Sorry in advance I am fairly new to the electronics world and Ethernet signaling. I am a physics postgrad student doing some hardware work rather than an electronics engineering background.

I have to connect two ethernet modules on a backplane PCB over Ethernet. The modules are pre-made and I have no choice but to use them. Since they are on the same board and in very close proximity I am trying to eliminate the magnetics modules.

One of the devices uses a KSZ9031 ethernet PHY by Microchip for which there is lots of documentation and the other uses a BCM54210 ethernet PHY by Broadcom which unfortunately there is virtually no published information.

I have been researching capacitive coupling of Ethernet and dug out this:


So it seems that by luck this is a good choice as it has termination and voltage mode interfaces on board. So far so good, just stick a capacitor in each line.

But this assumed that both ethernet PHYs are the same.

More generally both Microchip and Texas Instruments both publish documents like these:


which show me how to couple and terminate for PHYs that either have an internal DC bias or without an internal DC bias.

Since I do not know anything about the second PHY by Broadcom including termination and DC voltage bias how do I know which circuit type to pick?

Also, if anyone has any information they know about the Broadcom ethernet PHY I would love to know and if anyone has any general hints or experience with capacitively coupling ethernet that would be great too!

Thanks very much :)

  • \$\begingroup\$ They may not work correctly without their magnetics. If you think they will, what inspires you to believe this? \$\endgroup\$ – Andy aka Mar 10 at 14:48
  • \$\begingroup\$ For the Microchip one their own tech notes suggest it could work and for the Broadcom one the device in question is a Raspberry Pi Compute Module 4. On some forum posts some of the Raspberry Pi engineers were mentioning that with the appropriate knowledge and skills could capacitively couple the ethernet PHY instead of magnetics but unfortunately no further information on how. Thanks \$\endgroup\$ – sbash Mar 10 at 17:26
  • \$\begingroup\$ It still depends on your required CMRR design spec (dB) (or expectation) ?? Also what % is your R tolerance and is shared Vdd? and what % ripple? \$\endgroup\$ – Tony Stewart EE75 Mar 10 at 22:40

The magnetics serve an important BALUN function to enhance signal integrity of CMRR from crosstalk , mismatch reflections and stray noise from SMPS on ground paths.

At 1Gbps even with spectral compensation training signals for HDMI, even my ASUS tower has problems that are improved with some finger capacitance over the SMD magnetics on high contrast pixel zones with small eye pattern margin. However your backplane design may be better than my HDMI cables. So I use analog VGA for my 1m TV cable to the tower with 200kbps pixel rates or so. Yet my laptop had no issues. Earth Ground is not always “0V” at 1GHz!

  • \$\begingroup\$ Is the BALUN point negated by the fact that there isn't a medium transition into a twisted pair cable. They are controlled impedance differential pair traces all the way through? But thank you very much! I never assume anything I design would be better than your HDMI cables ;) \$\endgroup\$ – sbash Mar 10 at 20:15
  • \$\begingroup\$ I don't know if your design is Dual Stripline *best or coplanar or if vias exist and how well impedance matched it is and if electrical tested at fab shop or your resistor tolerance or if dispersion occurs with PCB spectral dependent dielectric constant and loss tangent. These aspects can affect PSRR & CMRR at UHF and signal integrity. So it depends on all these details. But if you test scattering parameters, and crosstalk you will know. \$\endgroup\$ – Tony Stewart EE75 Mar 11 at 11:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.