Signal Isolators and Galvanic Isolation, Explained

An analog reading that drifts on a rainy day, jumps when a big motor starts, or loses calibration over weeks is often not the sensor. It's a ground loop pushing current through your signal wiring. A signal isolator removes the electrical path that current needs.

Engineers chase phantom sensor faults for weeks that turn out to be ground or common-mode problems. Most aren't sure when to isolate, which method to pick, or whether loop-powered will work on their loop. This guide covers what galvanic isolation is, how it breaks ground loops, the three isolation methods and their trade-offs, how to read an isolation-voltage rating, loop-powered versus externally powered isolators, and when you need one.

TL;DR: A signal isolator passes a measurement from a sensor to a receiver without a direct electrical connection between the two - the signal crosses a barrier as light, a magnetic field, or a capacitively-coupled AC carrier. That broken conductive path stops ground-loop current (I = E/R) and blocks large common-mode voltages. Optical isolation is simple and EMI-immune but degrades over time; capacitive is fast and high-voltage; magnetic gives the best analog accuracy (Analog Devices MT-071, 2026).

This piece pairs naturally with the industrial sensors guide for the wider signal-wiring picture, and with 4-20mA signals if you're already wiring the loop this isolator would protect.

What Is Galvanic Isolation?

Galvanic isolation means there is no direct conductive path between an isolator's input and output. The signal crosses a barrier as light, a magnetic field, or a capacitively-coupled AC carrier, so no current can flow directly between the two sides. Break the wire, keep the signal.

The barrier itself is small: a gap of air or plastic, a tiny transformer core, or a layer of silicon dioxide. Its job is singular. It carries information across without carrying electrons across. Neither side shares a common return path, a power rail, or a chassis ground with the other, because a shared ground is exactly what lets unwanted current sneak in.

Isolation ratings also split into basic and reinforced. Basic isolation gives one layer of separation, adequate for routine noise rejection. Reinforced isolation adds a second, independent layer, so a single fault in the barrier doesn't collapse the whole separation, the standard for anything touching mains voltage or a hazardous location. Most process-signal isolators (4-20mA, thermocouple, RTD) only need basic isolation; power-electronics and safety applications usually specify reinforced.

An isolator can separate signal only, power only, or both, depending on the job. A signal-only isolator still shares a power supply between input and output; a fully isolated (3-port) unit separates all three. That distinction resurfaces when comparing loop-powered and externally powered designs.

Citation capsule: Galvanic isolation removes the conductive path between an isolator's input and output entirely - the signal crosses the barrier as light, a magnetic field, or a capacitively-coupled AC carrier, never as a direct current (Analog Devices MT-071, 2026). Basic isolation gives one separation layer; reinforced isolation adds a second independent layer for mains-connected or hazardous-location equipment. The decision rule: if the two sides could ever sit at different ground potentials, or one side could see a fault voltage, isolate rather than share a ground.

What Is a Ground Loop, and Why Does It Wreck Signals?

A ground loop forms when two grounded devices sit at slightly different ground potentials and a signal cable connects them anyway. The difference drives a circulating current through the signal wiring by Ohm's law, I = E/R, and that current rides on top of the real measurement (Wilkerson, 2026).

Earth isn't the perfect zero-resistance reference the word "ground" implies. It behaves like a resistor between any two points where equipment is bonded to it, and every real installation grounds equipment at more than one location because plant safety codes require it. Ground two devices far apart, measure between those ground points with a long-lead voltmeter, and you'll usually find some voltage - millivolts, sometimes many volts (Wilkerson, 2026). Wherever there's a potential difference and a conductive path, current flows.

That circulating current doesn't announce itself. It adds error to the signal: sometimes a slow drift, sometimes a jump correlated with something unrelated to the process, like a motor starting or rain soaking the earth around a ground rod. Recalibrating the transmitter doesn't fix it, because the fault isn't in the transmitter. The ground potential itself moves, so the correction only holds until conditions shift again.

Common-mode voltage is the related, more dangerous cousin. It's the voltage the signal carries relative to the receiver's ground reference, and it can reach thousands of volts in some industrial settings. A motor-control current-sensing application, for example, can present hundreds of volts AC directly on an isolated input (Analog Devices MT-071, 2026). An isolator ignores that voltage entirely rather than absorbing it.

I once traced a level transmitter's slow, cyclical drift that no recalibration would hold - it always tracked back within a few days. The two ends of that loop were grounded at cabinets fed from different parts of the plant's earthing system, and the drift correlated loosely with rain, which changes soil and ground-rod resistance. Isolating the signal between the two cabinets, instead of chasing the transmitter's calibration again, killed the drift for good.

Citation capsule: A ground loop forms when two grounded devices sit at different ground potentials, and the earth's resistance between them turns that potential difference into a circulating current: I = E/R (Wilkerson, 2026). That current adds error directly to the signal and cannot be fixed by recalibration, because the ground potential itself drifts. Common-mode voltage, the voltage a signal carries relative to the receiver's ground, can reach thousands of volts in some industrial settings, and isolation is built specifically to reject it rather than absorb it (Analog Devices MT-071, 2026).

How Does a Signal Isolator Work?

An isolator recreates the signal on the far side of the barrier without a wire ever crossing it. For a 4-20mA loop, it modulates the DC current onto a high-frequency AC carrier, couples that carrier across the barrier (usually a small transformer or a pair of capacitors), then demodulates and filters it back into a clean, proportional DC output (Analog Devices MT-071, 2026).

That modulate-couple-demodulate sequence is the trick behind every isolation method, whatever the physical barrier. A transformer or capacitor passes AC and blocks DC, so the isolator turns a steady 4-20mA signal into something that behaves like AC just long enough to make the crossing, then turns it back into steady DC on the other side.

Because the barrier blocks DC, it also blocks the DC path a ground loop's circulating current would otherwise use. A large common-mode voltage on the input side never reaches the output electronics; it has nowhere to go. The same barrier attenuates fast transients riding on top of that common-mode voltage, for reasons covered in the isolation-voltage section below.

The isolated output rides on its own local ground reference, independent of whatever ground reference the input side sees. That's the point of the exercise: the receiver's electronics never have to share a ground with a transmitter that might sit in a different part of the plant, at a different ground potential, or exposed to a fault voltage the receiver was never designed to handle.

Citation capsule: A 4-20mA signal isolator modulates the input DC current onto a high-frequency AC carrier, couples that carrier across a transformer or capacitive barrier, then demodulates and filters it back into a clean proportional DC output on the far side (Analog Devices MT-071, 2026). Because a transformer or capacitor blocks DC by nature, the barrier also blocks the DC path a ground-loop current would otherwise use, and the isolated output settles onto its own independent ground reference.

Optical, Capacitive, or Magnetic - Which Isolation Method Should You Use?

Isolators cross the barrier one of three ways, and each trades accuracy, speed, size, and voltage differently. Optical isolation (an LED shining across a gap to a photodetector) is simple and immune to electromagnetic interference, but it degrades over time and handles analog poorly. Capacitive isolation is fast, compact, and rated for high voltage. Magnetic (transformer) isolation gives the best analog accuracy but takes up more board space.

Pick optical for simple digital signals on a budget, capacitive for speed and high voltage in a compact footprint, and magnetic when analog accuracy matters more than board space.

Optical (optocoupler)

An optocoupler shines an LED across a small gap onto a photodetector, so the input and output share no conductive connection - just light crossing an insulating gap (Wurth, 2026). It's simple, cheap, and immune to electromagnetic interference, since a magnetic field or a stray voltage can't influence a beam of light.

Its weak point is the LED. Current-transfer ratio (CTR), the ratio of output current to LED drive current, degrades over the part's life because the LED's light output falls with age: thermal stress and current density in the junction zone create crystal defects that reduce its light-generating efficiency (Wurth, 2026). Under typical operating conditions, testing holds CTR degradation to no more than about 5% over 25 years (Wurth, 2026), a reasonable lifetime if you run the LED conservatively, but a real design constraint if you push it hard on current or temperature.

That drift barely matters for a digital ON/OFF signal, where a threshold has margin to spare. It's a real problem for a continuous analog value, where every percent of CTR drift becomes measurement error. That's why optocouplers dominate digital isolation and status signals, and show up far less often in precision analog isolators (Monolithic Power, 2026).

Capacitive and magnetic

A capacitive isolator couples the signal across a pair of tiny on-chip capacitors, typically built from silicon dioxide (SiO2), which has a breakdown field strength around 800 V per micron of dielectric thickness (Monolithic Power, 2026). Modern capacitive isolators built on that dielectric reach roughly 5.7 kVrms withstand and 8 kVpk transient ratings, as sourced examples from a reinforced-isolation part family (Texas Instruments, 2026). They are fast, compact, and comfortably high-voltage, though the specific kV figures vary by part.

A magnetic (transformer) isolator couples the signal through a small wound or planar transformer instead. It's the method with the best analog performance. Transformer-coupled isolation amplifiers reach 12 to 16-bit analog accuracy, and a high-speed transformer-isolated amplifier can hit as tight as 0.005% maximum linearity error, as a sourced example from a high-performance part (Analog Devices MT-071, 2026). Transformers are also naturally robust against transients, but the coil and core take up more board space than a capacitor pair, so magnetic isolators tend to run physically larger.

Citation capsule: Capacitive isolators built on SiO2 dielectric, with a breakdown field strength around 800 V/µm, reach roughly 5.7 kVrms withstand and 8 kVpk transient ratings as sourced examples from modern reinforced-isolation parts (Texas Instruments, 2026; Monolithic Power, 2026). Transformer-coupled magnetic isolators trade some of that voltage headroom for accuracy, reaching 12 to 16-bit performance and linearity error as tight as 0.005% on a high-speed example part (Analog Devices MT-071, 2026). Decision rule: pick optical for simple digital signals on a budget, capacitive for speed and high voltage in a compact footprint, and magnetic when analog accuracy matters more than board space.

How Do You Read Isolation-Voltage Ratings?

A datasheet lists several isolation voltages, and they mean different things. The continuous working voltage is what the barrier survives every day for its rated life; a higher withstand voltage is a short test the barrier must pass once; an even higher transient peak is what it tolerates for microseconds; and a surge-test level is the highest brief spike the part is proven against.

On a sourced reinforced-isolation example part, the working voltage (V_IOWM) is rated around 1.5 kVrms continuous for over 40 years. The withstand voltage (V_ISO) is a 60-second test at roughly 5.7 kVrms. The transient voltage (V_IOTM) is rated around 8 kVpk, and the part's actual surge test passes above 12.8 kV (Texas Instruments reinforced-isolation, 2026). Those four numbers describe the same barrier under four very different stress durations, from decades down to microseconds. Don't mistake the biggest number on the page for what the part survives continuously; it doesn't.

Basic versus reinforced isolation is a separate distinction, covered earlier: reinforced adds a second independent barrier layer, tested and rated as such, rather than just a bigger single gap. Creepage and clearance, the surface distance and through-air distance between the two sides of the barrier, set the working and withstand numbers on the datasheet; a wider physical gap has more insulation to break down.

When specifying an isolator, match the rating to what the application can throw at it. A 4-20mA loop sharing a cabinet with clean, low-voltage instrumentation rarely needs the highest surge rating on the shelf. A signal isolator next to a VFD, a welder, or mains switching gear is a different story, and that's exactly where the higher withstand and transient numbers earn their keep.

Citation capsule: A reinforced-isolation datasheet separates four voltage figures by test duration: a continuous working voltage around 1.5 kVrms, a 60-second withstand test around 5.7 kVrms, a transient peak around 8 kVpk, and an actual surge-test pass level above 12.8 kV (Texas Instruments reinforced-isolation, 2026). Decision rule: match the rating to the surrounding electrical environment - clean cabinets need the working-voltage number, while a panel near VFDs or switching gear needs the withstand and transient numbers to have real headroom.

Loop-Powered vs Externally Powered - What's the Difference?

Analog isolators come two ways. A loop-powered (2-wire) isolator draws its operating power directly from the 4-20mA loop, so there's no separate supply to install. An externally powered (3-wire or 4-wire) isolator has its own dedicated supply and typically delivers higher accuracy.

A loop-powered isolator introduces its own voltage drop into the loop, on the order of 6V for a typical device (Chenzhu, 2026). That drop has to fit inside your existing power budget. Take a worked example: a 24V loop supply with roughly 300 ohms of loop resistance at 20mA (6V of drop). Add an isolator dropping another 6V, and the transmitter sees about 12V (Chenzhu, 2026). If the transmitter needs 10V minimum, that loop works with headroom to spare; if it needs 14V, the loop-powered isolator starves it.

I've seen that failure once: a loop-powered isolator added to an existing 24V loop that had almost no spare voltage to begin with, and the transmitter stopped reporting correctly once the isolator's drop ate into the little headroom left. Raising the loop supply voltage solved it in minutes, but only after the budget math got done properly on paper.

Isolator wiring also splits into 2-port and 3-port designs. A 2-port isolator isolates the input and output signal circuits from each other, but its power circuit is not isolated from one of the two sides; the power supply is typically referenced to the output. A 3-port isolator isolates the input, output, and power circuits from one another completely, so the power supply can connect anywhere, input side, output side, or an independent ground, without compromising isolation (Analog Devices MT-071, 2026).

Externally powered isolators, freed from drawing operating current off the signal itself, generally deliver better accuracy and drive heavier output loads than a loop-powered unit can. That's the trade: loop-powered simplifies wiring (no extra supply, one less cable run to a field cabinet) at the cost of a voltage-drop budget to manage and somewhat looser accuracy. For guidance on which output topology you're wiring into on the receiving end, see wiring the sensor output.

Citation capsule: A loop-powered isolator typically drops around 6V of the loop's available voltage to run its own electronics, so the supply voltage minus that drop, minus wire and burden resistance, must still exceed the transmitter's minimum operating voltage (Chenzhu, 2026). A 2-port isolator leaves its power circuit referenced to one signal side; a 3-port isolator isolates input, output, and power completely, so the supply can connect anywhere without breaking isolation (Analog Devices MT-071, 2026). Decision rule: run the voltage-drop math before specifying loop-powered, and choose 3-port whenever the power source can't share a ground with either signal side.

When Do You Actually Need a Signal Isolator?

You need an isolator when grounded devices sit far apart, the plant is electrically noisy with VFDs and motors nearby, a sensor floats at a high common-mode voltage, or you're protecting an expensive PLC or DCS input from damage.

Isolate when two or more of these are true:

• A long cable run connects devices grounded in different areas of the plant
• The system has multiple, separate ground references rather than one
• VFDs, motors, welders, or other high-current switching loads sit nearby
• The sensor or signal source floats at, or can spike to, a high common-mode voltage
• You're protecting costly downstream equipment or a safety-critical loop from a fault on the field side

Isolation earns its keep on transients as much as on steady-state noise. An electrical fast transient (EFT) event that would otherwise drive current through a signal path for roughly 100 nanoseconds gets confined to as little as 6 nanoseconds once isolation is in place. With the voltage excursion held to a few volts instead of a much larger swing (Texas Instruments, 2026), that's the practical payoff of the barrier discussed earlier: it doesn't just block steady ground-loop current, it also shrinks the energy and duration of a fast electrical event before it reaches sensitive electronics.

Intrinsic safety barriers are a related but distinct device, and the two get confused. An IS barrier limits energy for hazardous-area classification, while a galvanic isolator breaks the conductive path for ground-loop and common-mode protection. Some devices do both, but they solve different problems, and a datasheet that claims one shouldn't be assumed to cover the other.

If none of the checklist items apply, a short, single-cabinet run with one ground reference and no nearby high-current switching, a direct wired connection is often adequate, and adding isolation only adds cost and a small amount of measurement error for no real benefit. Isolation is a targeted fix for a specific electrical condition, not a default practice to bolt onto every loop.

Citation capsule: Isolation shrinks transient events dramatically - a non-isolated system can carry an electrical fast transient current pulse for roughly 100 nanoseconds, while an isolated system confines the same event to as little as 6 nanoseconds with the voltage excursion held to a few volts (Texas Instruments, 2026). Decision rule: isolate when long cable runs, multiple grounds, nearby VFDs or motors, a floating high-common-mode sensor, or protection of costly equipment apply - two or more of those together make the case; skip it on a short, single-ground, quiet-plant run.

How Does Isolation Affect Analog Accuracy?

Isolation is not free. It can add a small amount of error and noise to a signal, so match the isolation method to the accuracy the application needs rather than defaulting to whatever's on the shelf.

Magnetic (transformer) isolators give the tightest analog accuracy of the three methods, reaching down to about 0.005% linearity error on a high-speed sourced example part, the natural choice when a measurement's precision matters (Analog Devices MT-071, 2026). Loop-powered units, whatever the barrier technology inside them, generally trade some of that accuracy for the simplicity of drawing power off the loop.

A small RC filter, on the order of 150 ohms plus 1 nF, is common at an isolator's output to smooth the residual carrier ripple left over from the modulate-demodulate process, cleaning up the last high-frequency noise before the signal reaches a receiving instrument. Where possible, scale and precondition the signal before it crosses the barrier. Amplifying a weak millivolt-level source like an RTD or thermocouple beforehand, much like RTD signal wiring already has to account for lead resistance, improves the accuracy that survives the isolation stage. For current sensing specifically, a galvanically isolated approach is already built into the physics of Hall current sensors, which measure a conductor's magnetic field with no direct electrical connection.

Bandwidth is the last factor to check. A fast-changing signal, like a vibration or high-speed process variable, needs an isolator rated for the frequencies actually present, not just the DC accuracy spec on the datasheet's front page. Push a wideband signal through an isolator sized for slow 4-20mA loops, and it rolls off or distorts exactly the fast content you needed to preserve.

Frequently Asked Questions

What is a signal isolator?

A signal isolator passes a measurement from a sensor to a receiver with no conductive path between input and output - the signal crosses a barrier as light, a magnetic field, or a capacitively-coupled AC carrier (Analog Devices MT-071, 2026).

What is the difference between galvanic isolation and a ground loop?

A ground loop is the fault - two grounds at different potentials driving a circulating current through your signal wiring (I = E/R). Galvanic isolation is the fix - it removes the conductive path that current needs (Wilkerson, 2026).

What types of signal isolators are there?

Three barrier types: optical (an LED and photodetector), capacitive (a signal coupled across tiny SiO2 capacitors), and magnetic (a small transformer). Each trades analog accuracy, speed, and isolation voltage differently (Monolithic Power, 2026).

Do I need a signal isolator on my 4-20mA loop?

Isolate when grounded devices sit far apart, the plant has VFDs or motors nearby, a sensor floats at a high common-mode voltage, or you are protecting an expensive PLC or DCS input from damage (Analog Devices MT-071, 2026).

Loop-powered vs externally powered isolator - what is the difference?

A loop-powered isolator draws its operating power from the 4-20mA loop itself and drops a few volts, e.g. roughly 6V (Chenzhu, 2026). An externally powered (3-port) isolator has its own supply and typically delivers higher accuracy.

Conclusion

Galvanic isolation removes the conductive path between a sensor and its receiver, which kills a ground loop's circulating current and blocks large common-mode voltages. The method you pick, optical, capacitive, or magnetic, trades analog accuracy, speed, and isolation voltage differently, so match the barrier to the job rather than grabbing whatever's in the parts bin.

Read the isolation-voltage ratings as four separate numbers, not one big kV figure, and budget the voltage drop on a loop-powered unit before you assume it will work on your existing loop. Isolate when grounds differ, cable runs are long, or the plant is electrically noisy, and skip it when none of that applies.

For the wider signal-wiring picture, start with the complete industrial sensors guide, and revisit 4-20mA signals for the loops these isolators are built to protect.