Terminating Resistors and Bus Length: Why It Has to Be 120 Ω

The figure 120 Ω is not a rule of thumb, nor a manufacturer stipulation; it is set by the characteristic impedance of the bus cable. The shielded twisted-pair cable used on CAN, Profibus and RS-485-class buses typically has a characteristic impedance of the order of 120 Ω; only when loads equal to that characteristic impedance are fitted at both ends of the transmission line is the signal energy fully absorbed and reflection avoided. One is not enough, three over-matches, and fitting one in the middle does nothing — it must be one at each physical end of the segment. This article explains the principle, the measurement method and the differences between fieldbuses; for general wiring practice see A Practical CANopen and EtherCAT Guide.

Why a mismatch causes trouble: reflection and superposition

In the face of a high-speed signal a length of wire is no longer "two points at the same potential", but a transmission line. When the signal reaches the end of the line, if the termination impedance is not equal to the cable characteristic impedance, part of the energy is reflected back and superimposes on the following signal, causing overshoot, ringing and waveform distortion. The level the receiver reads at the sample instant may therefore deviate from the true value, appearing as occasional bit errors, CRC failures and a rising error-frame count.

This explains three common site phenomena:

• Over a short distance at low speed, it "works" without termination. The round-trip time of the signal is far shorter than one bit time, so the reflection has already decayed before the sample instant and the problem is masked. Raise the speed or extend the line and the fault appears at once — the "it was always fine before" class of fault mostly comes from this.
• The fault is intermittent. The result of reflection superposition depends on the data content and the edge instant; not every frame is wrong, so it appears as occasional lost frames rather than a complete outage, which is harder to troubleshoot than a complete outage.
• Lengthening one cable breaks it. Cable length changes the timing of the reflection; ringing that previously "just missed" the sample instant lands on the sample point after the extension.

How many, and where

There is only one rule: one at each of the two physical ends of the segment; none on any intermediate node.

There are three common mistakes. First, fitting only one: the other end still reflects, and errors appear at high speed. Second, fitting three or more: in parallel the total load impedance is too low, the driver's load capability is pulled down, signal amplitude is insufficient and communication is likewise abnormal — this "over-match" is more hidden than a missing terminator, because at low speed it often still works. Third, fitting one on an intermediate node: the ends remain unmatched, reflection continues, and an extra load has been added in the middle.

Another easily overlooked point: stubs must be as short as possible. A branch taken from the trunk to a device produces reflection as well, and termination matching cannot be done on the stub. The higher the baud rate, the shorter the allowed stub. Engineering practice should use in–out wiring (the bus cable enters the device plug and leaves from the same plug), rather than "picking" a wire off the trunk.

Measurement method: measure resistance with power off

The fastest way to judge whether terminating resistors are in place as a pair is to measure resistance with power off. Taking CAN as an example, measure the DC resistance between CAN_H and CAN_L with the power off:

A precaution: measure with power off; a live measurement is both inaccurate and a risk. Some devices have internal dividers or protection circuits that affect the reading; if in doubt, unplug one end and measure the cable side alone. The measurement idea on the Profibus side is the same, but note that its termination network includes bias resistors, so interpretation of the reading follows the device manual; for detailed troubleshooting see Profibus Diagnostics and Terminating Resistors in Practice.

Differing termination requirements by fieldbus

The core information of this table is: terminating resistors belong only to bus-type physical layers in which "several nodes share one differential pair". Ethernet-class protocols use point-to-point links; impedance matching is provided by the PHY transformers and the cable specification, and no external terminator is needed. Applying CAN experience to PROFINET, or conversely omitting termination on a CAN segment, are both common concept mix-ups. For the corresponding note on the PROFINET side see PROFINET Diagnostics and Topology: What's Really Going On.

The inverse relationship of bus length and baud rate

CAN and Profibus follow the same physical law: the higher the baud rate, the shorter the allowed segment. The reason is that the round-trip propagation time on the bus must fall within one bit time to complete arbitration or sampling; raising the rate shortens the bit time and the tolerable propagation delay shrinks with it. CANopen supports the baud-rate steps 10 / 20 / 50 / 125 / 250 / 500 / 800 / 1000 kbps precisely so that the engineer can choose a suitable rate for the actual cable length.

The correct engineering order is: measure the actual cable run first → back-calculate the highest usable rate for that length → then drop one step for margin. The usable length also depends on cable specification, node count and connector count, and should follow the master manual. For how to configure rate and address see CANopen Node Address and Baud Rate Configuration in Practice; for systematic troubleshooting of interference and earthing see Troubleshooting Fieldbus Interference: Grounding, Shielding, Topology. Related products: the 194 CANopen series and the 195 Profibus series.