· Technical Guides · 4 min read
Modbus RTU Survival Guide: Noise, Termination, and Grounding Diagnosis
RS485 is not black magic, it is physics. The definitive guide to troubleshooting noise issues, ground loops, and incorrect termination in industrial networks.
“Don’t cut corners.” — Rule #1 of industrial cabling.
RS485 is the workhorse of the industry. Theoretically robust, capable of running 1200 meters and withstanding hostile environments. But practically, it is the number one source of headaches for integrators who underestimate the physics of differential transmission lines.
If you are seeing random timeouts, CRC errors, or devices vanishing when a VFD starts up, stop swapping cables at random. Let’s fix it with engineering, not luck.
1. The “Two Wire” Myth (GND is mandatory)
The most common mistake is believing RS485 only needs A (+) and B (-).
The Reality: RS485 is a differential signal, meaning the receiver measures the voltage difference between A and B. But these voltages must exist within a Common Mode range (-7V to +12V) relative to the transceiver chip’s ground.
If you connect two machines powered by different transformers without linking their references (GND/Common), the “floating” potential difference between them can be 50V, 100V, or more. Consequence: You burn the RS485 port or saturate the input, resulting in zero communication.
Solution: Always run a 3-conductor + shield cable (A, B, GND, Shield). Connect the GND (common) at all nodes.
2. Termination: 120 Ohm is not optional
On short lab cables (< 5m), you might get away without termination. In a plant with 100m of cable and frequency noise, it’s suicidal.
When the signal reaches the end of an open cable, it bounces (reflection) and travels backward, colliding with incoming data.
The Rule:
- 120Ω resistor across A and B at the first device.
- 120Ω resistor across A and B at the last device.
- NEVER on intermediate devices.
3. Bias (Polarization): Silence is not Zero
When no one is talking on the bus (idle state), the line is floating (high impedance). If a 380V cable runs nearby, it can induce small voltages on A and B. If the receiver sees a random difference > 200mV, it will think it’s a “Start Bit”. The UART receives garbage (0x00, 0xFF) and throws a Framing Error.
The Failsafe/Bias Solution: We need to force a known state when there is silence.
- Pull-Up on A (to 5V/3.3V).
- Pull-Down on B (to GND).
Many modern transceivers (and gateways like Moxa or Advantech) have “Pull High/Low Resistors” configurable via dip-switch. Enable them at only one point in the network (usually at the Master).
4. Diagnosis with Oscilloscope (The Naked Truth)
Throw away the multimeter. To see noise, you need speed.
- Connect probe Ch1 to A and Ch2 to B (both relative to GND).
- Use the Math (Ch1 - Ch2) function. This is the real differential signal the chip sees.
What to look for:
- Clean Square Signal: The amplitude (A-B) must be greater than 2V (typically 5V).
- Rounded Edges: Excess capacitance (cable too long or poor quality) or lack of termination.
- Spikes (Overshoot): Lack of termination, signal is bouncing.
- Common Mode Noise: If Ch1 and Ch2 move together like crazy waves but the subtraction (Math) is stable, you have ground noise, but RS485 is rejecting it (to a point). If the subtraction deforms, the noise won.
5. Panic Checklist (When nothing works)
If you are on the plant floor at 3 AM and nothing works:
- Topology: Is it a Daisy Chain? If there are stars or long “stubs” (T-taps), cut them.
- Cable: Is it Twisted Pair? A parallel door-bell wire DOES NOT work. The twist cancels electromagnetic noise.
- GND: Did you connect the third reference wire?
- Shield: Is the shield connected to Earth at ONE POINT ONLY? (If you connect it at both ends, you create a ground loop that acts as an antenna).
- IDs and Baudrate: Are there no two slaves with ID #1? (Happens to the best of us).
Conclusion: Making Modbus work isn’t luck. It’s respecting impedances and references. Use proper cable, terminate the lines, and don’t cut corners with grounding.
