EEVblog #1122 - Raspberry Pi 3 PoE Hat FAIL Investigation

The new official Power Over Ethernet (PoE) hat for the Raspberry Pi 3 B+ has been reported to fail with an overcurrent error when powering USB devices.
Dave investigates and narrows it down until, well, something goes horribly wrong...

Support the EEVblog through Patreon!

Stuff I recommend:

Donate With Bitcoin & Other Crypto Currencies!

[fromtheashesit]

Eben Upton of the Raspberry Pi Foundation told The Register that the problem lay with "an interaction between the fairly low-frequency switching regulator on the HAT, and one of the two brands of USB current limiting switch that we use on the main board".

Upton continued: "Because the regulator operates at a fairly low frequency, each time it switches it moves quite a large chunk of energy into the three USB reservoir caps via the current limiting switch: this large instantaneous current is fooling the switch into thinking that a genuine over-current event is occurring."

[maikmerten]

2:08 - Isolation slot? I thought this gap is so the camera flatflex-cable can reach to the connector on the RPi.

[PlaywithJunk]

I found a solution and a colleague has already posted it in the forums. I cut the +5VDC PCB track near the fan connector and inserted a 4.7uH inductor in series to the output of the hat. I also added two 100uF ceramic caps to the 5V output behind the inductor. There are unpopulated PCB spots that can be used for the caps.
The result has a current draw on the USB ports of up to 1000mA without overcurrent event. With better analysis of the problem, this could possibly improved. But it shows that the problem can be solved by better filtering of the supply output.

[richardgoebel]

Congratulations Dave! You got rid of that horrendous ripple!

[mikeselectricstuff]

Should have used way bigger pads for that SM connector

[brainndamage]

They should've used through-hole pin headers so you could use those long pin headers designed for stacking. This basically prevents using the GPIO for anything.

[steverobbins]

I wonder why they used such an old PD chip? The data sheet says it's IEEE 802.3af compliant. The improved 802.3at standard (look closely, it went from 'f' to 't') was release like ten years ago I think. FYI, I was part of the IEEE task force that wrote that spec. The good old days.

[damians.]

How about EMC compliance? I did a PoE design this year similar to this one but using a TI controller running at 240kHz and the conducted emissions were just horrible. Had to use a big ass filter inductor as well as a large common mode choke to get it to a reasonable level. Not seeing any adequate input filter on this thing so it may very well not pass any standard.

[taitano]

You know... This thing is open source. You could make a pretty penny by making an improved HAT for the Pi. I'd love to buy an EEV-HAT.

[dosgos]

Please do a follow up video to finalise troubleshooting. Learn a ton from these videos. Even your off-the-cuff comments have good bits of information. Thanks!

[DalekmunTwo]

"I've killed it"
Don't worry, it didn't deserve to live anyway.

[geekrulz]

That isn't a isolation slot, but a slot for the extra camera module

[tylerandrews]

The MP 8007 is only an IEEE 802.3af chipset so it’s only rated up to 12.95 watts at the PD. Have you checked the combined loads to see whether the Pi is drawing too much power to add any significant wattage on the USB ports? It really needs the MP 8008 (PoE+).

[spikester]

PoE maxes out at 15W and you might be pulling near the 3.5A @5V capacity of the B+'s internal DC-DC. It's only PoE+ 802.3(at) that does a full 30W. There are also conversion losses to factor in there. Try checking the transient current draw.

[guesswhotoo]

Different failures on Pi3B in power supply section. After burning in 2 units for about 3 weeks, all 3 inductors in power supply section are over heating and cooking the PCB in the local vicinity of these parts. Obviously why + version got the supply area modified.

[NallTWK]

I bought an unofficial POE hat by Navolabs because it was much smaller so I could fit my pi into a case. I have yet to experience any problems with it, and I may buy another pi with the same POE hat. It may not work with all POE injectors though. My injector is a TPLink.

[poldiderbus]

At my last job, we designed a class 4(25W) PoE-hat for our rpi based IoT sensor nodes - for me it felt not quite right, that the murata reached >80°C(easily) under normal working conditions. So in the end I used thermally conductive silicone for glueing them to our finned aluminum housing..
.. oh, by the way: Its not a good idea to use the old rpi 3(and I guess the even older ones) in a PoE- environment - we've gone through a lot of pain before we realized that the rpi lan-chip is totally missing tvs-protection and prone to get killed by PoE transients.

[PeterCCamilleri]

I wonder... I you connect an capacitor in across the 5V output if it would lower the noise?

[NicoDsSBCs]

Hi Dave.
I recently made a video where I show that Raspberry has changed there frequency governor so that at 60C it clocks to 1200Mhz instead of doing that at 70C(when new/benchmarks were done). And that the Rpi 3b+ still says it`s running at 1400Mhz(or other overclock speed). This means without a fan it now runs slower than the 3B when stressed. I`ve made a tread about it in the rpi forum. First they laughed it away. They changed the tread title and comments. They terminated the tread so no new comments can be made, and it doesn`t appear again. Important information about this was removed. They came with the excuse that nobody notices it so they don`t care. When new this all worked fine.
It is a small thing. But we want to know what the real clockspeed is instead of a fake number.
Here my video. Links to the forum in the discription. It`s not my best video, but the info is all there.
I love sbc`s. I want them to be as good as possible.
Greetings.

[bjornroesbeke]

Five different power supplies, three different cables and the Pi is still complaining about the lack of available power. I had my hopes up for this PoE hat!