Hi Matt and the group,
You raise some interesting points with the POE measurement:
1) in some implementations the spare pairs are sometimes used for power transfer, this may require current monitoring of all 8 wires.
2) The higher power applications (POE+ and POE++) use all 8 wires.
3) Back to back transformers may help, so that you can measure the currents between the center taps?
You also need to make sure that the measuring circuit doesn't interfere with the classification process.
Have you considered using a PD followed by the measurement circuit followed by a PSE? This would be a special mid-span unit?
I hear what you saying about the LTC2946 resolution. 12 bits is equivalent to using a 3 1/2 digit meter at half of full scale. There is a separate ADIN pin which can be used for the lower voltage applications.
The total unadjusted error (TUE) includes, the divider resistor tempco, the reference drift, all over the full operating temperature of the part. Like you say it depends on how much accuracy you need.
I wish you well with your campaign.
Regards,
Jay_Diddy_B
We allow for primary and secondary pair simultaneous measurement. 8 wire measurement is overkill as the current travelling into one transformer pair is the same current which travels out another. In the unlikely event that you have damaged return path windings you will not be able to detect this without full 8 wire current measurement and I did look into having a manual switch for flicking between primary and secondary pair measurement but it was a clunky solution better left to the likes of Cisco (who may need this ultimate level of scrutiny on a production line). These devices could be rewired to be used in this way, with pairs of two LogIT units, for the hardware hacker. If you do damage ethernet transformer windings or suspect substandard parts on a product under development one of the best ways to check this is to use a device like the LogIT.PoE and plug in an identical known good device to cross-check the measurements.
One issue of major consideration though is impedance matched data lines. We have gone to a lot of effort in this area to ensure data integrity. More wires measured means more impedance mismatch. These devices are supposed to be transparent to data so impedance is important.
I can't go into too much more detail but be assured that we have total current consumption covered for PoE devices independent of transformer coil mismatch.
The PoE handshake ID resistances are orders of magnitude higher than the sense resistors so there is absolutely no issues with class identification at all levels from the 7W standard to the less official 92W class.