4 wire remote voltage sense implementation

[Tinkerlad]

Hi All,

First off I will describe my project for context followed by my problem.

I am building a cluster of SBCs, namely I am using the Orange Pi Zero. The OPi Zero lists PoE as a feature however this is not true PoE and isnt standard compliant, what it actually means is the the PoE power pins are expected to carry 5v power. More details here:

https://linux-sunxi.org/Xunlong_Orange_Pi_Zero#Passive_PoE
https://parglescouk.wordpress.com/2017/04/14/getting-the-orange-pi-zero-working-with-poe/

I plan to have a few of these but dont want to have to put the 5v regulator on the Pi side of the Ethernet cable.

So here is my plan, I want to make a board that will have an RJ45 jack on either end, no magnetics. The data signals will pass straight through. Then I will have a power input to a (preferably switchmode for efficiency/heat) that will place 5v on the power lines in the RJ45 that is facing the Pi.

Now here is my problem, whilst the 'passive poe' uses two pairs of power conductors, I am worried about the voltage drop over the cables. So my plan is to do the following, only use one pair as power conductors and use the remaining pair as a sense return to do remote voltage sense (the pairs are connected at the Pi just out of the Pis ethernet jack).

But I cannot find anything on implementing 4 wire remote sense... I have found application notes on 2 wire virtual remote sense... But this is not what I am after.
Can anyone point me in the right direction on how to implement this? Also thoughts on the project?

[Tinkerlad]

After looking around some more..... I want to try just extending the feedback to the switching regulator along the second pair so that it senses the remote load, the feedback current should be minimal so shouldn't be affected by the distance I think... I have a couple of MP1584 modules laying around I might try modifying.

[viperidae]

You'd probably need an instrumentation amp to sense the remote voltage, since the switch mode controller probably expects the feedback referenced to ground.
I'd be concerned about oscillations too, since you're added a bunch of inductance and capacitance to the feedback loop.