If you want to get straight on with building, skip this bit!
The basic principles of Power over Ethernet (PoE) are straighforward: at one end of the cable, either near the network hub or built into it, is an 'injector' which takes DC power and the network signals, and combines them onto one cable. At the far end of the cable is a 'splitter' which does the reverse, taking the cable input and producing a DC power output and a network connection.
There are two main types of PoE systems. 'Active' systems have circuitry at both ends so that the injector only turns power on when a compatible device is detected on the cable. This is the sort you'll find on commercial network switches with PoE built in. 'Passive' systems are commonly found in cheaper applications (e.g. home CCTV) - here the power is applied all the time and no detection circuitry is used.
A passive system uses spare wires in a CAT-5 network cable to supply power - a 10- or 100-Mbit Ethernet connection only needs 4 of the 8 available wires to operate, leaving 4 free for power. Gigabit Ethernet connections need all 8, so don't work with passive PoE. Fortunately just about all Gigabit devices will fall back to 100-Mbit operation if only 4 wires are connected.
This project uses a simple passive PoE arrangement. The only refinement is to use a higher DC supply voltage (12-24V) than the Pi needs, and to step it down to 5V at the Pi end. The advantages to doing this are:
- supplying 5V directly at one end of the cable means the Pi would get less than 5V, due to the resistance in the cable. This would make the Pi less reliable as the cable got longer.
- the current in the cable is less, meaning that less electrical power is wasted.
- the step-down regulator provides (some) protection for the Pi if the cable is mis-connected.