Forums

Ethernet transformer vs. RJ-45

Started by ericbogomolny July 15, 2009
Hello,
I have seen in several layout guidelines that the Ethernet transformer
should be placed as close to the RJ-45 connector as possible, while the
close placement of a PHY to the transformer is not as critical.  Why is
that?  It seems counter-intuitive: I would think that the PHY should as
close to the transformer as possible, while the lines between the
transformer and the RJ-45 are just differential pairs and, thus, are not
much different from the CAT5 cable.  So, why is the close placement of the
transformer to the RJ-45 is more critical than the close placement of the
PHY to the transformer?
Thank you,
Eric.



ericbogomolny wrote:
> Hello, > I have seen in several layout guidelines that the Ethernet transformer > should be placed as close to the RJ-45 connector as possible, while the > close placement of a PHY to the transformer is not as critical. Why is > that? It seems counter-intuitive: I would think that the PHY should as > close to the transformer as possible, while the lines between the > transformer and the RJ-45 are just differential pairs and, thus, are not > much different from the CAT5 cable. So, why is the close placement of the > transformer to the RJ-45 is more critical than the close placement of the > PHY to the transformer? > Thank you, > Eric. >
This is just a guess at a theory... Perhaps it is because on the PHY side, the lines are a differential pair referenced to the board's ground, while after the transformer there is no common reference. There could therefore be a significant voltage difference between the transformed pair and the board's ground, or significant common-mode noise.
"David Brown" <david@westcontrol.removethisbit.com> wrote in message 
news:4a603031$0$26345$8404b019@news.wineasy.se...
> ericbogomolny wrote: >> Hello, >> I have seen in several layout guidelines that the Ethernet transformer >> should be placed as close to the RJ-45 connector as possible, while the >> close placement of a PHY to the transformer is not as critical. Why is >> that? It seems counter-intuitive: I would think that the PHY should as >> close to the transformer as possible, while the lines between the >> transformer and the RJ-45 are just differential pairs and, thus, are not >> much different from the CAT5 cable. So, why is the close placement of >> the >> transformer to the RJ-45 is more critical than the close placement of the >> PHY to the transformer? >> Thank you, >> Eric. >> > > This is just a guess at a theory... > > Perhaps it is because on the PHY side, the lines are a differential pair > referenced to the board's ground, while after the transformer there is no > common reference. There could therefore be a significant voltage > difference between the transformed pair and the board's ground, or > significant common-mode noise.
My theory is.... All fast signals weather differental or not should be short as possible. The Micro, Ethernet PHY, Magnetics and R45 should be as short as possible. Joe
On Jul 17, 8:35=A0am, "Joe G \(Home\)" <jo...@optusnet.com.au> wrote:

> >> I have seen in several layout guidelines that the Ethernet transformer > >> should be placed as close to the RJ-45 connector as possible, while th=
e
> >> close placement of a PHY to the transformer is not as critical.
> My theory is.... > > All fast signals weather differental or not should be short as possible. > > The Micro, Ethernet PHY, Magnetics and R45 should be as short as possible=
. And then you plug a 50-foot patch cord into the RJ45...
cs_posting@hotmail.com wrote:
> On Jul 17, 8:35 am, "Joe G \(Home\)" <jo...@optusnet.com.au> wrote: > >>>> I have seen in several layout guidelines that the Ethernet transformer >>>> should be placed as close to the RJ-45 connector as possible, while the >>>> close placement of a PHY to the transformer is not as critical. > >> My theory is.... >> >> All fast signals weather differental or not should be short as possible. >> >> The Micro, Ethernet PHY, Magnetics and R45 should be as short as possible. > > And then you plug a 50-foot patch cord into the RJ45... > >
I think a general rule about impedance controlled transmission lines is that you want to have any unavoidable impedance "bumps" done with over as short a distance as possible. Thus the path from the transformer, over the pcb, and into the RJ45 should be short, while the smooth impedance cable can be longer. Basically, you want the electrons to go over the speed bumps before they have had a chance to accelerate to full speed along the straight cable. (Perhaps someone who actually knows the answer could tell us.)
"ericbogomolny" ...
> Hello, > I have seen in several layout guidelines that the Ethernet transformer > should be placed as close to the RJ-45 connector as possible, while the > close placement of a PHY to the transformer is not as critical. Why is > that? It seems counter-intuitive: I would think that the PHY should as > close to the transformer as possible, while the lines between the > transformer and the RJ-45 are just differential pairs and, thus, are not > much different from the CAT5 cable. So, why is the close placement of the > transformer to the RJ-45 is more critical than the close placement of the > PHY to the transformer?
The lines between RJ45 and transformer need 1500V isolation to the rest of the world, the lines to the PHY do not. Just think about running long 1500V isolated lines over a crowded PCB... Arie
Arie de Muynck wrote:

> > "ericbogomolny" ... >> Hello, >> I have seen in several layout guidelines that the Ethernet transformer >> should be placed as close to the RJ-45 connector as possible, while the >> close placement of a PHY to the transformer is not as critical. Why is >> that? It seems counter-intuitive: I would think that the PHY should as >> close to the transformer as possible, while the lines between the >> transformer and the RJ-45 are just differential pairs and, thus, are not >> much different from the CAT5 cable. So, why is the close placement of >> the transformer to the RJ-45 is more critical than the close placement of >> the PHY to the transformer? > > The lines between RJ45 and transformer need 1500V isolation to the rest of > the world, the lines to the PHY do not. Just think about running long > 1500V isolated lines over a crowded PCB... > > Arie
Correct - another similar issue is 3 phase. If one end of the ethernet was powered up by one phase, and the other receiving end was on a different phase, you can get minute leakage and 415 volts. You don't want that all over your system.