3
\$\begingroup\$

Where are the right places to put termination resistors in RS-485 networks? Specifically in full-duplex networks, I see app notes either use the termination at the driver node (Maxim (Page 15), TI (Page 2)), while others suggest otherwise (ADI). You don't really need to open these datasheets - the conundrum is essentially about these resistors:

enter image description here

Not whether you need them or not (I know in a few cases everything work without them), but are there any benefits to including them when not necessary, or maybe even drawbacks? Unfortunately I'm not too familiar with the transmission line theory, from some light reading (in particular, this TI "blog") it sounds like having a resistor on the driver side will help dampen any waves reflected from the receiver quicker, which should help with EMI. But what does everyone else think?

\$\endgroup\$
3
  • \$\begingroup\$ ADI appnote image and text below it about termination contradict each other. So the image is a conceptual diagram and does not reflect the text how RS-485 buses should be terminated. \$\endgroup\$
    – Justme
    Jul 2, 2021 at 0:42
  • \$\begingroup\$ @Justme are you talking about this line: "...RS-485 applications require termin-ation at the master node and the slave node furthest from the master."? In the "Parallel Termination" subsection, they go on to say "In a full duplex configuration only the master receiver and most remote slave receiver need to be terminated." Which seems consistent with the image? \$\endgroup\$
    – avg
    Jul 2, 2021 at 2:04
  • \$\begingroup\$ No, I mean in the "Parallel termination" it also says "When two or more drivers share a pair of wires, each end of the link has a termination resistor equal to the characteristic impedance of the cable" which again makes sense and agrees with other manufacturers. But then they have drawn a picture which does not adhere to this rule, and you also found text that contradicts with this rule they just said. \$\endgroup\$
    – Justme
    Jul 2, 2021 at 8:01

4 Answers 4

5
\$\begingroup\$

Short answer : The picture in ADI appnote contradicts the text in ADI appnote, so it basically means all sources agree. The drawing you posted shows a full-duplex multi-drop RS-485 bus where multiple transmitters are connected along the same bus - which is more than enough to determine that it's not a RS-422 bus and so at least the upper bus must have dual termination. And for the sake of symmetry, or to allow the placement of the host controller anywhere on the bus instead of just at the end of it, the lower bus should (or must) also have dual termination.

Longer answer : The difference is whether the whole network is built up to RS-422 or RS-485 standards, it's not just about chips and terminations but also the placement of the devices.

The RS-422 specs define that there must only be one single transmitter and up to ten receivers on a bus, and the driver must only be able to drive a single termination which at the end of the bus. This also means that the transmitter must also be at the beginning of the bus to prevent reflections. And usually the transmitter is always enabled. So the electrical specs do not allow terminating the bus at both ends as it will violate the specs.

The RS-485 specs define that a transmitter must be able to drive a bus with termination at both ends. This allows for placing the transmitter anywhere on the bus, and even multiple transmitters if only one is enabled at one given time. But as that is half-duplex, sometimes there are two buses to make it a full duplex configuration, just like on RS-422, but it allows for many devices on a full-duplex bus, instead of just two devices on a full-duplex RS-422 bus.

If RS-485 devices are used like RS-422 devices, i.e. transmitter always at one end of the bus and always enabled, then the second termination is not absolutely necessary, as it is not on RS-422 bus either. It's just that since RS-485 drivers can drive a doubly terminated bus, it makes no harm there, but double termination makes it incompatible with RS-422 chips. But with RS-485, having the double termination allows to place the transmitter anywhere on the bus, not just at the end. Double termination consumes more power obviously, but with RS-485, that can easily be handle by for example turning the driver off when not transmitting.

So the difference is which standard is used, and that limits how the buses are terminated, and how the devices can be connected to it.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks for your reply! Versatility of the transmitter placement is a really good point. That would probably explain the tendency of datasheets and app notes for to suggest to terminate at both ends, and adjust on a per-need basis (like if you had a single driver, for example). \$\endgroup\$
    – avg
    Jul 2, 2021 at 0:05
3
\$\begingroup\$

It depends on what the topology looks like and what data rates you need but I'd go with terminating each end. This needs to be done for each differential pair (because the whole idea of using terminating loads is to impedance match the cable and the load to reduce reflections and other non matching effects)

Data transmission lines should always be terminated and stubs should be as short as possible to avoid signal reflections on the line. Proper termination requires the matching of the terminating resistors, RT, to the characteristic impedance, Z0, of the transmission cable. Because the RS-485 standard recommends cables withZ0 = 120 W, the cable trunk is commonly terminated with 120-W resistors, one at each cable end (see Figure6-1, left).

enter image description here

Source: https://www.ti.com/lit/pdf/slla272

\$\endgroup\$
4
  • \$\begingroup\$ In the doc they also have a cool method for terminating to reduce common mode noise, it's worth a read \$\endgroup\$
    – Voltage Spike
    Jul 1, 2021 at 22:50
  • \$\begingroup\$ Thanks for sharing your thoughts on this - I tend to think the same. You've linked a really nice app note from TI (I had a link to it in the question, too! :). As far as big chip manufacturers go, terminating at the driver end of the line in a full-duplex system seems to be recommended, at least for versatility. \$\endgroup\$
    – avg
    Jul 2, 2021 at 0:04
  • \$\begingroup\$ You mean the split termination? Funny because on TI forums they say that they found it not so useful with RS-485, as it is more suitable for CAN bus. They don't mention why, but I suspect it's because drivers are allowed to drive common mode between -1 and +3 volts, and the common mode can vary 0.2V between logic 0 and 1 states. So different drivers need to charge the bus to their own common mode when they start to transmit and that current then flows in the common return wire. I'd be curious to know more if it is useful or not. \$\endgroup\$
    – Justme
    Jul 2, 2021 at 0:26
  • \$\begingroup\$ Yeah the split one. I've never tried it, maybe it's time \$\endgroup\$
    – Voltage Spike
    Jul 2, 2021 at 0:40
3
\$\begingroup\$

Terminations go at the physical ends of the lines, basically simple as, at least for RS485 where there can be multiple possible transmitters (RS422 is single terminated at the far end and the transmitter MUST be at one end of the bus).

Any line terminated at the far end into its characteristic impedance looks to a driver like an infinite line, no energy reflects.

If you extend this to a line with a driver tapped onto it in the middle (RS485), then providing both directions are terminated the driver sees two 'infinite' lines and again no reflections.

The key however, is that this is mostly not about what devices close to the ends of the line see, but about what a device half way along the line sees, and it is here that termination really starts to matter. If you consider an unterminated line, then that receiver at the half way point will see your original signal plus a delayed version coming back from the reflection, that is quite capable of screwing things up, while the receiver right at the end of the line will just see your transmitted signal at twice the expected amplitude (generally fine).

\$\endgroup\$
0
\$\begingroup\$

Assumptions may be incorrect.

There is no contraction on the interweb to requiring a matched differential impedance to the cable at the bus termination. Nor is there any contradiction on terminating both ends of a full duplex bus.

The common cable standard is a 120 ohm differential twisted pair but you can use other cables and change the loads to match it.

Matching makes a huge difference on ripple reduction although the best method is to bias a string with an RF shunt cap to earth ground splitting two 60 ohm resistors at both ends for CM noise attenuation.

The source impedance of each driver is often lower than the bus and have a complex non-linear relation with voltage bias but if there are no end reflections then there won't be source reflections to worry about at each source.

\$\endgroup\$
3
  • \$\begingroup\$ Thanks for your comments, Tony! I'm not sure I understand what you mean when you say there is no contradiction regarding termination on both ends of a full-duplex bus - at least the ADI datasheet explicitly urges to only terminate the receivers. But again, I think I'm misunderstanding what you mean. Also - unless that would take us too far off-topic - can you please add a little bit more information on the effect of the source impedance on termination? I've seen it mentioned before, tried finding more, but haven't had much luck so far. Thank you! \$\endgroup\$
    – avg
    Jul 1, 2021 at 23:47
  • \$\begingroup\$ By the way, I tried keeping the question generic, but in my specific case I'm working with a few third-party RS-485 devices that may or may not be using precision resistors. So I was thinking it's safe to assume there may be some reflections? \$\endgroup\$
    – avg
    Jul 1, 2021 at 23:53
  • \$\begingroup\$ @avg But ADI appnote image and text below it about termination contradict each other. So the image is a conceptual diagram and is not entirely correct about RS-485 termination, do read the text below it to find out. \$\endgroup\$
    – Justme
    Jul 2, 2021 at 0:39

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.