We are having serious issues with DP83848CVV chip failures on a product we have designed and are starting to roll out into the field.
The most common issue: units are deployed and initially work. After some time the product no longer connects. When we analyze the board it turns out that the TXP and TXN signals are not proper. They are either greatly reduced in amplitude, or we see no RF output at all. Replacing the DP83848 fixes the issue.
In a few cases this symptom occurs with brand-new boards. In the case of brand-new boards, the probable cause might be ESD damage to the DP83848 during assembly by the CM. They are usually very diligent and we are not seeing ESD damage to any other parts on the board. Appears the DP83848 might be unusually sensitive? The data sheet however seems to say that the DP83848 has very good ESD specs.
Other details:
* This is a POE system using a TI TPS23754 POE power supply. That part of the board is working reliably, does not fail.
* Layout is based on the DP83848 POE dev kit--layers, layout, etc.
* Circuitry is very similar to the dev kit. The circuitry seems to work very well all except for this failure.
* We are using a MagJack that combines RJ-45, magnetics, and resistors and caps on the line side. Ratings are consistent with POE voltage and current specs and temp specs. We aren't seeing failures of this part (replacing the DP83848 fixes a failed board without replacing the jack).
* Ethernet cables are run 5 to 50 feet in an outdoor location to a server. The cables are CAT5. They are protected physically by raceways.
* We have patched all recent boards using a TVS array on the TXP and TXN signals. Part number is USBLC6-2SC6Y. Evidence so far says that this reduces the incidence of failure but does not eliminate failures.
* We think we are obviously going to have to add protective circuitry, on-board or off-board.
We would appreciate very much any comments and discussion as to why these particular signals are so vulnerable--in the field and even during board assembly. Thanks.
