FEDEVEL
Platform forum

oscargomezf , 09-01-2016, 08:46 AM

Hi,

I'm working on a new design with a Giga PHY ethernet. There is an evaluation board that uses the Giga PHY ethernet I need, so I'm going to purchasing it. It's this: http://ethernetfmc.com. My design it's going to be faster because I can check everything before designing and manufacturing anything.

The problem is the Giga PHY ethernet is an IC from Marvell, so to access the datasheet and all the documentation I have to wait to sign a DNA. I was researching about other Giga PHY ethernet and I found this part number KSZ9031RNX from Micrel (nowadays microchip). It's very similar and the questions I've got are common to both.

1º.- The datasheet only told about using LAN transformer + RJ45 without magnetics. But I would like to use an RJ-45 with transformers. I usually work with Wurth Electronic for connectors, power chokes, inductors and TVS, everything I can because you can get free samples and you can buy the vast majority part numbers without an MOQ and also you can get the footprint to use with Altium (with 3d-step models), so I'm delighted with Wurth Elektronics.

I was trying to find a good RJ-45 LAN transformer and this part number seems to be cute: 749 911 161 4A, because it's ready for Low EMI:

WE-RJ45 LAN: RJ45 LAN Transformer Low EMI. 10/100/1000 Base-T High-Performance Low EMI.

Do you think I could use an RJ45 LAN Transformer instead of a LAN transformer + RJ-45 connector (without transformer)? I don't know what advantages I would have to use a LAN transformer + RJ-45 connector (without transformer).

2º.- Another important question about length matching in RXDx, RX_CLK, RX_CTRL and TXDx, TX_CLK, TX_CTRL. I've read in the product brief of the IC 88E1518 of Marvell the following: "Four RGMII timing modes including integrated delays • This eliminates the need for adding additional trace delays on the PCB" Does anyone what it means? I mean you don't have to match length, doesn't I?


Best regards.

robertferanec , 09-02-2016, 09:34 AM

1) Have a look at OpenRex - we used there KSZ9031RNX and RJ45 with integrated transformer. However! After testing it, we found out, that KSZ9031RNX has problems with low temperatures and Micrel confirmed, that is a silicon issue. Therefore, instead of using KSZ9031 we switched back to KSZ9021 which we used and tested in our previous designs and which works fine in the whole industrial temperature range. The chips are not 100% pin to pin compatible, but you can have same design for both chips - have a look at OpenRex schematic (page 15 ) >

2) I think, the first RGMII specification required this delay to be added on PCB track length. I guess, soon people found out, that this important point is going to be missed by many engineers and silicon manufacturers added option to adjust delay through registers inside the chips. So, basically, you can route all the track the same length and adjust delay inside KSZ9021. Also, I think KSZ9031 has already this delay added directly in silicon, but you still can do some changes through registers.

oscargomezf , 09-28-2016, 03:01 AM

Hi Robert,


I'm trying to understand how to design a gigabit phy ethernet, and I read this paragraph in this document of LAN7500/LAN7500i Layout Guidelines:



On the page 18/24 5.3 section, point 6:

6. To maximize ESD performance, the designer should consider selecting a discrete transformer as
opposed to an integrated magnetic/RJ45 module. This may simplify routing and allow greater
separation in the Ethernet front end to enhance ESD/susceptibility performance.


What do you think?

Best regards.

robertferanec , 10-01-2016, 07:27 AM

It can be possible - distances between individual coils inside the integrated RJ45 are probably smaller (it's more packed) than distances between coils inside a standalone transformer, so yes, standalone transformer may probably deliver better ESD protection comparing to the integrated RJ45. You can compare datasheets, they should say it there.