethernet can literally run on wet string (specifically, brine soaked), albeit at 10BASE-T and only to a fraction of its nominal range.  You can't do much wrong to get any connection.  How good of a connection, and over what distance, is what actually matters.

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline Someone

  • Super Contributor
  • ***
  • Posts: 5156
  • Country: au
    • send complaints here
Re: routing ethernet signals between PHY and MAGJACK
« Reply #2 on: May 19, 2020, 08:52:47 am »
Fractions of a micron are optical wavelengths. Minimising mismatch in trace length is to keep the differential signals coherent (in phase) at the load. How much energy is there being carried along those traces at those wavelengths?
 

Offline phil from seattleTopic starter

  • Super Contributor
  • ***
  • Posts: 1063
  • Country: us
Re: routing ethernet signals between PHY and MAGJACK
« Reply #3 on: May 19, 2020, 03:50:05 pm »
Thanks guys, that was my thinking. All the guides just specify exact length and, short of a deep dive, I wasn't sure of the limits of the statement.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf