We've integrated the Arduino Ethernet PoE into a Rev.1 product for RAD purposes.
Out of 5 delivered products all 5 have now failed within one month.
Out of the 3 devices in the office, 2 have failed during development (in an ESD safe environment).
Sometimes the failure occurs spontaneously during operation, sometimes when cycling the PoE connection. PoE source is professional 802.3at switch.
The Arduino is electrically isolated from any ESD damage via external connections and has an ESD safe enclosure with plenty of room above and below for good convective heat dissipation. It is driving only 35mA of external circuitry - well within specified capabilities.
The symptoms are always the same:
- Board powers up just fine (via PoE or programming header)
- Processor is responding just fine
- Ethernet connectivity is lost
Further troubleshooting shows that the W5500 chip is no longer responding correctly.
The W5500 chip is running excessively hot, reaching ~75*C within a minute, clearly indicating something is very wrong.
It's been noted by others that the W5500 is power hungry and should get fairly hot, but not like this - no way. The chip should consume about 0.6W of power, not enough for this sort of heating, but certainly enough to get warm to the touch. Perhaps, if this is a hot running chip, it was not a wise design choice to cover it with the PoE transformer essentially blocking off convection cooling from the chip.
I am not going to speculate on whether this is the reason for failure or not, but whatever the reason, these boards appear to fail very quickly even when used in a professionally designed environment.
It is our assessment that these boards should be taken off the market or be redesigned as they do not comply with market regulations.
As a comparison, we also have several products designed around the discrete W5500 and also the WIZ820io module.
They have all been running 24/7 for 6+ months now without any issues (related to Ethernet).