WiringProposal

RFC: Improvements to PoE wiring convention

I'd like to suggest an improvement to the wiring described by the NYC Wireless POE wiring page. This informal RFC (Request For Comments) is designed to elicit comments from people who might be able to point out any disadvantage(s) of my suggestion, because I am currently aware of no drawbacks.

I'll start with the assumptions that are in common with the NYC Wireless POE wiring, and then follow with what I would suggest we change.

1) We can't (shouldn't) change ethernet wiring. It uses the orange and green wires, which correspond to pairs 2 (two) and 3 (three) in both EIA/TIA T568A & T568B.

2) Unless other signals are needed (e.g. serial console), that leaves the blue and brown wires for Power. All four wires should be used for minimum DC voltage drop on long Cat5Cable runs.

Where I differ from NYC Wireless is in how to assign the spare wires to the positive and negative leads from the Power source. NYC Wireless puts positive on both brown leads and negative on both blue leads. I am suggesting that we instead use blue/white and brown/white for positive, and solid blue and brown for negative. This will place a +/- signal on each twisted pair, and should improve noise rejection. Of course, not everyone is going to have problems with noise on their power lines, but it's better to keep the noise out if you can do it for free, because otherwise the filter capacitor(s) and regulator in the device on the remote end of the POE cable will be working harder to filter out any noise, and everybody knows that will generate more heat and reduce the lifetime of your remote device.

There seems to be a number of people on this list who understand twisted-pair wiring and other aspects of shielding, so I'm hoping to hear constructive criticism. Can anyone find a problem with the PoE wiring changes I suggest?

Note: I've already heard from EricJohanson, who recently measured the open-circuit resistance of each wire on a length of Cat5Cable. According to his measurements, and a little basic electronics resistance calculations, my wiring would result in a slightly lower resistance on the lines used for Power, which should help reduce the DC voltage drop and extend the usable length of PoE cables (but perhaps not by a significant amount). But, I'm not sure if all Cat5Cable exhibits the same resistance variances that Eric measured.

Brian Willoughby

Comments:

You asked for it

Twisted pair noise rejection works because when the wires are twisted close together, externally induced noise on both wires is the same, so the noise is cancelled out when you measure the difference between each wire. Also, radiated noise is reduced because the differential signal causes the wires to radiate equal and opposite signals, which cancels out.

But the really important thing is; radiated noise is caused by current variations in the wire. Wires used for signaling, signal by varying the voltage, which varies the current, which radiates RF noise. DC Power, on the other hand, has a fixed voltage, and the current remains fairly constant. This means radiated noise from DC power is nearly nothing.

DC power lines can get induced noise, but typically devices are pretty forgiving about a bit of noise on their DC lines, and have filters for it. The fact that the two + and - wires will be tied together at each end pretty much negates any "equal offset" benefits you might get for each individual pair. The Open Circuit resistance of ethernet wires is supposed to be the same for all wires, so you shouldn't get any voltage drop variation.

So the wiring you propose will give little benefit. Far more worrying and important, is you have moved the + and - connectors next to each other in the plug/socket, increasing the chance of shorting them.

Donovan Baarda