Polarity of Magnetics (Ethernet)

[GFTCTP]

Hello everyone,

I'm designing a ethernet PCB with LAN9354 and i have a doubt in polarity of magnetics.

The evaluation board schematic has the follow wiring (between PHY/SWITCH and MAGNETICS):



And the connector that I would be use:



My doubt: In evaluation board, the polarity in magnetics is on TD+ and RD+, and the polarity of chock mode is near to RJ45 side. In my connector JXD0-2005NL, the polarity of magnetics is oposit... Whats wiring option is correct (option1 or option2)?



Thanks in advance.

[ajb]

The magnetics don't care about the polarity of the pairs, in fact the PHY doesn't care either (well, maybe some PHYs do, but that would be very unusual).  Even if you use the 'right' polarity throughout your design, it may end up connected to another device that has the polarity swapped, so it really doesn't matter.  PHYs are generally designed to tolerate this, so you're free to swap polarity within pairs in whatever way makes your layout easier.  You can even swap TX and RX pairs if your PHY supports auto MDI-X, which most do these days, but check the datasheet to be sure.  Usually it's enabled by default, if not you can use the PHY's management interface to configure it.

EDIT: just noticed your part number, the LAN9354 actually calls out automatic polarity correction along with auto MDI-X in the first page of the datasheet.  That's unusual, most devices don't even mention polarity correction because it's just assumed, except maybe in calling out a bit in a status register that indicates the link polarity.  I guess someone wanted to fill out the feature list a little to make it look better

[jonpaul]

The transformer for isolation and Z match is center tapped and the dots show the winding starts. The windings are correct.

Ethernet and many other protocols are using Manchester, phase or other modulation that is not ploarity of signal sensitive.

Finally the CM choles dots show the windings are bifilar and wound in the same sense to forma a CM choke.

However the CM choke dots can be either to left or right as long as both are in the same direction.

Hope this clarifies,

Jon

[nctnico]

The magnetics don't care about the polarity of the pairs, in fact the PHY doesn't care either (well, maybe some PHYs do, but that would be very unusual).  Even if you use the 'right' polarity throughout your design, it may end up connected to another device that has the polarity swapped, so it really doesn't matter.  PHYs are generally designed to tolerate this, so you're free to swap polarity within pairs in whatever way makes your layout easier.  You can even swap TX and RX pairs if your PHY supports auto MDI-X, which most do these days, but check the datasheet to be sure.  Usually it's enabled by default, if not you can use the PHY's management interface to configure it.

I'd be careful with swapping polarity. By definition all ethernet cabling has a fixed polarity. Maybe it matters for 10Mbit which uses a different encoding.

To the OP: polarity of transformers doesn't matter for as long as the dots line up with the same signal (+in to +out). Also make sure to include TVS diodes and the option to insert EMI suppression beads in the lines between between the magnetics and the phy. Otherwise you might not pass testing as this crappy range of (formerly SMCS) ethernet chips push out a lot of HF junk.