RS-485 Communications

[agross144]

Because my current application uses 485 comms for an industrial application I figured I'd start here and move the thread if I have to. My current setup uses a standard 485 communications network through a multiplexer (refereed to as a repeater) to the PLC to save comms channels. At one of the plants where this system is setup they were adding new equipment and arc welding way to far from their ground point and ended up killing the system. As it turns out they blew a comms channel, the repeater, and several nodes. So as me and my crew are putting the pieces back together we started having some very strange 485 problems. After replacing all the equipment on the damaged runs including the cabling we started having problems talking with nodes in the middle of the new run. Figuring it had to be a faulty repeater I replaced that again with the same results nodes at the beginning of the chain would communicate while the ones in the middle would not. After spending countless hours slogging through boards and other possible problems we got exasperated and started trying all sorts of weird stuff to get the network back up. As it turns out we tied all the node strings together and ran a single pair of wires back (as in a star topology) and IT FREAKING WORKED?! I was always told that you could not operate a 485 network in that configuration or without a terminating resistor, but somehow it's working. Now for the second oddball part, after taking all the bad equipment back to the shop to test it I find that apparently the number of the node makes a difference as to weather or not it works. I have no idea what is going on but if anyone has some information about why a star topology would work for RS485 or why a node number would cause comms problems I would appreciate the insight.

 

[danw2]

Well, with a star topology, each node is a "spur" and reflections from spurs are known to be problematic. Does that mean it won't work at all? Nope, it depends on conditions. The reference note cited at the end shows a star topology with a repeater on page 16.

Attempted baud rates, spur length, total link length, grounding, isolation, common mode, biasing and terminating or lack thereof all contribute to whether a system works or not.

Your spur topology might work at 9600 but fail miserably at 19.2k, or the first 5 nodes work OK until the 6th node is attached, and then none work, due to biasing or loading. Or what works on the bench fails when the much higher common mode in the field saturates the drivers.

The biggest fallacy is 2 wire RS-485, where ground reference is needed somewhere, but many systems ignore 3rd wire grouding altogether. Heck, the 2 nodes worked on the bench as a Beta prototype, didn't it? Then ship it ! !

Out of curiosity, do you know if the repeater isolated the damage from one segment to the other segment side of the repeater? Most commercial repeaters are isolated, side-to-side, but I'm sure someone sells a non-isolated repeater, there's always a market for cheap.

The best working reference I know of is here:

http://sss-mag.com/pdf/485appno.pdf

National Instruments' web note on biasing is pretty good, too:

http://www.ni.com/support/serial/resinfo.htm

 

[agross144]

The repeater we use does not isolate which is why I think we had so much damage, I'm currently looking at some isolating repeaters from rs485.com that optically couple the tx and rx lines. Thanks fro the info as I've been testing different setups I've been slowly picking up what you've been describing. # of nodes, correctly sizing termination resistors seems to be the biggest factor as I've got 10 nodes across 1000ft of cable and to get that network bumping I had to go back to my old box of resistors and start at 120 ohms and work my way down. I've also noticed that is takes a minimum of 200mV to get a successful com signal transmit. by throwing my meter across the tx and rx pins I can see a spike and if the spike doesn't get over 200mV the network will not work until I get a resistor to pull the network.

 

[danw2]

Here's a really interesting app note on field troubleshooting RS-485 issues:

http://www.robustdc.com/library/an006.pdf