Field Notes: The Sensor That Chattered Near the VFD
Week 3 of commissioning. An analog proximity sensor (0–10 V) measuring web tension on a converting line. The reading was jumping ±2 V even when the web was stationary. The VFD on the conveyor motor was two meters away. The signal cable ran parallel to the motor cable for about a meter. We spent two days chasing it.
Day one: chasing ghosts
We swapped the sensor. No change. We swapped the analog input card. No change. We checked the sensor’s supply voltage — 24 V, clean. The signal itself was a clean DC voltage at the sensor terminal, but at the PLC input it was noisy. The noise started when the VFD ran. When the VFD was off, the reading was stable. That told us something: the VFD was radiating or conducting noise into the signal cable.
We checked the grounding. The sensor was grounded to the machine frame. The VFD was grounded to a different point on the same frame. The ground loop between them was about 1.5 V. That’s the noise source — not magnetic radiation, but a ground potential difference between the sensor and the PLC.
Day two: the fix
Separate the cables. We rerouted the sensor cable away from the motor cable. They crossed at 90° instead of running parallel. The noise dropped by half. Not enough.
Shield the signal cable. The original cable was unshielded. We replaced it with a twisted-pair shielded cable (3×0.34 mm², shield drain wire grounded at the PLC end only). The shield grounded at one end — grounding both ends creates a ground loop. The noise dropped to ±0.2 V. Acceptable.
Add a ferrite bead. We clipped a ferrite ring (anti-interference bead) on the cable at the PLC end. The high-frequency switching noise from the VFD (20–50 MHz) was absorbed. The reading settled to ±0.05 V. Done.
What the cable routing looked like
| Condition | Noise at PLC input | Root cause |
|---|---|---|
| Parallel to motor cable, unshielded | ±2.0 V | Inductive coupling + ground loop |
| Crossed at 90°, unshielded | ±1.0 V | Ground loop only |
| Crossed at 90°, shielded, one-end ground | ±0.2 V | Residual high-frequency |
| Added ferrite bead | ±0.05 V | Clean |
The rules I now apply
Analog signal cables (4–20 mA, 0–10 V, thermocouple) need three things that digital sensor cables don’t:
- Twisted pair. The two signal wires twist together. External interference induces the same noise in both wires, and the differential input cancels it.
- Shield, grounded at one end. The shield blocks electric field coupling. Ground it at the PLC end only. Grounding both ends creates a ground loop — current flows through the shield, which becomes another noise source.
- Separation from motor cables. Keep signal cables at least 200 mm away from VFD output cables. If they must cross, cross at 90°. Never run parallel in the same cable carrier shelf.
The VFD itself
The VFD is the noise source because it switches the motor at 4–16 kHz. The switching edges are fast (dv/dt up to 5000 V/μs), which radiates broadband noise. For VFD installations, the motor cable should be shielded (or in metal conduit), the shield grounded at both ends (VFD and motor), and the VFD should have an EMC filter if the machine is in an office environment.
We didn’t change the VFD. We changed the signal cable. The sensor that chattered for two days now reads clean. The customer never knew how close we came to blaming the sensor for a grounding problem.
The shortcut: If the analog reading jumps when a VFD or motor runs, it’s not the sensor. It’s the cable. Shield it, twist it, separate it from power cables, ground the shield at one end. The ferrite bead is the last 10%.
