pic: Power Over Ethernet for 2017

This diagram explains how to create a custom wire to power the OM5P-AC and OM5P-AN through the 18-24V Ethernet port while compiling with R63.

Thank you for this resource. At out meeting tomorrow, one of the things on the agenda is tackling POE. This should make the process much quicker.

Thanks for this, but a few questions.

  1. Will any inspectors not understand this and fail it because they have never seen it? I know that impossible for people to answer but whats your opinion.

  2. How reliable is this for vibration? I know last year was a tough game for robots and many people has the radio power fail them, but is this any better? If seems like it would be but do people know for sure?

  3. I assume you have to make up your own custom cable by using an RJ-45 end and then just running the conductors where they need to go?


  1. Definitely bring along a diagram with you to explain to your inspectors what this is.

  2. I didn’t have experience with it personally, but I heard lots of anecdotal evidence that this was the go-to solution for vibration-related radio power issues for many teams last season.

  3. Yup. It’s essentially making a custom/more compact version of one of these cables: Passive PoE Injector Cable Set : ID 435 : $5.95 : Adafruit Industries, Unique & fun DIY electronics and kits. There are several available on Amazon as well, if you want to avoid making your own cable.

We used the Adafruit cable last year and liked it a lot. It’s also nice because you can use the POE and barrel jack together and have some redundancy.

Also Mark McLoud posted, in another thread, this cable:

It looks like the adafruit ones. Didn’t see it when I posted the picture. Also here is a Q&A saying this sort of thing is legal:


Your inspector’s understanding of it will vary.

Awesome ! Thanks.

I’m assuming you mean run a separate cable for the barrel jack and not trying to plug in both the Ethernet pigtail and the barrel jack from one of the Adafruit cables into the radio right?

Just so it’s clear to others, that would not result in powering the “PoE” port on the radio as the Ethernet pigtails on those cables do not have power on the pins, the Ethernet cable you connect to the female Ethernet jack on the cable does.

We are planning on using a passive injector similar to these: https://smile.amazon.com/gp/aw/d/B00S5ROQO6/ref=mp_s_a_1_38?ie=UTF8&qid=1486479541&sr=8-38&pi=AC_SX236_SY340_QL65&keywords=passive+poe

We will be removing the barrel plug and soldering in a whip that can plug directly into the VRM.

Note passive poe injectors are explicitly allowed per the blue box of R51. So as long as you can explain I would find it hard to fail inspection.

The advantage of using something like the one I linked is the reduction of single point failures. There’s only one cable plugging into the radio (not an Ethernet cable AND a barrel plug). This is at a cost of a single additional RJ45 connection and the loss of a barrel jack connection (which is the one that usually fails you since there is no positive retention. With the previously linked cables, you’re reducing the wires that need to run across the robot, but you have increased the number of failure points as there are two additional RJ45 plugs and one additional barrel jack connection. So reliability has actually gone down.

I think the idea is to take the power for the Radio and split it with something like this:

You then plug one of the male plugs directly into the radio, and the second male plug into:

The male Ethernet connector goes into the RoboRio, and you use a CAT5 or CAT6 cable to connect the female end into the radio.

Thus, the radio receives power directly, and through the Ethernet cable.

Interesting. I think I like the pigtail option better because we can use electrical tape to keep the plug from coming apart.

FYI, this was covered in LRI training as well. So as long as you really are using a passive injector, your worst case scenario is having to escalate to the LRI.

Yes I mean running a separate power cable for the barrel jack, and they are both powered off the VRM. You will know you did it correctly when you can unplug the barrel jack or the Ethernet cable and the radio still maintains power as long as one of them is still connected.

I don’t quite follow this logic.
My power injector uses three connections:

  1. roborio RJ45
  2. radio RJ45
  3. direct wired into VRM after cutting the barrel connector off

And an additional Ethernet cable is required from the passive POE injector to the radio.
So, four connections altogether.

Rob’s solution has three total connections (and a few more user modifications).

Your solution appears to require one additional Ethernet cable adding RJ45 failure points over and above those others even after you’ve bypassed the power barrel connector (the user modification is also another potential failure point).

While the number of potential failure points has increased, the likelihood of a failure of any one of those points has decreased. Hence, the overall reliability has increased.

In our solution, we are powering both ways. So the risk of failure of the original path is unchanged, but now there is a second path that has to independently fail.

How do you comply with wire gauge restrictions when doing this? Most ethernet cables are 23-26 ga. (usually 24) but 22 ga. is required in the manual for 12v 2a. Would it be legal to crimp 24ga (either from an ethernet cable or individual) wire into the RJ connector then solder the +v pair to a 22ga wire heading to the VRM and the same for the ground?

I was assuming you were using the connector link attributed to you (https://www.amazon.com/iCreatin-Passive-Injector-Splitter-Connector/dp/B00NRHNPUA/ref=sr_1_19?ie=UTF8&qid=1486416072&sr=8-19)
on both ends. Apparently I’ve misunderstood the planned implementation. But my comment of things being less reliable assumed you were plugging the dongle with the male barrel into the radio, along with the RJ45 end.
The other dongle plugged into the RIO, then a power extension plugged into the female barrel to get to a VRM. An Ethernet cable plugged in to bridge the gap between the two dongles. So to tally things up I was expecting 4 Ethernet connections and 2 power connections. Basically exactly the layout depicted in the last two images of the product page linked.

If you’re only using one of the cables, cutting the barrel jack off one end, and plugging it into the VRM directly… then I would agree you have one fewer Ethernet connections in this implementation than what I was suggesting. As far as I can tell, nobody suggested this implementation before I posted.
In this configuration, you will need the VRM immediately next to the RIO unless you’re going to solder extensions to the power leads (which per your comment - you apparently consider to be introducing a failure point). Personally I trust our teams ability to produce reliable solder joints, so I consider this a wash.

Your solution appears to require one additional Ethernet cable adding RJ45 failure points over and above those others even after you’ve bypassed the power barrel connector (the user modification is also another potential failure point).

The component I suggest effectively turns two RJ45 connections and a barrel jack into 4 RJ45 connections with zero barrel jack connections. Personally I consider the RJ45 connections to be one of, if not the most, reliable connections on the robot. To prove a point, I’ve lifted up a kitbot chassis by an Ethernet cable, maintaining active comms. The objective for me in using POE is to remove the Barrel plug from the equation, as in my experience this is the only component which has contributed to comms problems on our robots. We usually hot glue them in… I’ve never had to hot glue an RJ45 cable into a socket.

I did consider using a cable like the ones you’ve linked. My concern with them was the quality of the crimps and whether or not the cables would be twisted pairs or (non-compliant) straight through in the dongle themselves. When I was doing my research at the beginning of the season, the pictures few pictures of the dongles I saw online looked to be questionably small gauge straight through conductors. I didn’t have any of these on hand to check myself though. So I made a judgement call to go with the PCB version like the ones I linked earlier.

All that said, no need to defend your choices to me. I was just sharing what we were planning on doing since I hadn’t seen a POE injector like the one we’re using in this thread yet…

Ah, too many assumptions.
You can’t actually use that injector in the configuration you were thinking of.
Power is only injected into the female end, not the male end.

This is the picture I posted of my setup when I talked about it: https://www.chiefdelphi.com/forums/attachment.php?attachmentid=21527&d=1484316250

22awg =0.326 mm^2

The passive power injectors use pairs of conductors for sending power and ground (a total of four conductors used).
24awg = 0.205 mm^2
So you effectively you have 0.410 mm^2 cross sectional area of copper which exceeds the 22awg requirement.

At least that’s how I was planning on explaining it away if the question came up.

If you read what I wrote, I did assume the female end was where the power was being injected from. I assumed it was being used literally the way it’s depicted in the product posting.
edit: specifically:

The other dongle plugged into the RIO, then a power extension plugged into the female barrel to get to a VRM.