Ethernet Termination Issues

Question

I recently designed and fabricated some prototype boards based on the W5500, but I mistakenly used regular Bob Smith Termination for the differential traces coming out of my RJ45 jack with magnetics. Communication between my microcontroller and my W5500 via SPI seems to work OK, and the W5500 can blink the LINK and ACT leds, but the W5500 is unable to acquire and IP over DHCP or otherwise interact with the ethernet port on my router. My current suspicion is that my Bob Smith termination is at fault. I’ve noticed that in the reference schematics, the W5500 has a 6.8nF capacitor on both of the RX lines, and the 49.9R resistors on the RX lines are terminated to ground instead of +3V3. What functions do these 6.8nF capacitors perform (I’m assuming an AC-pass filter of some sort)? Would the chip still work without them? What kind of issues might be caused by terminating the RX lines to +3V3 with the 49.9R resistors instead of GND? I have uploaded my schematic as a PDF. The W5500 is on the second page. oops.pdf (244.9 KB) Thanks for your time!

Eugeny · Answer

CoolNamesAllTaken: What functions do these 6.8nF capacitors perform (I’m assuming an AC-pass filter of some sort)? No, they are DC decoupling. I suspect for the cases when you mistakenly connect CT to the 3.3V (or have transformer with only one CT - see section “RJ45 with integrated Transformer and connected CT”). CoolNamesAllTaken: Would the chip still work without them? I think you have proven that it will not, or better to say functionality is not guaranteed. Change your circuit and board according to the reference circuit diagram. I hope you’ve seen these documents before designing your board. Also notice that W5500 is not MDIX capable.

CoolNamesAllTaken · Answer

Thanks Eugeny, I’ve revised my board to follow the reference diagram RJ45 termination, and have phase-matched my differential pairs out of an abundance of caution. I’ll update here if things go wrong. After looking at the “RJ45 with Integrated Transformer and connected CT” schematic, it looks like things are expected to fail if the RX+ and RX- lines are connected to +3.3V in any way–that seems to be my mistake. I wonder why the DC decoupling capacitors are present in the reference design even when the center tap of the RX transformer is not connected to +3.3V… In any case, thanks again for your input! I’m using a router with my board (no devices in between), so I believe that I wouldn’t require MDIX functionality. Hopefully the board works this next time around.