1
\$\begingroup\$

I have an application where we use RS485 to communicate messages to a series of LED Displays on the bus. Our first version of the board seemed to work fine with no problems, however we have made a new revision that is using a different RS485 Transceiver and now some displays do not receive their messages reliably. However not all displays on the bus are being affected. Only one or two are being affected. We never had to use any biasing resistors or terminating resistors with the first version so I can't see that as the problem for this case. The only thing I can see that might be causing the problem is the change in RS485 Transceiver that was used.

The first 485 Transceiver we used was a ST485C, however, the new board is using a SN65HVD485E 485 transceiver. Is it possible that the change in chip has caused a difference in performance? I know there is only one way to find out, but I just wanted to know if this was a reasonable causation?

We also use Cat5 UTP cable and use the following pin configuration. This is T568a standard.

  1. RS485 B
  2. RS485 B
  3. RS485 A
  4. GND
  5. GND
  6. RS485 A
  7. VCC
  8. VCC

I know that the above wiring doesn't follow RS485 standard wiring, but it worked with the last Revision and I can't see how this could be the reason for the current problems.

\$\endgroup\$
6
  • \$\begingroup\$ using the twisted wire pairs, like you did, is the correct way to go when using ethernet drop cables \$\endgroup\$
    – jsotola
    Commented May 11, 2018 at 1:05
  • \$\begingroup\$ one thing that leaps out when i googled the two chips are the words Half-Duplex \$\endgroup\$
    – jsotola
    Commented May 11, 2018 at 1:07
  • 1
    \$\begingroup\$ @jsotola - but he's not using the twisted pairs. 1-2 is a pair and should have A-B on it, twisted together (or 3-6, 4-5, 7-8). Instead he's bundled each pair together and effectively just has 2 non-twisted wires. May as well have just used some 2-wire zip-cord. \$\endgroup\$
    – brhans
    Commented May 11, 2018 at 2:14
  • \$\begingroup\$ I was told that the twisted pairs themselves are twisted within the cable. Could that twisting be enough to be good enough for RS485? i.e. the green pair and the orange pair are also (to less of a degree) twisted about each other. \$\endgroup\$
    – Jadboy20
    Commented May 11, 2018 at 2:25
  • 4
    \$\begingroup\$ @Jadboy20 Not in a useful way they, also this is far more about conductor to conductor spacing then twist (Spacing impacts impedance, twisting is about noise reduction). How fast are you running this link, and over what distance? RS485 run at low speed over shortish distances can be fine with no termination and random cable, but that is always a bit of a crapshoot (And devices near the end of the line working, but ones in the middle not is classic for unterminated lines). Also check the common mode range at the far end of the cable, I have been stuffed by ground cable resistance before now. \$\endgroup\$
    – Dan Mills
    Commented May 11, 2018 at 11:43

2 Answers 2

1
\$\begingroup\$

From looking at both datasheets the SN65HVD485E draws 50 times the current on its data input DI, or about 100 uA. The ST485C draws about 2 uA on its DI pin with a 24 K load, so there is a major difference.

In addition the changes in input loading can cause reflections of the signal from end to end, such that at some places it is strong while in others it is weak due to 'running into itself' out of phase.

Since both IC's are supposed to work as advertised I would make sure they are all the same. The mix of 485 types maybe more of a problem then having just one type for all points of use.

The load resistor is still needed at each end of the line.

\$\endgroup\$
1
  • \$\begingroup\$ We changed the chip and that did the trick. I know the issue is caused by reflections in the line though so I need to make sure that my EOL devices are terminated. Thank you. \$\endgroup\$
    – Jadboy20
    Commented May 25, 2018 at 3:55
1
\$\begingroup\$

You would need terminating and biasing resistors at each end of the transmission line. Tha cable has to have a characteristics impedance of 120 ohms, twisted pair. Ethernet cable has twisted pairs, but 100 ohm impedance, a little mismatch but not critical. It depends on cable length if the cross section of Cat5 is enough.

I usually use shielded twisted pair with 0.5mm^2 cross section for approx. 100m, like LiYCY TP 4x2x0.5.

The RS485 transceivers are all compliant and compatible, I guess you were just lucky the first time, but you would need to improve it anyway if you want a stable and robust RS485 network.

schematic

simulate this circuit – Schematic created using CircuitLab

$$R_{T_{eq}}=\dfrac{R_{term}\cdot (R_{bias_{up}} + R_{bias_{dw}}) }{R_{term}+R_{bias_{up}}+R_{bias_{dw}}}$$

For given example the equivalent Rt is 112 ohms. You can tune resistors to match used cable. R_bias shall not be less than 560 ohm.

\$\endgroup\$
2
  • \$\begingroup\$ The cable's characteristic impedance does not need to be 120 ohms (or any specific value). The critical factor is that your terminating resistors match the characteristic impedance of the cable. Also don't forget to account for the biasing resistors when calculating the value of the terminator (they're effectively in series with each other and in parallel with the terminator). \$\endgroup\$
    – brhans
    Commented May 11, 2018 at 16:16
  • \$\begingroup\$ @brhans you are correct, I have added some additional info. \$\endgroup\$ Commented May 11, 2018 at 18:04

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.