Magnetics in Ethernet connected devices

[ibrahimNazir]

I am about to make a schematics for ethernet switch and processor. I am using Ethernet switch IC which has 5 ports one port is connected to processor's buit-in phy. so I only left with 4 ports which are connected to ethernet devices that are present on same board. I want to know that is it fine to connect these devices without ethenet magnetics/transformers interface becuase they are on same board no external LAN wires are present.

 

[KlausST]

Hi,

As so often:

Read the datasheets.
In each device's datasheet you will find a specification about input- and output-voltages.
If all inputs fit to all outputs according: signal levels and common mode range, then I see no problem.

Klaus

 

[BMaarsen]

What type of drivers do the PHY's have, current mode or voltage mode? Most PHY's have an Auto MDI/MDIX feature, I assume that will not work now, but you don't need it. Maybe it avoids proper operation of the link. You can disabled this feature in the PHY.

 

[andre_luis]

I want to know that is it fine to connect these devices without ethenet magnetics/transformers interface becuase they are on same board no external LAN wires are present.

The quality and length of the ethernet cabling, as well as the presence of electromagnetic noise in the vicinity is something that indicate how much this transformer will be mandatory or not. I work with an industrial equipment that has an unshielded patch cable of about 20cm, with some solenoids, one compressor and few small motors around, and never revealed any problem, but as said above, reliability may vary depending on the characteristics of each chipset.

 

[FvM]

All recent PHYs (except some old 10 MBit devices) have current source outputs designed for external parallel termination. The receivers are differential amplifiers and are also using parallel termination.

I don't see a problem with Auto MDI-X feature. It requires that RX and TX ports have the same voltage reference, usually Vdd. So there will be less common mode interferences.

 

[BMaarsen]

We use mostly Micrel and Marvell PHY's and they use more and more voltage source outputs ... the last PHY with current source outputs we used was in 2006.

 

[FvM]

I believe what you name "voltage source output" are differential current sources with internal parallel termination, the preferred implementation for fast drivers (GBit/s). But there is no functional difference to series terminated voltage sources, if a chip designer would prefer it for any reason.

A suggestion from a Micrel application note that should work with most PHYs designed for external termination



Old style transceivers with internal Rx port bias (non Auto MDI-X) should use this circuit



In case of internal termination, bias chokes may be necessary to mimic the transformers DC path.

 

[BMaarsen]

You are right ... pictures are better than words ;-)