Power over Ethernet (PoE) basics and beyond

Other stuff engineers should know:

- PoE endpoints should have isolation barriers, factor this into cost and size estimates

- Don't skimp on TVS

- ideal diode full bridge rectifiers are really cool and you should use them (in more power entries than just PoE)

Awesome I’m just doing a POE design now.

- Was having a conversation today about isolation and grounding for POE (product has a metal case). Do you have a reference? Or standard?

- TVS ahead of the bridge right?

- Do you have a part recommendation or reference design for ideal diode POE?

> ideal diode full bridge rectifiers

For the less electronically inclined, an "ideal diode" surprisingly does not contain plain diodes, it refers to actively controlling MOSFETs to function as diodes.

They're more efficient and quite amazing in PoE applications in particular!

> - Was having a conversation today about isolation and grounding for POE (product has a metal case). Do you have a reference? Or standard?

Huh. I'm not the GP poster but interesting question. AFAIK there is no proper ground reference on the LAN cable. I'm not sure I've ever seen a metal case… wait, I do, outdoor wifi APs have metal cases sometimes.

If you find out, report back ;D

> - Do you have a part recommendation or reference design for ideal diode POE?

I've done a PoE device (802.3at, 25W) and just went with TI's reference design; the higher power ones use ideal diodes, sometimes there's multiple circuit variants.

(It's not worth mucking with the PoE design for small-scale builds; the reference design might be a bit more expensive but you get that money back on way less trouble to deal with.)

> Don't skimp on TVS

This. I know from a friend (ahem) that if you do, you will discover problems in production deployments, when it's too late to fix things.

The chassis net should be passed through the cable shield and the power isolated to force the current return through the cable in case power ground is bonded to chassis (which is commonly done, not sure why, I prefer a 1MOhm standoff). While not perfect this link is concise and provides handles to the relevant specs.

https://www.brainboxes.com/faq/power-isolation-in-poe-ethern...

Yeah, TVS before any other silicon junction. It's nice to throw a single-use medium or slow blow SM fuse before the TVS to open circuit in device faults.

This is going to be individual preference. I like the density and low design risk of fully integrated solutions like Microchip's PD70224. As long as you spec your FETs appropriately you can't go wrong with TI or AD options (VDS of at least 100 V, ID of at least as much current as you want to cram through with healthy headroom, RDSon that makes you happy, VGS that's compatible with the datasheet charge pump, size and cost that doesn't make you weep). When in doubt, stay very close to the datasheet's design.

I just saw that the PD70224 is not recommended for new designs. What an awful day to have eyes.

Oh, it's been superseded by the PD70288. Much lower RDSon, but a huge 8x8 package. The charge pump is mysteriously gone and there is now a UVLO of 24V. This is more PoE-specific, which is less generally interesting to me.

If only someone would sell me an ideal diode full bridge rectifier IC with integrated FETs, OCP circuit breaking, UVLO, OVLO, a fault flag, control input, and current monitor, I'd never buy a different power entry IC.

I use the FDMQ8205. It's an old part, a little pricey, but keeps the board footprint low. It also has a sufficiently high UVLO, so it acts like regular diodes during the classification phase and you don't need to factor those in to the Rcls values.

Amazing indeed- since PoE is DC.

Indeed, DC with unknown polarity. The switching frequency of the rectifier is amazingly low ;D