RS-485 Bus Design: Termination, Fail-Safe Bias, Stubs, and Validation

Define the Physical Bus

Record cable differential impedance and propagation velocity, total trunk length, each stub length, node count, transceiver output rise and fall times, supply range, ground potential difference, and expected surge environment. RS-485 specifies an electrical interface, not packet framing or application recovery. The Serial Interface Guide compares several buses; this guide addresses the installed RS-485 physical layer in depth.

Terminate the Ends, Not Every Node

For a linear trunk, place one differential resistor near each physical end when termination is needed. Select each resistor close to cable Z0. Two 120 Ω end resistors create a 60 Ω differential load from a driver's perspective before receiver and bias-network loads. Adding a third full terminator drops that idealized parallel load to 40 Ω. Use the RS-485 Termination Calculator to check match, effective load, differential current, power, and split-termination values.

For Z0 = RT = 120 Ω, the ideal local reflection coefficient is zero. With Vdiff = 2 V, one resistor dissipates 4/120 = 33.3 mW; two dissipate 66.7 mW total, while the driver sees roughly 60 Ω. Power rating must cover the worst sustained differential voltage, temperature, tolerance, fault scenario, and any required derating. A split termination uses two approximately RT/2 resistors; any midpoint capacitor or common-mode reference needs an explicit noise and safety design.

Treat Edge Time and Stubs as a Timing Problem

Reflections are launched by edges, not by the baud-rate label alone. A stub creates a branch discontinuity; its returning edge can interfere with the main waveform. Use cable velocity to estimate one-way and round-trip delay, then compare with the fastest specified transition and the receiver sampling window. No universal stub-length limit follows from one bit rate. Keep intermediate connections short and avoid a star or ring when designing a simple linear multidrop trunk. The Communication Latency Calculator estimates propagation delay but does not model reflections or topology.

For a 2 m stub and vprop = 2×10⁸ m/s, the stub round trip is 20 ns. That may be material for a fast edge even if the bit interval is much longer. Examine the installed eye or waveform at both ends and at the worst stub, rather than deciding from this ratio alone.

Establish a Defined Idle State

When every driver is disabled, terminators tend to bring the differential voltage toward zero. A legacy receiver may have an indeterminate output there; external fail-safe pull-up and pull-down resistors can create a defined differential state. Modern receivers may guarantee an idle, open, or shorted-line output without external bias, but these are device-specific guarantees. Read the exact threshold, hysteresis, common-mode, and fault conditions rather than assuming all receivers behave alike.

For a simplified bias circuit with a pull-up Rup, pull-down Rdown, and equivalent differential termination Req, the nominal idle difference is VsupplyReq/(Rup + Req + Rdown). For 5 V, 680 Ω, 680 Ω, and two 120 Ω terminators (Req = 60 Ω), the idealized result is about 0.211 V. That is only a nominal example: supply tolerance, resistor tolerance, receiver input loads, leakage, common-mode limits, and required noise margin must be included. The termination calculator does not size this bias network, and the example is not a universal resistor recommendation.

Common-Mode, Ground, and Protection

Differential reception rejects some common noise; it does not make unlimited ground offset harmless. Check the transceiver's guaranteed input common-mode range over the full installation, including ground shifts, cable shield scheme, powered-off nodes, and surge events. Galvanic isolation and protection devices may be necessary where those conditions exceed device ratings. Placement and grounding of TVS protection, termination, and shield connections affect both surge paths and signal integrity. Follow the chosen component datasheets and applicable installation requirements; a resistor-only calculator cannot establish electrical safety or EMC compliance.

Verification Matrix

Probe with a suitable differential method and short measurement loop. A long ground lead can invent ringing. Capture the far-end eye or worst patterns at temperature and voltage corners. If the network uses a higher-layer protocol such as Modbus RTU, validate framing and retries separately from this physical-layer review.

Common Mistakes

• Terminating every intermediate node.
• Using 120 Ω without checking cable Z0.
• Deciding termination from baud rate but ignoring edge time.
• Using a star layout with long reflective branches.
• Assuming a calculated reflection coefficient predicts the whole bus eye.
• Treating bias resistors as free of driver loading.
• Assuming every receiver needs external bias, or none does.
• Using a nominal 200 mV idle result without tolerance margin.
• Ignoring powered-off nodes and common-mode range.
• Treating differential signaling as isolation.
• Confusing propagation delay with protocol latency.
• Claiming electrical safety or EMC compliance from a termination calculation.

Primary Engineering Sources

For topology and fail-safe boundaries, consult Analog Devices AN-960, Analog Devices AN-727, and TI RS-485 implementation guidance. Verify the selected cable and transceiver against their own current datasheets; examples here are not a substitute for their guaranteed ratings.