Shield grounding: one end, both ends, and when the rule flips
2026-06-02 · wiring
Two engineers will argue about this for an hour and both be right, because they are describing different frequency ranges and neither says so out loud.
The low-frequency case
Ground a shield at both ends of a long cable run and you have created a loop between two points of the earthing system. Those two points are rarely at the same potential — a few volts is normal in a plant with big drives — and the resulting current flows down the shield. That current couples into the signal pair and shows up as mains-frequency noise on your measurement. Hence the classic rule: ground the shield at one end only, at the controller side, for analogue signal cable.
The high-frequency case
Above roughly a megahertz the same shield stops behaving like a wire and starts behaving like a transmission line. A shield grounded at one end is an antenna with a free end; it picks up radiated noise and has nowhere to put it. Here you want the shield bonded at both ends, with a low-inductance connection — a 360° gland or clamp, not a pigtail, because a 50 mm pigtail is several tens of nanohenries and undoes the point of the exercise.
How to hold both truths
The usual reconciliation: bond at both ends, but make one of those bonds through a small capacitor, so it is a short circuit at radio frequency and an open circuit at 50 Hz. Many drive manufacturers build this into the terminal arrangement. Where you cannot do that, decide by what dominates: an RTD run alongside a VFD cable is a high-frequency problem and wants both ends; a thermocouple crossing a quiet yard is a low-frequency problem and wants one.
Measure before believing anyone
Clamp an AC current probe around the shield. If you read tens of milliamps at mains frequency, your both-ends bond is costing you. If you read nothing there but the signal is noisy when a nearby drive runs, your single-ended shield is not doing its job. Ten minutes with a clamp meter settles arguments that outlast several project meetings.
Correction, 2026-06-09: the original version said a pigtail adds "a few nanohenries". Measured on the bench it is closer to 20–30 nH for 50 mm of conductor. Corrected above.