(To preface, I will probably just run 4 wires, with RS485 on 2 and power on the others, but this came to my head, and I started thinking about it.)
I was thinking of a simple way to send simple low speed data (9600 baud?), from a single source out to multiple devices using just the two wires also used to power the devices. No return data is needed.
Situation: 12 or 24 volt system of perhaps 40 'nodes' spread over a 200 foot x 200 foot area, each 'node' is a simple microprocessor with it's own software address, running a few RGB LEDs. Harsh environment, only has to last 2 weeks (Burning Man). The data would be simple commands, with no more that say 8 bytes per command (with address, checksum etc.) Data would be sent from a central controller that simply coordinates the colors and timing etc of the LEDs. Devices will be added in a 'star' pattern, but with multiple nodes on each line. Each node will run a processor at 5 volts, and perhaps 300 mA of 12 volts for the LEDs at each node.
My thought was if I use a power supply between 16 to 24 volts (24 volts is easy to find) and at each node used a 12 volt regulator for the LEDs, and a 5 volt regulator for the processor, that would prevent voltage drop problems with long lines. If, at the power supply, I put 3 or more high current diodes in series, then my 24 volt supply becomes more like 22 or less. By putting an FET parallel to the diodes, I could now switch the power line from 22 to 24 volts. With the appropriate driver, a microprocessor (5 volt) serial TX line could 'modulate' the voltage.
At each node, I could have a single supply comparator on the 5 volt supply, with resistor voltage divider off of the 22 volt modulated line as the input to the comparator, with it's output feeding into the processor RX line for serial data. In an ideal situation, a simple low pass filter and voltage divider could provide the reference for the comparator, in actuality, the 22 volt supply will probably vary a bit, maybe not enough to worry about, and perhaps I just use a lot of diodes to make more of a voltage difference, so that the reference is guaranteed to fall inside the voltage drop range.
Any thoughts? Does this seem doable. Since the data is just broadcast out, with no need to get a response, this seemed the simplest way to go.