1
\$\begingroup\$

at my work there are a network of identical devices each of which control some 50 location-guidance light units. These devices or controllers are all linked together and then to a main server/computer via Rs485. The Light modules are linked Together in series via a bus/rail and to the controller pcb's one 12v output socket. I am investigating regular malfunction of these lights ( lights get stuck), and was suggested that there is interference with other high power machinery cables or internet cables that confuses the system or the light modules. When lights get stuck/freeze they either need to be unpluged from board to get workin again, or when this does not help all lights need disconnecting one from bus rail by one until a faulty light is found and system goess back to normal op. So by the looks of it the light modules go faulty after a certain time or a certainn amont of interference current. The interference is possible in the rs485 as its run with loads of other cables to the location of the controllers, but not likely in the 12 v bus i think. The rs485 has a shield but its not connected to any points as all connectors are plastic and shield is not connected to any pins either. Would it be a possible solution to replace for metal hood on at least one end? Also some suggested on different thread that shield should connect to signal ground as that is normally connected to mains earth throug h device plug. Is that why all these connectors made of plastic? And the person who installed didn't know they have to connect shield to pin5? Any help would be appreciated!

Edit: Light module array on bus bar is in parallel connection not in series.

Shield of rs485 seems to have continuity to earth just not connected in db9 shells.. although I have not checked each connector yet. Signal Grnd has continuity to PE and extraneous parts of rail. When things go wrong, not all controllers and their light arrays are affected. Rail/bus is powered by controller (26.6V) which in turn is mains powered.

\$\endgroup\$
12
  • \$\begingroup\$ put baluns on each cable to improve immunity (CM chokes) for the >100kHz band \$\endgroup\$ Commented Dec 21, 2016 at 9:11
  • \$\begingroup\$ If the RS-485 interface is not isolated (typically optically for data and galvanicly for the interface power) on one end at least then a very common problem with large networks is the limited common mode voltage tolerance which is only about 10 V. If the remote devices have no local power make sure they have no local potential connections and let them float and take all ground references from one central point. The shield should be connected at one end at least depending on conditions. The metal shell forms a fairly small window in shielding if the shield wire is terminated. \$\endgroup\$
    – KalleMP
    Commented Dec 21, 2016 at 9:22
  • \$\begingroup\$ The shield can be a problem if there are ground fault currents flowing, otherwise you got a nice antenna to couple radiated transient ingress glitches to latchup your devices. Check ground currents or add Balun this can be a (bad) reason for someone leaving one end open . CM impulse noise can easily become DM signals if balance is poor.ie >1% especially line surge currents or arcs , motor stop switches etc which may benefit from Y caps. The lack of shell shield is tiny compared with 100' antenna wire shield voltage \$\endgroup\$ Commented Dec 21, 2016 at 13:26
  • \$\begingroup\$ What physical distance does the network cover? Short (a few hundred feet) or long (thousands of feet)? \$\endgroup\$ Commented Dec 24, 2016 at 20:39
  • \$\begingroup\$ There should be a script on SE that filters posts for walls of text. Bonus points for lack of capitalization and punctuation. \$\endgroup\$
    – Bort
    Commented Dec 24, 2016 at 20:46

3 Answers 3

2
\$\begingroup\$

Although RS-485 is a robust interface, it can be brought to its knees with common-mode problems. Physical distance between nodes and not having a really solid ground that ties all the nodes together will make this worse.

Most of the RS-485 interface chips that I use can handle common-mode voltage in the range of -7 to +12 Volts.

If you haven a situation where some nodes might experience transients that exceed that range, you will have problems.

The easiest way to deal with this is to install isolators on the RS-485 network where needed. The challenge lies in identifying those locations.

One of the projects that I did many years ago was to provide isolation at every slave node. The network was a daisy-chains scheme with termination at the very end.

System worked well in a very electrically-noisy environment.

\$\endgroup\$
1
  • \$\begingroup\$ The location is given, as the problems occur in this specific network involving less then 20 drop points. I could perhaps isolate all of them? I had my manager order a bunch of ferrite beads to clip on to the rs485 cable going in and coming out of each controller. will that do , or is there any other isolation technique that you recommend? Is there a good way of monitoring the common voltage between pins of my rs485? \$\endgroup\$ Commented Dec 25, 2016 at 13:01
2
\$\begingroup\$

enter image description here Wherever your problem is coming from , you have two obvious solutions, reduce the offending egress source with snubbers or improve the immunity of the so-called shielded balanced differential signal.

How this is done depends entirely on your analytic test skills to define the above threat levels.

I would do this with 50MHz current probes and 200MHz FET diff probes. What you have may differ, but can still be done with ingenuity, such as connect unterminate shields with 50 Ohms and measure voltage drop with 2 perfectly balanced probes that flatline in A-B mode. ( no probe gnd clip using removed tipclip with pin& exposed gnd probe barrels only ) Then for current meter, use 10M probe with shorted tip using gnd clip wire as a loop around cable and then stimulate potential impulse event. Then measure diff. voltage at each end and measure amplitude and edge jitter for violations to safe operating area from CM input range.

Once you determine the source of the egress and ingress and measured it level exactly., the solution is easy. If you can't wait for an event. Create one with arcing handdrill motors on a parallel extension cable. Discharge 100V in a <1 microseconds on a loop wire from a charged ultralow 1uF ESR E-cap. That will safely give 1GV/s transient on a 1m antenna that will span the spectrum to a few hundred MHz

That's all I have to say for now.

\$\endgroup\$
11
  • \$\begingroup\$ Thanks again Tony, but I am a mechanical engineering graduate with passion for electrical and my test skills are not fully developed yet, so the stuff you wrote is a bit to complex!! Thanks for the effort though. Recommend some books or videos on how to do this please if you know any good ones ! \$\endgroup\$ Commented Dec 24, 2016 at 21:13
  • \$\begingroup\$ I'll add a picture and note where grounds are connected to shields and not connected at other end. Use same wire and Terminations are always at ends but not middle if any taps. \$\endgroup\$ Commented Dec 25, 2016 at 7:10
  • \$\begingroup\$ my topology is a bit different Tony, in two respects: \$\endgroup\$ Commented Dec 25, 2016 at 13:09
  • \$\begingroup\$ the cable used is not tp but a three core , and secondly the signal grnd seems to be connected to main earth. Don` t know why that is, as the two topology seems fairly similar in their design. Maybe its the fact that the drop points in my case are all mains powered? I thought about disconnecting their PE but I dread the possible negative consequences. \$\endgroup\$ Commented Dec 25, 2016 at 13:18
  • \$\begingroup\$ I suggest you rewire to the offending location using AWG24 2x shielded Twisted pairs for +/- and common as shown and earth shield grounded only at one end if you want to fix it. 3 core is poorly balanced and earthing incorrectly induces transients and steady state ground currents. Take heed. Connecting COM to shield defeats the purpose. Dont guess about AC power connections yet. \$\endgroup\$ Commented Dec 25, 2016 at 21:07
1
\$\begingroup\$

I thought it would be nice to give an update to you who tried to help. As I mentioned it was suggested to me that that the location suffered interference issues. Turns out no one investigated the power supply, which is running on its upper tolerance all the time ( 27 V DC instead of 24). This is due to that all products made in Europe expect 230 V AV and some variation, but in the UK it's constant 240 available. This combined with the constant low temperature at location should- I believe- put enough stress on the light modules to fail at an accelerated rate. We Lowered the voltage temporarily by installing a wirewound resistor in series with the input transformer and so far no issues. If this remains the case we will look into upgrading to a regulated SMPS.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.