RS485/RS422 Bus Termination

RS485/RS422 Termination and Biasing

RS485/RS422 links can show intermittent errors even when "120Ω at each end" is present. In practice, reflections and noise depend on where termination is placed, how long stubs are, whether the bus has proper biasing, and how fast the driver edges are.

Termination is only part of the story. Placement matters, multi-drop stubs can dominate, and biasing is often required to keep the bus in a defined idle state. A design that looks correct schematically can still create impedance discontinuities in the physical cable topology.

This guide explains when termination is needed (based on edge rate and cable delay), where to place it, how to choose biasing, and how to validate signal integrity with measurements.

Termination exists because cables are transmission lines. When an electrical signal propagates along a cable, it behaves as an electromagnetic wave traveling at a velocity determined by the cable's characteristic impedance, typically expressed as $Z_0 = \sqrt{L/C}$, where L is the inductance per unit length and C is the capacitance per unit length. For typical twisted-pair cables used in RS485/RS422 applications, this characteristic impedance usually falls between 100 and 120 ohms. When this traveling wave encounters an impedance discontinuity - such as an unterminated cable end - a portion of the signal energy reflects back toward the source. The reflection coefficient is given by $\Gamma = \frac{Z_L - Z_0}{Z_L + Z_0}$, where $Z_L$ is the load impedance. An open circuit ($Z_L = \infty$) produces a reflection coefficient of +1, meaning the entire signal reflects with the same polarity, while a short circuit ($Z_L = 0$) produces a reflection coefficient of -1, reflecting the signal with inverted polarity.

The decision of whether to implement termination depends primarily on the relationship between the signal edge rate (rise/fall time) and the cable propagation delay. A common rule of thumb is that a cable behaves like a transmission line when the one-way propagation delay is a significant fraction of the rise time (roughly on the order of 1/6 to 1/2 of $t_r$, depending on how conservative you want to be). For RS485/RS422 signals propagating at roughly 2×10⁸ m/s in typical twisted-pair cables, the delay is about 5 nanoseconds per meter. The important nuance is that bit rate is not the same as edge rate: a 1 Mbps system with a slew-rate-limited 100 ns edge may tolerate longer runs than a 1 Mbps system with 10 ns edges. This is why some designs work un-terminated on short cables, while others need termination even at modest data rates. In practice, termination decisions also depend on node count, stub lengths, receiver thresholds, and how much ringing you can tolerate.

Plain resistive termination is the simplest and most common approach for RS485/RS422 buses. This approach places a resistor matching the cable's characteristic impedance across the differential pair at each end of the bus. For a typical 120-ohm cable, 120-ohm termination resistors are installed between the A (or +) and B (or -) lines at the two physical endpoints of the network. This configuration ensures that signals reaching the cable ends encounter an impedance match, preventing reflections. Power dissipation in the termination resistors matters, especially in systems with continuous communication. The differential voltage depends on the driver and load - typically 2-5V. With a 3V differential signal, each 120-ohm termination resistor dissipates approximately $P = V^2/R = 9/120 = 75$ mW. At a worst-case 5V, it's about 208 mW. Size the resistor power rating accordingly.

Where you put the termination matters as much as the value. It goes at the actual physical endpoints of the bus - not necessarily the master device or the most important node. A common mistake is terminating at a node in the middle of the bus, which leaves unterminated stubs that cause reflections. In multi-drop setups, figure out where the two real endpoints are by tracing the cable routing. If you get this wrong, you'll see ringing on signal edges, inter-symbol interference at higher data rates, and intermittent failures that get worse with temperature.

AC termination saves power by blocking DC current through the termination. This technique places a capacitor in series with the termination resistor, creating a high-pass characteristic that presents the proper termination impedance for fast signal edges while appearing as an open circuit for DC signals. A common starting point is a resistor near the cable impedance (often ~120 ohms) with a capacitor chosen so the RC time constant is long compared to an individual bit time, but short enough to pass edges cleanly. For example, 120 ohms in series with 100 nF gives a corner frequency of $f_c = 1/(2\pi RC) \\approx 13.3$ kHz - but whether that is appropriate depends on your protocol and line coding. This approach can be valuable in battery-powered applications or systems with long idle periods. However, AC termination can cause baseline wander with certain data patterns and may not be suitable for protocols with long strings of identical bits.

Multi-drop networks introduce additional termination complexities beyond simple point-to-point connections. Each stub connection from the main bus to a node creates a potential reflection point, with the stub acting as an unterminated transmission line during signal transitions. A useful guideline is to keep stubs short compared to the rise time's spatial extent (often quoted as less than ~1/10 of $t_r \\cdot v$). For a 100-nanosecond rise time and ~2×10⁸ m/s velocity, that suggests stubs on the order of a couple meters or less; for 10 ns edges, it's more like a few tens of centimeters. This is why "keep stubs under ~0.3 m" is a decent default for fast-edge systems, but not a universal limit. Some designs implement distributed termination (multiple higher-value resistors that sum to the effective termination) - it can work, but it requires careful analysis, especially when nodes can be disconnected.

Bias resistors are technically separate from termination, but they work together. RS485 receivers need a minimum differential voltage (typically 200mV) to reliably determine the bus state. When all transmitters are in high-impedance mode - such as during bus turnaround or in multi-master systems between transmissions - the bus can float to an indeterminate state without proper biasing. Traditional fail-safe biasing uses pull-up and pull-down resistors to establish a defined idle state. A common configuration uses a 560-ohm pull-up on the A line to positive supply and a 560-ohm pull-down on the B line to ground. Combined with 120-ohm termination at each end, this creates a voltage divider that maintains approximately 250mV differential voltage in the idle state. The interaction between bias and termination resistors must be carefully analyzed: $V_{idle} = V_{supply} \times \frac{R_{term}||R_{term}}{R_{bias} + R_{term}||R_{term}}$, where $R_{term}||R_{term}$ represents the parallel combination of the two termination resistors.

The cable itself matters. While RS485/RS422 systems commonly use 120-ohm twisted-pair cables, actual impedance can vary from 100 to 150 ohms depending on construction, dielectric materials, and twist rate. Category 5e and Category 6 Ethernet cables, with their 100-ohm characteristic impedance, are increasingly used for RS485 applications, requiring corresponding 100-ohm termination. Shielded cables introduce additional considerations - the shield should be grounded at one point only to avoid ground loops, and the shield's effect on characteristic impedance must be considered. Cable attenuation increases with frequency, following approximately $\alpha = k\sqrt{f}$, where k depends on the cable construction. This frequency-dependent loss naturally dampens high-frequency reflections in long cables, sometimes reducing the criticality of precise termination.

Some newer transceivers include switchable termination so you don't waste power when the bus is idle - activate it during communication, disable it during sleep. Active termination circuits can also synthesize the right AC impedance while drawing less DC current, using op-amps or dedicated termination ICs. More complex, but useful for systems with widely varying operating modes or tight power budgets.

An oscilloscope will tell you if your termination is working. Connect differentially across the bus and look at edge quality. Clean transitions with minimal overshoot mean the termination is correct. Ringing that crosses receiver thresholds means under-termination. Rounded edges and weak amplitude mean over-termination (termination resistance too low). TDR measurements can directly measure cable impedance and find discontinuities along the bus. Eye diagrams at the receiver show how much margin you have. Do these measurements under worst-case conditions: maximum cable length, highest data rate, full node loading.

Temperature changes shift cable impedance and resistor values, which can push a marginal design into failure. Cable impedance and attenuation can shift with temperature due to dimensional changes and dielectric behavior, and termination/bias resistor values drift according to their temperature coefficients. In outdoor or industrial applications with wide temperature swings, those shifts can move a design from "barely works" to intermittent failures. Designing with adequate margins, using stable resistors where it matters, and validating performance across the full operating temperature range prevents unpleasant surprises. Some critical applications implement temperature-aware or switchable termination, but this level of sophistication is usually unnecessary if the physical layer has reasonable margin.

If you have intermittent errors that get worse at higher data rates, suspect termination first. Probe the bus at different points along the cable - ringing that crosses receiver thresholds points to under-termination, sluggish rise times point to over-termination. A network analyzer can measure bus impedance across frequency and reveal resonances from bad termination or long stubs. Common mistakes: terminating at intermediate nodes instead of endpoints, wrong resistor values, and forgetting that bias resistors load the bus too. A protocol analyzer helps separate physical-layer problems from software/protocol issues.

Modern transceivers have relaxed some traditional requirements. Many now include enhanced fail-safe that maintains valid output states down to 50mV differential, reducing bias resistor requirements. Slew-rate-limited drivers slow down edges to reduce EMI and reflections, which effectively extends the cable length that works without termination. Extended common-mode range and improved rejection help tolerate ground potential differences that would have caused failures in older parts. But these features complement proper termination - they don't replace it. A design that relies solely on transceiver enhancements without good termination fundamentals may work on the bench but fail in the field.

When termination is needed, place it at the physical cable endpoints and match it to the cable's characteristic impedance. Include appropriate biasing so the bus has a defined idle state (some modern transceivers have built-in failsafe, but many systems still benefit from external bias - designed to meet the receiver thresholds while not overloading the drivers). Keep stub lengths minimal and route the main bus in a daisy-chain topology rather than star configuration. Use quality resistors with appropriate power ratings and temperature coefficients. Verify signal quality with an oscilloscope during commissioning. Document the termination scheme clearly, as field modifications without understanding the termination design often create problems.

If you're dealing with RS485/RS422 challenges, whether that's intermittent errors, signal integrity issues, or designing a network that needs to work reliably over long distances, I'd be happy to take a look. I've designed and troubleshot differential communication systems across diverse industrial applications.

Sometimes the fix is straightforward once you understand what's actually happening on the bus. An oscilloscope measurement often reveals problems that aren't obvious from the symptom description. Reach out if you'd like to discuss your system, these problems usually have clear solutions once properly diagnosed.