A 4-20 mA reading that slowly drifts, carries a steady 60 Hz hum, or jumps around for no reason the process can explain is often not an instrument problem at all - it is a wiring problem called a ground loop. Ground loops arise when a signal circuit is connected to earth in more than one place and those places are not at exactly the same voltage. This guide explains the loop mechanism specifically, distinct from general grounding practice: how a potential difference between grounds drives stray current through the signal path, the telltale symptoms, the cures, and how a ground loop shows up as flapping analog tags at a remote RTU.
Ground Loop in one line: A ground loop is an unintended current path that forms when a signal circuit is grounded at two or more points that sit at slightly different electrical potentials. The potential difference between those grounds drives a stray circulating current through the signal wiring, and because that current shares the same conductors as the measurement, it corrupts analog readings - producing offset, drift, and 60 Hz hum - and in severe cases can damage I/O. The cure is to break the loop, usually by grounding the circuit at a single point or by galvanically isolating it.
The mechanism starts with a fact people wish were not true: two ground points in a real facility are almost never at exactly the same voltage. Earth is not a perfect equipotential conductor. The building steel near a large motor, the ground at a field instrument a hundred meters away, and the ground bar in the control room can all sit at slightly different potentials because of the currents flowing through the earth and the impedance of the ground paths between them. That small voltage difference is harmless until you give it a wire to push current through.
A ground loop gives it exactly that. If a signal circuit - say a 4-20 mA loop - is connected to ground at the transmitter end and also, unintentionally, to ground at the I/O end, the two ground points and the signal wiring between them now form a closed loop. The potential difference between the grounds drives a current around that loop, and critically, part of that stray current flows through the same conductors that carry the measurement signal. The I/O module cannot tell the difference between the genuine loop current from the transmitter and the added stray current from the ground-potential difference; it measures the sum. The result is a reading that is offset or drifting by an amount that depends on a voltage difference that has nothing to do with the process.
Because the dominant source of those ground-potential differences in an industrial plant is the AC power system, the stray current often carries the power-line frequency, which is why ground loops so characteristically inject 60 Hz hum (or 50 Hz elsewhere) into a signal. The unwanted current can also vary as nearby loads switch on and off, so the corruption is not a fixed offset but a wandering, sometimes noisy error that tracks the electrical environment rather than the process being measured.
Ground loops announce themselves through a recognizable set of analog-input symptoms, and learning to spot the pattern saves enormous time. The signature complaint is an analog reading that is wrong in a way that does not correspond to the process: a level or temperature that reads a bit high or low with no physical reason, that drifts over time, or that carries a persistent low-level noise or hum. Filtering the input may smooth the noise but cannot remove a steady offset, so a reading that is quiet but simply incorrect can be a ground loop just as much as a noisy one.
A strong clue is that the error correlates with the electrical environment rather than the process. If the noise or offset changes when a large motor, VFD, or heater cycles, or appears at exactly the power-line frequency, suspicion should point immediately at a ground loop rather than the instrument. Another classic tell is that the reading changes when you lift one ground connection - if disconnecting a shield ground or an instrument ground makes the problem vanish or shift, you have essentially found the second ground point that was closing the loop.
Ground loops are not only a signal-quality nuisance; in severe cases they are a hardware hazard. A large potential difference - during a fault, a lightning event, or on a long run between buildings at very different ground potentials - can drive enough current through the signal wiring to damage the input circuitry of an I/O module. So a pattern of analog channels that read strangely and occasionally fail, especially on long runs to remote equipment, is worth investigating as a grounding problem rather than writing off as bad luck with modules.
The cure follows directly from the cause: if the problem is two grounds, the fix is to stop having two. Single-point grounding is the foundational principle - a given signal circuit and its shield should be earthed at exactly one place, so there is no second ground point to close a loop with. In practice that usually means grounding the shield at one end only, typically the control-system end, and leaving it floating at the other, and making sure a device is not inadvertently grounding the signal common where the design intends it to float. Careful, deliberate grounding at design time prevents most ground loops before they exist.
When single-point grounding is impractical - because the field device insists on grounding the signal, or the two ends genuinely must both be earthed for safety - the answer is galvanic isolation. An isolator or an isolated input channel breaks the direct electrical connection between the field side and the I/O side, passing the signal across a transformer or optical barrier while blocking any DC path for stray current to circulate. With no continuous conductive loop, the ground-potential difference has nothing to push current through, and the reading is clean regardless of how far apart the two ground potentials are. This is why isolated analog inputs and signal isolators are so common on long runs and between separately grounded areas.
In a cloud SCADA context, a ground loop typically first shows up not as a wiring diagram but as behavior on the screen: an analog tag from a remote RTU that flaps, drifts, or carries a rhythmic wobble that no operator action or real process change explains. Because a platform such as Merobix trends and time-stamps every value, that misbehavior is captured as history, and the pattern itself is diagnostic - a tag whose noise appears at the power-line frequency, or whose offset shifts whenever nearby equipment cycles, points to a ground loop rather than a failing sensor. Being able to see, from the office, that a specific channel at a specific site has been quietly wandering lets a technician arrive already suspecting the grounding and shield termination, and confirm the cure by watching the same tag settle down after the loop is broken - all without the traditional afternoon of chasing the loop with a meter to discover the second ground.
When a signal circuit is grounded at two points that sit at slightly different voltages, that potential difference drives a stray current around the loop formed by the wiring and the two grounds. Part of that current flows through the same conductors carrying the measurement, so the I/O module measures the real loop current plus the stray current and reports the sum. The result is an offset, drift, or noise on the reading that has nothing to do with the process.
Analog readings that are offset, drifting, or carry a steady 60 Hz (or 50 Hz) hum without a matching process change. A strong clue is that the error tracks the electrical environment - it changes when a large motor or VFD cycles, or sits at the power-line frequency - rather than the process. Another tell is that lifting one ground connection makes the problem vanish or shift, which points straight at the second ground closing the loop.
Break the loop. The primary cure is single-point grounding: earth each signal circuit and its shield at exactly one place so there is no second ground to circulate current. When both ends must be grounded or a field device insists on grounding the signal, use galvanic isolation - an isolator or isolated input channel that passes the signal across a barrier while blocking any DC path, so the ground-potential difference has nothing to push current through.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.