While bringing up a board with Power over Ethernet on it for an RFID/NFC project I came upon an issue with linear tech design note 361. It is a really interesting design which uses the LTC4267. The LTC4267 is a Power over Ethernet interface and a 300kHz dc-dc converter in one package. It is quite a nice chip if a little expensive, like most LT chips. The design note also includes a battery circuit which acts like a battery back up as well as a temporary power booster that lets a circuit draw more than the PoE spec allows. It is really awesome for remote devices that need to be stable like a door lock controller or similar device. I am not using the battery circuit to keep costs down.
The problem I had while bringing this board up was the LTC4267 not getting past the detection stage of PoE. This is the stage where the power sourcing equipment (switch or PoE injector) sends out a small voltage to see if there is a powered device (rfid controller, ip cam, etc) at the other end. With the ethernet cable disconnected I was able to inject -48V in to the system and the dc-dc converter was outputting pretty close to 5V so I knew the LTC4267 was fine. I then connected the board to a PoE switch and probed across the positive and negative inputs (connected to VPortP and VPortN of the LTC4267 respectively) and got the following waveform:
I knew something was up after double checking my schematic and the design note and seeing nothing wrong so I posted on the eevblog forums. Jay_Diddy_B was quick to point out that the input capacitance I had was too high for the PoE spec (802.3af). The LTC4267 datasheet shows an example schematic with 4.7uF capacitor (C5) from VPortN to GNDA (the correct way) rather than VPortN to VPortP like design note 361 shows it (the incorrect way). This was apparently adding too much capacitance for PoE to initiate properly. The difference in the two is that the LTC4267 connects GNDA after classification which prevents the capacitor (which is connecte to VPortP and GNDA) from charging.
All that I had to do to get everything working was cut the trace to the negative pin of capacitor C5 and connect it to GNDA. After this it powered right up!
I'm pretty excited since this board is by far the most complex board I have brought up and everything has worked so far except for this one small issue!
Here is the board powered up:
The wires are all test points for different voltage rails.
Thanks again to Jay_Diddy_B!