RS-485 bench troubleshooting: separate termination, stubs and idle bias first

Draw the electrical path first

Draw both ends of the trunk, every branch point, the transceiver interfaces and built-in device termination on one diagram. When half-duplex communication shares one signal pair, the usual arrangement is a bus trunk with short stubs. Termination matches the transmission line’s characteristic impedance; it does not mean putting one resistor at every device. ADI’s AN-960 explains the basic relationship between parallel termination at both ends and receiver fail-safe protection. AN-960

For common 120 Ω cable, 120 Ω termination is often used at each of the two electrical ends. Count device “120 Ω” switches, on-board resistors and external resistors, rather than just the two resistors outside the terminals. For a different cable specification, recheck its differential characteristic impedance and the complete circuit.

With power off and other parallel paths isolated, two ideal 120 Ω resistors in parallel give 60 Ω. This is a configuration check using Ohm’s law, not a substitute for measuring characteristic impedance. Bias networks, transceivers and protection devices also affect actual meter readings. Record the circuit conditions before interpreting a reading, rather than immediately attributing a value other than 60 Ω to a harness fault.

Use different comparisons for three symptoms

These associations help plan checks; waveform shape alone does not determine the root cause. Confirm that addressing, frame format, transmission-direction control and message records are unchanged before comparing electrical conditions. A check with only one node transmitting should also remain separate from an idle check in which all nodes have released the bus.

Stub constraints depend on rise time

Section 6 of TI’s The RS-485 Design Guide, SLLA272D, gives a stub screening equation:

L ≤ tᵣ × v × c ÷ 10

Here L is stub length, tᵣ is the driver output’s 10% to 90% rise time, v is the cable propagation velocity as a fraction of the speed of light, and c is approximately 3×10⁸ meters/second. This equation applies to the stub conditions in that design guide; it is not proof of suitability for every device. SLLA272D, Section 6

Assuming a cable propagation velocity of 0.67c and keeping other conditions the same gives:

These are demonstration calculations, using no measured edge from a specific device. If a one-meter stub develops problems after a transceiver replacement, comparing only the devices’ maximum baud rates is insufficient. Reducing the UART baud rate does not necessarily change the rise time of each device transition. Next, obtain edge specifications or waveforms for the actual transceiver under the relevant supply and load conditions, and check the actual stub path.

Omitting termination is a separate decision

In SSZTB23A, revised in March 2026, TI discusses how some links may not need termination when the network round-trip propagation time is very short relative to a single bit time. That depends on conditions for reflection decay and receiver sampling. This addresses a different problem from the preceding section’s edge-based stub screening. A single “low speed” label cannot replace both judgments. SSZTB23A, pages 1–3

Continuing to use 0.67c as an assumption, a 100-meter line has a round-trip propagation time of approximately 1 microsecond. At 9600 bits/second, one bit lasts approximately 104 microseconds; at 115200 bits/second, approximately 8.68 microseconds. Changing the rate changes the relative time available for reflections to decay. However, these figures omit the nodes, connectors, device output characteristics and actual sampling point, so they do not directly justify removing termination.

Where low power consumption matters, other termination methods can also be compared. Parallel termination loads the driver. AC termination combines a resistor and capacitor to reduce steady-state current, but introduces RC effects. Its suitability depends on the link’s waveforms, data rate and component design, rather than standby current alone.

Idle bias does not repair transmission-line reflections

Termination addresses transmission-line matching; idle bias establishes a defined receiver state when no driver is transmitting. AN-960 distinguishes conventional receiver thresholds, external bias and receivers with true fail-safe functionality. A device’s support for open-circuit protection cannot be extended into a claim that it needs no external design measures under all idle, short-circuit and noise conditions. AN-960, Fail-Safe Biasing

For an actual check, first establish the device’s receiver thresholds, hysteresis and fail-safe specifications. Put every driver into its non-transmitting state, then observe idle differential voltage, receiver output and frame-error records. If external bias needs adjustment, change only that item and account for termination at both ends and the loading of every node. Do not copy resistor values from another board.

Check the common-mode range separately. An adequate difference between the two signal voltages does not establish that their voltages relative to the receiver’s reference ground remain within the permitted range. For different supply domains, ground-potential differences or isolation arrangements, use the equipment circuit and transceiver documentation to address the issue; adding termination does not resolve it. TI RS-485 design questions, Section 4

A traceable bench comparison

On a low-voltage bench isolated from the site’s power systems, prepare the two end devices, cable with known specifications, defined transmit and receive message logs, and measurement tools suitable for the interface and its common-mode range. Follow the equipment instructions to confirm supply, probe connections and reference points first. Turn power off before changing resistors or wiring. Without suitable differential measurement conditions, do not try connecting an ordinary probe ground clip to an unknown potential.

1. Save the baseline. Record the transceiver model, cable model and length, stub length, termination and bias, supplies, baud rate, message format and driver-enable timing. Save the current repeatable errors or waveforms.
2. Choose one variable. Use the symptom table above to change only termination configuration, stub length or idle bias. Keep all other parameters at the baseline.
3. Retest within the same observation window. Define the number of transmitted frames or duration, waveform observation points, error definition and ending conditions in advance. Save successful frames, error frames, timeouts and corresponding waveforms; “communication normal” is not enough.
4. Restore the baseline for a check. Where conditions allow, restore the original configuration and see whether the symptom returns with it. If it does not, retain other possible explanations, such as temperature, connection state or an intermittent event.
5. Decide the next step. If a change affects the symptom, expand validation to the actual range of nodes, supplies and environments. If changes are inconsistent, continue checking direction control, contacts, power supply or measurement method rather than adding more changes at once.

The attachments provide a blank record sheet, an example with planned values and field instructions. Received frames, error frames, voltages, measured rise time and conclusions are all blank in the example; users should enter actual results. The sheet generates no automatic pass decision. Acceptance values come from the specific system requirements, device conditions and agreed test design.

Recheck the original bench conclusion whenever a device or cable is replaced, the trunk or stubs change, nodes are added or removed, termination switches change, or the supply reference changes. If heating, abnormal current, device-limit violations or uncertainty about the measurement reference appears, stop using that configuration and check the circuit and measurement conditions first.

Related application note: Single-pair Ethernet and power bench evaluation. Both require clearly defined data paths and supply conditions, but RS-485 termination, thresholds and test conclusions cannot be transferred to an Ethernet interface.

English summary

This note separates three RS-485 troubleshooting questions: bus-end termination, stub reflections and idle-bus receiver behavior. It provides a rise-time calculation, a separate propagation-time example, and a repeatable bench comparison with one changed variable. The worked values are assumptions, and the attached measurement fields are blank. MTTJ has not performed a hardware test or established product qualification limits.

Sources checked: September 30, 2026. Principal materials are TI SLLA272D (revised May 2021), SSZTB23A (revised March 2026), ADI AN-960 and the relevant sections of TI’s RS-485 design questions. This note was researched and edited with AI assistance. MTTJ prepared the demonstration calculations and bench steps and has not performed physical testing.