DIY PoE on the cheap - DDY's (failed) attempt

[DDY]

It didn't actually fail, it works, sort-of. I won't say what went wrong, yet, because I'm going to make a game out of it, guess right and get a gold e-sticker. HINT: You can see it (or lack thereof) in the pics below.

Anyway, here's the write up.

My goal was to place my IP camera somewhere far from a power socket but within reach of a network cable. The obvious solution is PoE (Power Over Ethernet), my attempt at PoE involves sending DC power though an active network CAT5 cable by using the two unused pairs when running at 100/10, note this approach isn't 'proper' PoE, .i.e. isn't standard IEEE 802.3.

This would be trivial if the camera ran at 12V or greater, all I'd have to do is make a simple splitter. But the camera I have runs at 5V which on a decent length of network cable will result in significant losses from the higher current when running at low voltages. The solution here was to send a higher voltage though the cable and step it down to 5V at the IP camera end.

Another goal was to do this using only the parts I had lying around, this included;

19V Laptop charger
2x Double RJ-45 wall sockets
LT3083 Power regulator
Various other components, resistors, caps, proto board etc.

The idea was to send 19V down the ethernet cable and step it down to 5V using the regulator, all housed in compact RJ-45 wall sockets.

So I cobbled this together;


Power regulator circuit testing


The two RJ-45 sockets


Regulator circuit stuck together on PCB and crammed in to one of the sockets, this is the device (e.g. IP Cam) end of the PoE system.


Another view.


This is the 'injector' end of the PoE system. A similar system without the power regulator would resemble this on both ends. The red cable running out is an inline DC jack socket where the power adapter is connected.


Everything hooked up and working. The black box is my multimeter ammeter adapter, I built it to make current measuring easier. The DMMs left and right are measuring voltage and current respectively.

HINT 2: Again, it works but with one major issue preventing me using it with high load appliances with potential for injury if the case housing the regulator is open.

EDIT: Yes the datasheet in the background is for a LT1185 rather than the LT3083, but that's not what went wrong

[DanceswithUnix]

Sooo, dropping 19V down to 5V with a linear regulator?
That's a 14V drop with an indicated 0.61A, so 8.5W to dissipate on a little heatsink.

Did it actually melt anything?

[DDY]

Sooo, dropping 19V down to 5V with a linear regulator?
That's a 14V drop with an indicated 0.61A, so 8.5W to dissipate on a little heatsink.

Did it actually melt anything?

That's it

The 0.61A current draw is when both pan and tilt motors are turning, I made the mistake of going by my measurement of 0.2A when neither motors are turning!

I only had the thing running for about a minute knowing I've made a mistake when 0.6A appeared on the DMM, so no damage done. Not having an IR thermometer to hand I resorted to using water to estimate the heatsink temp - which evaporated almost instantly! I reckon quite a bit over 100°C

Here's the residue the drop of water left, I couldn't catch the boiling action because I wasn't fast enough with the shutter release!

[watercooled]

Yeah ~19v down to 5v on a linear regulator had me wondering. Even 1W would probably break spec temperature with that tiny heatsink in an enclosed case. They're often happy to run at over 100C but at risk of melting stuff around them. With over 8 watts though, well, Intel Atom isn't rated much higher than that and look at the size of the passive heatsinks, and they still get hot in an enclosed case!

On the subject of IR thermometers, they also have their problems, especially when measuring reflective surfaces like uncoated metals - you can be essentially reading the temperature of whatever is reflecting off of it, and small devices can be awkward to measure accurately anyway. Adding a sticker or a bit of tape can help, but I find a thermocouple in a dot of water (given time to stabilise) is generally accurate enough.

For a fix, you could use a buck converter. You can get basic pre-assembled boards dirt cheap on eBay, or another shop I found (ships from overseas though): http://www.prodctodc.com/

Edit: After a quick look on eBay for something with UK shipping, this looks suitable: http://www.ebay.co.uk/itm/LM2596-DC-...item3f1b17ec1a

[DDY]

thermocouple in a dot of water (given time to stabilise) is generally accurate enough.

Water on a heatsink at over 100°C?

For a fix, you could use a buck converter. You can get basic pre-assembled boards dirt cheap on eBay, or another shop I found (ships from overseas though): http://www.prodctodc.com/

A buck convertor was my first idea but I couldn't find one in my box o' components, so I went with the next best thing

I've already requested a few samples for DC step down regulators* with the obvious downside of waiting about a month to get them and having to move my goal post a bit, from making a PoE kit only with parts already in my inventory to being a tight git for not wanting to spend a penny on this project

EDIT: * Buck regs of course!

[watercooled]

Well, if the water immediately vaporises it's probably safe to assume it's a bit warmer than you'd like it.

In the meantime, you could experiment with lowering the input voltage so it's just above the reg's dropout voltage under load, and dissipating the higher voltage when idle would be less of a problem. A 9v PSU might be a good place to start.

[DDY]

It didn't go so well with 12V either so I gave up and didn't try anything lower, but now that you mention it I'll give it a go on 9V and see how that works out.