Power-over-differential

[Oznog]

I'm trying to get around the limitations of the number of wires in my cable. I need to have some fairly high speed differential serial comm as well as supply 3.3V @ 125mA (max). The best cable choice by far has only 3 wires, including the ground.

Well, I did have an idea there. Diff is such that one is high while one is low. If I put a could of schottky diodes off the lines, one will always be high and feed a Vdd cap for the receiving device except for the transition period.

But 125mA is more than say a RS485 driver can put out. One thought there is connect the differential lines to the source Vdd through a pair of inductors, so the Vdd provides most of the DC current and the driver just does the AC switching.

Is there any sense to this, or is it just whacked?

EDIT: Well, of course it's wacked, when the line is pulled low the driver will have to pull against the 125mA or whatever current in the inductor. Duh. Is there a way to do what I'm thinking of? I mean in theory it's possible anyways but how to actually implement it and deal with the noise issues is tough.

 

[blueroomelectronics]

Interesting idea, I've got a super low power thermostat design that may just work with. A latching relay is the only high power device and that will only have to cycle once every 10 minutes.

 

[Oznog]

Well that Dallas 1-wire protocol allows for devices powered by the signal line, however, that's much less power and FAR less bandwidth. The 1-wire does work with multiple devices on the same line, BTW.

Using a relay might be a bit too much current, however, just using a MOSFET- or if the signal's AC, a mosfet pair used as a passgate or a packaged Solid State Relay- will get the current down to where this is easy.

 

[blueroomelectronics]

A reasonably sized capacitor should allow the relay to function, I chose a latching relay specifically for its low power requirements.

 

[mneary]

Look into "power over Ethernet". Not quite the same, but the principle is that if your signal is fast enough and is DC-balanced (8b-10b) you can send the current without disturbing the data.

 

[Oznog]

Actually I looked at S/PDIF and that used biphase mark code which is DC-balanced. It would be possible to provide power and signal just by isolating the AC from the DC.

However, PICs can't produce biphase mark code without bit-banging and there's not enough processing power at this bandwidth for that sort of bit-banging.

 

[mneary]

You could build your own "RS422" drivers. A pair of PNPs (or P-channel) transistors, each driven in turn. To strip the power, use a pair of Schottky diodes. To recover the data use a voltage divider to reduce the voltage on each line to 1/2 the power supply.