Distance between MAC and PHY ?

[lpcware]

Content originally posted in LPCWare by GPh on Fri May 09 08:13:34 MST 2014
Hi,

We have some mechanical restriction and we need to move ethernet on addons card. the two cards are in the same box but with link cable about 6cm.

Can we put the Ethernet PHY (LAN8720) on addon card or we need to put the PHY near the LPC1768 and move RJ45 only on addons card ?

Thanks for your reply, Philippe.

[lpcware]

Content originally posted in LPCWare by GPh on Wed Nov 05 02:31:21 MST 2014
The project was in standby and restart now   :)

Thanks a lot for your reply.
I put the PHY near the LPC and put RJ45 and other components on add-ons.
It's seem to be less dangerous !

[lpcware]

Content originally posted in LPCWare by pierre on Tue Jul 29 06:16:45 MST 2014
The problem you have here is that RMII link between MAC and PHY uses only one clock for both directions, and it is quite fast. It is not like a source-synchronous link, which can be pretty long, provided clock and data lines don't have too much skew.

Suppose you have a board with a MAC and a 50MHz clock. Each clock edge travels along the cable, then the PHY reacts and outputs some data, which has to travel along the cable to reach the MAC again before the next clock edge. On a PCB, you'd place the clock between the PHY and the MAC, so clock travel time to both chips should be roughly equal, and you'd only consider a one-way propagation time on your signals. But in your case, you have to consider the round-trip time... It may work or not, you will certainly need a GND wire between each signal wire in the cable.

On the other hand, twisted pair ethernet is extremely robust. If your cable is roughly the right impedance, 6cm of ribbon isn't going to matter a lot versus 50m of Cat5...

[lpcware]

Content originally posted in LPCWare by hlsa on Tue May 13 23:06:02 MST 2014
Hello Philippe,

I cannot give you a definite advice (and I assume no one can, since this topic is too complex). However, I can give you three hints:

1. Have a look at AN-1469 from TI, especially at chapter 5.2 (Max Trace length). This application note is for a TI Phy, but the recommendations should apply to other devices as well.

2. Several years ago I made my first ethernet device, where I had a similar problem. I put the phy far away from the MAC, but close to the RJ45 connector. This was a catastrophe. The connection between the MAC and phy is very critical. (In addition, I placed the connection traces in parallel to a motor power connection. Bad mistake.)
Since this time, the first thing I do in a new device, is making sure that the phy and MAC are placed next to the RJ45 connector. I am doing this step before the mechanical design.
If I need to chose between long traces from MAC to phy and from phy to RJ45, I would chose the latter one.

3. I do have a prototype of a new device, where the PHY is put on an add-on card. However, the distance between MAC and Phy is about 2 inches and the connector between the main print and the add-on card is full with GND pins. Everything works fine.

Best regards,
Holger