You could even power the system over a single pair. 1. Analog data from a typical modem IC can co-exist on a DC power feed. 2. If speed is not a problem (which it almost certainly isn't here) you could use DC signaling on the pair AND draw power off when the data signal is high. Example: This is harder to explain than to do!. I'll have to get an automated Protel to ASCII converter :-)) Power is fed from an eg 12 volt supply. Line resistance is low over this distance - say 10 ohms. Minimum acceptable voltage on line due to loading is, say, 7 volts allowing 2 volts headroom for a regulator at each station. Allowable drop is therefore (12-7) = 5 volts. If the 10 ohms was the ONLY resistance then you could draw 5v/10r = 0.5amp. So, place a, say, 20 ohm resistor in series with the power supply. Total R is now 30 ohm max. If you short the line at any point you get a maximum of 12/20 = 0.6amp and minimum of 12/(20+10)=0.4 amp. Each PIC station has: A voltage regulator. A transistor which can short the line. The ability to sense line voltage. If there are 10 stations drawing 10ma each you have 100ma continuous drain. Assume it is all at the far end of the line (worst case). The idle voltage is then V = 12 - (0.1 x (10+20))= 9 volts. If ANY PIC wants to transmit it shorts the line with its transistor and ALL other stations see the line voltage drop to a low value = logic 0. Power feed also stops during this "logic low" condition. Each PIC needs enough reservoir capacitor to hold up during the worst case string of logic lows. Suitable design can ensure that the line is always "high" long enough to keep stations alive. This system is wasteful of power (in this example 12x12/20 = 7 watts is dissipated in the 20 ohm feed resistor when the closest station is transmitting) but the power can be tailored to "just work" in your case and the feed station can probably be mains powered. This sort of system will work effortlessly at low baud rates (1200 baud) and can probably be made to work at much higher ones. The stations can be very low cost (PIC, regulator (possibly only a zener), small TO92 transistor (minimal dissipation off or on)) National Semiconductor once suggested something like this with their now ancient 14469 AART transceiver IC. The application notes may still be available somewhere. . DIAGRAM: Consumer Warning The following diagram made sense in WordPad in Windows 95 using Courier font. When copied to IE4 email it makes none. Depending on your browser YMMV. V+ ---- | R R <-- This R allows supply to be shorted for signalling. R | | <-- Data and Power line --> ___________________________________________________-> more PICS | | | __________ ____________ |C | | | | | |/ ----- | Regulator|---| PIC |--RRRR--| |__________| | | ___________| b|\ | C | |E | C | | _______________|______|_______|_________________|___________ ___ PIC will also need to sense line voltage (not shown) C is essential to keep PIC powered during data low -----Original Message----- From: Eric Smith To: PICLIST@MITVMA.MIT.EDU Date: Tuesday, 17 February 1998 09:30 Subject: Re: OT: PIC power supply >> This summer I'm going to be working on a project with a multi-drop >> network of PICs connected to a Linux box. The network cable will >> be about 75 long. Some of the devices will be out in a garden shed and >> I would like to power them off a Ni-Cad battery/Solar battery >> combination >> I'm wondering if a switcher setup would be more effecient (I'm worried >> about the Ni-Cad). Any comments, ideas, pointers (I love pointers :-} )? > >Why not power the nodes over the cable? I'd use RS-485 over two-pair >cable, and use the second pair for 12V DC power. Saves a lot of messing about >with batteries and solar cells. > >Cheers, >Eric >