The Sensor That Randomly Triggered: EMI, Grounding, and Shielding
A proximity sensor (inductive, PNP output) on a packaging machine. It triggered randomly — a false signal about once an hour. The PLC counted it as a part present. The machine double-stacked boxes. The sensor wasn’t broken. The signal was being corrupted by electromagnetic interference from the nearby heat sealer. This is how to trace and fix random sensor triggers.
The diagnostic
We disconnected the sensor and put a spare in its place. Same random triggers. The sensor wasn’t the problem. We swapped the cable. Same. We moved the sensor 1 m away from the heat sealer. The triggers stopped. The heat sealer’s heating element (AC, switched at high current) was radiating noise that coupled into the sensor’s cable.
The EMI sources
| Source | Frequency | How to mitigate |
|---|---|---|
| VFD / servo drive | 4-16 kHz switching | Shielded cable, separate routing |
| Contactors / motor starters | Switching transients | Snubber across contacts |
| Heat sealer / heater | 50/60 Hz AC | Separate routing, shielded sensor cable |
| Welding (RF) | High frequency | Faraday shield, fiber optic |
| Radio (walkie-talkies) | VHF/UHF | Keep antennas away from cables |
What we did
1. Shielded cable. The original sensor cable was unshielded. We replaced it with a shielded cable. The shield grounded at the PLC end only (grounding both ends creates a ground loop). The random triggers dropped from 1 per hour to 1 per day.
2. Separate the routing. We rerouted the sensor cable away from the heat sealer’s power cable. They used to run parallel in the same cable tray. Now they’re in separate trays, 200 mm apart, crossing at 90°. The triggers stopped entirely.
3. Add a capacitor across the sensor output. We added a 100 nF capacitor between the sensor signal and common at the PLC input. The capacitor filters short spikes (nanosecond transients) that the PLC counts as signals. The sensor’s real signal (a steady DC level) passes through. The noise doesn’t.
The PLC input filtering
Many PLCs have input filtering (a small RC filter on each digital input). We enabled the filter on that input. The PLC now requires the signal to be steady for 10 ms before it counts it. A 1 ms noise spike is ignored. This is the easiest fix — it’s a PLC setting, not a wiring change.
The sensor type
An inductive proximity sensor (PNP/NPN) is a digital output. It’s either ON or OFF. Noise on the cable can make it appear to switch. A sensor with a solid-state output and a shielded cable is more immune. We also verified the sensor’s switching frequency — if it’s set too sensitive (high frequency response), it picks up noise. We set it to the slowest response that still meets the cycle time.
The EMI rule: Random sensor triggers are almost always EMI, not a bad sensor. Use shielded cable (ground at one end). Separate signal cables from power cables. Enable PLC input filtering. Add a capacitor across the input. If the trigger moves when you move the sensor, it’s a shielding problem.
