How to Make a Power Over Ethernet (PoE) Adapter | Hacker News Reader

How to Make a Power Over Ethernet (PoE) Adapter

tech2.in.com

41 pointsmnazim19 comments

Kind of misleading.

1) This isn't "PoE" in any sense of being able to power a PoE-compliant device. It's hacking up an Ethernet cable to carry voltage on 2 pairs not used for 10/100 Ethernet.

2) The article recommends a 7805 power regulator if you need to step down 12V to 5V (as in the example). While this would work, 780x regulators essentially throw off the unwanted excess voltage as heat, and total heat output is going to be a factor of the voltage drop and current. 12V-5V for a router (could be 1A or more) could end up being quite a bit of heat. You'll likely need to use an adequate heat sink with it. There are more efficient (and still easy to use) switching regulators available.

You are correct, but you can splice DC power onto one end of an Ethernet cable exactly like this to power PoE devices. It is quite effective, but is more dangerous than using an injector because it will supply power to non-PoE devices.

It's called "Passive PoE" and some devices support it out-of-the-box (e.g. WRAP board)

This is a neat hack for 100mbit networks, where only 2 of the 4 twisted pairs are in use. Unfortunately, this wont work for gigabit networks, where all 4 twisted pairs are used for data transfer.

There are gigabit PoE injectors, you can find cheap ones on ebay :)

IEEE 802.3at uses phantom power, all 4 pairs can transmit data.

I'm guessing they would take the existing voltages on each line and merely amplify them or similar. So the differential signal would just be using much higher voltage differentials. HVDS[1]?

I'd never thought of it this way, but I suppose networking devices could be built to subtract the excess voltages and noise into a power regulation circuit, which would power the device. In effect, a switch or router like this could be called "unpowered" as it's not using the traditional wallplug powerbrick.

[1]https://en.wikipedia.org/wiki/HVDS

Close... Ethernet is isolated through magnetic transformers on both sides, so any existing common mode noise or DC voltage is removed leaving only the AC data signal.

PoE (the standard, 802.3af/at) works by injecting DC power after the transformer on one side, and removing the DC before the transformer on the other side and using it to power the device. There is also a signaling mechanism to make sure the end device is PoE-compliant.

Passive PoE as seen in the article simply takes out the two pairs of Cat 5 that are not used for data transmission, and uses them for power.

It works in a pinch, but more caution is required to ensure the voltage injected is compatible with the device, there is not too much voltage drop in the cable, and there is not too much current going into the cable. Also, since these pairs are used for gigabit data, this method cannot be used with 1000baseT.

Mbit, not mbit, there's a huge difference (about 1000000000x)

I've never heard of a 100 microbit specification. I'm not even sure we can work with fractional bits. So how could I be referencing something that nobody talks about and conceptually can't exist?

You can buy adaptor lead sets from ebay and other suppliers that plug onto the data cable ends and give you an Rj45 port and a DC plug/socket pair - this is a much neater and quicker option.

eg: http://linitx.com/product/passive-poe-injectorsplitter-set/1...

This is a very bad idea from a system architect point of view. You would probably get another cable in there in the same amount of time, while having a much higher test coverage for your parts.

Physical connectors are there for the same purpose as type enforcement in strongly typed languages. They make things "just work".

Plus, what if you are ill and you want someone else fix an issue? What if you want to scale up the network later and need all wires?

If you don't hack you will need no docs.

This is called passive PoE, and you can get a special jack to do this for about $5.

We do this all the time with devices carrying the DMX protocol. As others have said, it works only in networks that aren't using all 4 pairs of ethernet cable -- which is the case in DMX.

The biggest issue when doing this is always sizing the power supply to the receiving device.

Wouldn't this method, especially with a switching power supply, introduce a lot of interference in the cabling?

Any pickup or noise in the power supply and router would be inductively/capacitavely coupled to the signal cabling inside the cable's shield, right?

> inside the cable's shield

There are no shield in Unshielded Twisted Pair cables used by most 100 Mbit/s networks and on photos in the article. Twisted pair by itself filters most of the noise because on every twist polarity of noise changes so it subtract itself.

Reminds me of the etherkiller: http://www.fiftythree.org/etherkiller/

if your house burns down and they find that, goodbye insurance payout.

Are handmade electrical devices generally a defense against homeowners' insurance payout?

Reference?

You'll have to find a copy of a homeowners policy for reference. What you're looking for is a clause that releases the insurer from responsibility in the event of gross negligence on the part of the homeowner; such clauses do exist. To be gross negligence, they'd basically have to show that you knew what you were doing would possibly lead to burning down your house.

Reply on news.ycombinator.com