End of Line Termination Concept
Most of the buses we use are specified with end of line termination in mind. In the field these terminations are not always present, especially in short runs. Do we need them?
Here’s an analogy I like that illustrates why we need end of line terminations: a bus behaves like a rope held between two people. When one person wants to talk they whip pulses down the rope towards their partner. The other person feels these pulses and figures out what the message means from them.
Those pulses don’t go away, though – they’ll reflect back at the person who whipped it down the line. It’ll keep reflecting back and forth, a little weaker each time as losses eat it up. These reflections are confusing to both people, who waste time trying to decode this false message over and over again.
A bus terminated on both ends is like a rope fixed on either side with springs. When the pulses reach the end, the person there can feel them and then the pulses are absorbed by the springs. If the springs are appropriately sized, the pulses will be “fully dampened” and there will be no reflection back.
For a visual look at this problem, check out the visual results of termination as seen by an oscilloscope.
RS-485/RS-422 End of Line Termination
Since we’re typically using RS-485, the end of line termination is selected based on something called the characteristic impedance of the cable you’re using. This is something the manufacturer stamps on the box the cable comes in, so it’s no secret. Typically we use something around 108-120 ohms. You don’t need to use exactly that value, but try to be close and don’t use resistors that are much less than it! Some other bus electrical standards will use more complicated devices as their end of line terminations.
Do you always need an end of line termination? Not really, no. It depends on two things: baud rate and bus length. The lower your baud rate, the less reflection will matter. The greater the distance, the more reflection will matter.
High speed buses much more sensitive because everything happens at a much lower time scale – a single bit transferred at 38400 baud lasts only 26 microseconds. For a 9600 baud bit, it’s 104 microseconds. A voltage spike in the wrong place that lasts the length of one bit length can cause communication errors.
Longer buses are more sensitive to reflection because any reflection will tend to dissipate after several trips back and forth down the bus; for a longer bus, that dissipation takes longer. If it’s long enough, controllers could send new messages on the bus while the older reflections are still bouncing around.
Getting Away With Lazy
You can reach a ballpark estimate of when end of line termination is needed can by calculating two things: the propagation delay and the bit pulse width.
Propagation delay is how long a bit takes to go down the cable. Since a voltage pulse travels at a fraction the speed of light (let’s say it’s 66% the speed of light) the propagation delay is 0.66 * c / (length of cable).
The bit pulse width is how long a single bit pulse lasts in time, which we addressed earlier. It’s as simple as 1 / (baud rate). Since baud rate is in Hertz, which is cycles per second, 1/(baud rate) gives us the seconds per cycles (bits).
We now have two time values we can compare. A general guideline is that you can avoid using end of line termination when your propagation delay is much less than the bit pulse time. I.E.: pulses travel so fast across the bus that a reflection will dissipate before it interferes with the interpretation of future bits.
It’s not pure science and could still depend on other factors. Generally speaking, there’s not much reason to cheap out on a few 108 ohm resistors. Bus problems are very labour intensive to troubleshoot.
If you want to be cheap, here’s a table of rough estimates I’ve made of maximum lengths for different baudrates before termination is required:
|Max Length for PD << BP (m)
|Max Length for PD << PD (ft)