Lesson 2/1712%
MODULE 02 — WHY ISOLATION

Why Optocouplers are Used

Module 1 showed what an optocoupler is. This module answers why you would put one in a design — and makes the answer concrete. A 3.3 V microcontroller wired directly to a 400 V motor controller is one fault away from destruction, one ground offset away from data corruption, and one surge away from hurting somebody. We work through safety, ground loops, noise immunity, fault protection, mixed-voltage interfacing, and the regulatory framework — with simulators, calculators, nine worked design examples, and a decision tree that tells you when an optocoupler is (and is not) the right tool.

Learning Objectives

Explain why galvanic isolation is required
Identify safety hazards in electronic systems
Eliminate ground loops with optical isolation
Design interfaces between voltage domains
Reduce EMI and noise coupling using isolation
Improve system reliability and fault containment
Apply IEC/UL/VDE isolation standards
Decide when an optocoupler is the right choice

Consider the most tempting shortcut in industrial electronics: a 3.3 V microcontroller connected directly to the control input of a 400 V motor drive. It works on the bench. Now ask the questions a design review would ask:

What happens during a fault?

An IGBT shorts, a winding flashes over, or a wiring error lands 400 V (or a rectified 565 V DC bus) on the shared signal line. Every component between the drive and the MCU becomes a fuse — including the MCU, its power rail, and everything else on that rail.

What if the industrial ground rises 100 V?

Motor return currents and ground faults routinely lift a machine's local earth tens of volts above the control room's. With a direct connection, that difference drives current through your signal wiring and shifts every logic threshold — data corruption first, burned traces second.

What about a lightning strike?

An indirect strike hundreds of meters away induces kilovolt transients on long cable runs and raises ground potential across the site. A direct electrical path delivers that energy straight into the lowest-impedance (and most expensive) silicon it can find.

Why is direct connection dangerous?

Because a conductor cannot distinguish signal from fault. Whatever voltage, noise, or surge exists on one side is offered to the other, at the full available fault energy, forever.

Interactive Fault Simulator

Toggle the isolation barrier, then press Inject 400 V fault to see where the fault current goes. Without isolation the fault current (red) pours through the signal line into the MCU. With an optocoupler in the path, the fault stops at the barrier — only light crosses.

System healthy. Signal flowing normally.

Shock hazards in numbers

Isolation requirements exist because the human body is a resistor of roughly 1–2 kΩ (hand-to-hand, wet skin much lower) connected to ground. Safety standards treat voltages above roughly 42.4 V peak AC or 60 V DC as hazardous to touch. What current does to a person:

Current through body (50/60 Hz) Physiological effect
0.5–2 mA Perception threshold, tingling
5–10 mA Painful shock; muscle control affected
10–20 mA "Let-go" threshold exceeded — cannot release the conductor
30–50 mA Respiratory difficulty; possible asphyxia
50–100 mA+ Ventricular fibrillation risk — potentially lethal
≥ 10 μA (patient-connected) Medical limit: current applied directly to the heart can fibrillate at microamp levels

Galvanic isolation protects on every axis at once: user protection (no conductive path from hazardous voltage to touchable parts), operator and patient safety (leakage currents limited to microamps), equipment protection (faults cannot propagate into the control side), functional safety (E-stop and interlock signals remain trustworthy in fault conditions), and fire prevention (fault energy cannot dump into wiring never sized for it). It is required wherever high-voltage supplies, industrial machinery, grid-connected equipment, laboratory instruments, or medical devices meet human beings.

Fault Current Path Animation

A fault connects the mains side to the signal interface while a person touches the equipment. Toggle the barrier and watch the current path.

Without isolation → fault current flows through the user

Medical equipment isolation

Medical devices are the most demanding isolation environment because the "user" may be an anesthetized patient with electrodes bypassing the skin's protective resistance — ECG front ends, patient monitors, defibrillators, infusion pumps, and ultrasound probes all make deliberate electrical contact. IEC 60601-1 therefore limits patient leakage current to microamps and defines protection in terms of MOPP (Means of Patient Protection) and MOOP (Means of Operator Protection). Each "means" is one independent barrier; a mains-to-patient path needs two MOPP — which a single reinforced-insulation optocoupler (VDE 0884-17 certified, 8 mm creepage) can provide as 2×MOPP in one component.

Requirement (230 V mains) 1×MOPP 2×MOPP 1×MOOP
Insulation Basic Reinforced / double Basic
Test voltage 1.5 kVrms 4 kVrms 1.5 kVrms
Creepage 4 mm 8 mm 2.5 mm

Industrial automation safety

In a factory cabinet, isolation is the containment strategy: PLC I/O racks isolate every field signal so that a ground fault on one sensor loop, a welded contactor, or a miswired 230 V line onto a 24 V terminal damages one input channel — not the backplane, not the CPU, and not the emergency-shutdown chain. Motor drives isolate gate drivers and feedback so a power-stage explosion (they happen) leaves the controller alive to report it. Robotics and safety systems isolate E-stop loops so the stop command survives the very faults it exists to handle.

Power supply isolation

Every offline AC/DC supply — flyback, forward, or LLC — has a transformer carrying power across the mains barrier. But regulation needs information to flow backwards: the secondary-side output voltage must steer the primary-side PWM controller. That return path is the classic optocoupler application. A TL431 shunt regulator on the secondary compares the output against its reference and drives the optocoupler LED; the phototransistor pulls the controller's feedback pin on the primary. Power crosses in the transformer; the error signal crosses as light.

Animated Flyback Feedback Loop

Watch the regulation loop: output drifts high → TL431 conducts harder → LED brightens → feedback pin pulled → duty cycle shrinks → output returns. The dashed line is the isolation barrier — power crosses in the transformer, information crosses in the optocoupler.

Why ground loops occur

"Ground" is a convenient fiction. Two earth connections 100 m apart are joined by soil, structural steel, and cable shields — all of which carry return currents from motors, welders, and rectifiers. The result is a ground loop: connect two grounded devices with a signal cable, and the volts of difference between their grounds drive amps of circulating current through your cable shield and signal return. Causes stack up in industrial environments: ground reference differences, long cable runs, multiple grounding points, earth potential differences after faults, and magnetically induced loop EMF from nearby high-current conductors.

Ground Loop Simulator

Two cabinets share a signal cable. Set the ground potential difference and cable resistance, then break the loop with an optocoupler and watch the circulating current disappear.

Ground potential difference 5 V
Loop resistance (cable + shield) 2 Ω
Circulating loop current
2.5 A
Noise on signal return
Link status
Before isolation
Ground loop current
Noise superimposed on signal
Communication failure
After isolation
Loop broken — no conductive path
No circulating current
Reliable communication

The same loop-breaking logic prevents circulating currents wherever ground potentials legitimately differ: battery banks whose negative rails float relative to chassis, motor drives whose "power ground" bounces with every PWM edge, multi-cabinet control systems, and DC fast chargers where the vehicle and the charger have separate earthing. Break the metallic path, and each system keeps its own reference in peace.

Signal Comparison: Isolated vs Non-Isolated

A clean 0–5 V data signal is sent between the cabinets while common-mode noise (ground bounce + 50 Hz hum + switching hash) is present. Drag the noise slider and compare what the receiver sees.

Injected common-mode noise 3 V

Common-mode voltage

Common-mode voltage is the voltage both signal conductors share relative to earth. A motor-drive phase node swings the entire gate-drive circuit hundreds of volts in tens of nanoseconds; a solar string floats hundreds of volts above ground; an RS-485 link between buildings rides on whatever the two earths disagree by. Electronics connected directly across such domains see this voltage as if it were signal. An optocoupler sees it only as stress across its barrier — the detector never knows.

Term Definition Typical spec
CMR (Common-Mode Rejection) How completely a static/low-frequency common-mode voltage is kept out of the output Expressed in dB, or as max VCM
CMTI (Common-Mode Transient Immunity) The fastest common-mode dv/dt the device tolerates without corrupting its output state 15–50 kV/μs (gate drivers 100+ kV/μs)

CMTI matters because of the ~0.5 pF barrier capacitance: a fast common-mode edge drives displacement current \(i = C \cdot dv/dt\) through that capacitance into the detector node. At 50 kV/μs, 0.5 pF passes 25 mA — enough to falsely switch an unprotected output. IGBT, SiC-MOSFET, and GaN gate drivers live on exactly such edges (SiC and GaN reaching 100–200 kV/μs), which is why gate-drive optocouplers advertise CMTI as a headline spec and use internal shields between barrier and detector.

Common-Mode Transient Demo

A half-bridge switch node slews between 0 and 800 V. Set the slew rate and the coupler's CMTI rating; the demo shows whether the output stays clean or glitches.

Switch-node slew rate 50 kV/μs
Displacement current through 0.5 pF barrier
Output integrity

EMI protection

Interference reaches circuits by four couplings: conducted (along shared wiring), capacitive (electric field between adjacent conductors), inductive (magnetic field linking loops), and radiated (far-field pickup on cables acting as antennas). Galvanic isolation attacks the two worst offenders directly: it severs the conducted path entirely, and its sub-picofarad barrier presents ~300 kΩ at 1 MHz to capacitive coupling — versus milliohms for a copper connection. Loop area reduction (Module 11) then handles the magnetic terms. The optical link itself neither emits nor receives electromagnetic interference — photons in silicone are immune to both E and H fields at these energies.

Switching noise: dv/dt, di/dt, ground bounce

PWM systems — buck converters, motor bridges, Class-D stages — generate the noise they must then survive. Every edge produces dv/dt (displacement currents through every stray capacitance), di/dt (inductive kick \(v = L \cdot di/dt\) across every nanohenry of wiring — 10 A/ns through 5 nH is 50 V), and ground bounce as return currents lift the "quiet" reference. The animated waveforms below show a gate signal, the resulting switch-node edge, and the ground bounce a directly-connected controller would see versus the clean reference an isolated controller keeps.

High-voltage transients arrive from many directions: ESD (kilovolts from a charged human, sub-nanosecond edges), inductive kickback (any interrupted coil — relay, solenoid, contactor — flies back to whatever voltage it needs), utility switching (capacitor-bank and load switching on the grid), and industrial switching (contactors and drives sharing your feeder). The isolation barrier is the last line that keeps these events on the side where they happened.

IEC 61000-4-5 installation class Surge (line-to-earth) Typical environment
Class 1 0.5 kV Well-protected electronics room
Class 2 1 kV Office / light commercial
Class 3 2 kV Industrial, separate earthing
Class 4 4 kV Heavy industrial, outdoor cabling
Class 5 6–10 kV Exposed / utility / lightning-prone sites

Lightning rarely strikes the equipment; it doesn't need to. An indirect strike raises local ground potential by kilovolts and induces surges on every long cable run — the reason outdoor equipment, solar installations, and plant-wide networks specify isolation plus surge protection as a hierarchy: gas discharge tubes and MOVs clamp the bulk energy, series impedance limits current, TVS diodes clamp the remnant, and the galvanic barrier withstands what leaks past (its VIOSM/VIOTM impulse rating — Module 1) without ever offering the surge a path into the protected domain.

Fault containment architecture

System architects use isolation to segment designs into fault domains, so that any single failure is contained where it occurs. The fault tree below reads bottom-up: without barriers, every leaf event propagates to the top; with barriers, propagation stops at the dashed lines.

System destroyed Controller domain Communication domain Power domain ISOLATION BARRIERS — propagation stops here Field wiring short Surge on comms line IGBT failure Ground fault Faults below the barrier line are contained in their domain — the failure never reaches the top event.
Fail-safe design. Isolation also simplifies the safety argument: an optocoupler-based E-stop input is dark (off = safe) when unpowered, when the LED fails, when the cable is cut, and when the field supply dies. All credible failures land on the safe side — the property safety assessors call "fail-safe by construction."

Real products are voltage zoos. Logic runs at 3.3 V (or 1.8/5 V), industrial I/O standardized on 24 V decades ago, telecom and server backplanes use 12/48 V, control power arrives at 110 V DC in substations, mains is 230/400 V AC, and EV batteries reach 800 V. These domains must exchange signals constantly — and every boundary crossing is a design decision: level-shift (shared ground, same fault domain) or isolate (independent domains).

3.3 V
MCU / FPGA logic
24 V
Industrial I/O
48 V
Telecom / PoE / batteries
230 VAC
Mains
400 VAC
3-phase industrial
800 V
EV battery bus

Worked interface: MCU GPIO → optocoupler → 24 V PLC input

3.3 V MCU side 24 V PLC side PC817 GPIO R1 GND A Rc +24 V to PLC IN GND B (PLC) isolation barrier
Resistor calculations

LED side, targeting \(I_F\) = 5 mA from a 3.3 V GPIO (VF ≈ 1.2 V):

\( R_1 = \dfrac{3.3 - 1.2}{5\,\text{mA}} = 420\,\Omega \rightarrow \textbf{390–430}\,\Omega \) standard value.

Check the GPIO can source 5 mA (most can; use a transistor buffer if not). Output side: PLC sink-type input typically needs > 2 mA at > 11 V to register ON. With CTRmin 130% (B bin) derated ×0.5 EOL: IC = 5 × 1.3 × 0.5 = 3.25 mA — comfortably above threshold, margin 1.6×.

Power check: R1 dissipates I²R = 10.5 mW; LED 6 mW — trivial, any 0402/0603 works.

FPGA to industrial equipment

FPGAs raise the stakes on both sides of the interface: their I/O banks are expensive, low-voltage (1.2–3.3 V), ESD-sensitive, and often routed to fine-pitch BGAs that cannot be field-repaired — while their applications (machine vision triggers, motion-control encoders, industrial Ethernet PHY resets) reach into electrically hostile machinery. The isolation calculus is asymmetrical: a 30-cent optocoupler protecting a $300 FPGA on a $3000 assembly. Fast logic-output couplers (or digital isolators for >10 Mbps links) are standard practice on every FPGA pin that leaves the cabinet; the FPGA's speed advantage is preserved by choosing the coupler family to match the signal rate, not by omitting the barrier.

Standards exist because the market cannot verify safety by inspection: a buyer cannot see whether a barrier withstands 4 kV. Certification converts invisible engineering into legal market access — CE, UL, and CB marks are entry tickets to their respective markets, and the isolation component's certificates are load-bearing documents in every product safety file. The key standards:

IEC 61010-1 — lab & industrial

Measurement, control, and laboratory equipment. Defines measurement categories (CAT I–IV), creepage/clearance by working voltage and pollution degree, and Hi-Pot test levels. Governs multimeters, PLCs, and test instruments.

IEC 60601-1 — medical

Medical electrical equipment. Patient vs operator protection (MOPP/MOOP), leakage current limits in µA, applied-part classes (B/BF/CF), and 2×MOPP reinforced barriers between mains and patient.

UL 1577 — component isolation test

The optocoupler's own recognition standard: defines the VISO withstand rating (e.g. 5 kVrms, 60 s) and 100% production Hi-Pot testing (1 s at 120% of rated). Says nothing about continuous working voltage — that is VDE 0884's job.

VDE 0884-17 / IEC 60747-17 — safe barrier physics

Certifies the barrier for life: rated working voltage VIORM, partial-discharge testing (detects microscopic voids before they become failures), impulse withstand, and the reinforced/basic classification — the certificate that legally supports "reinforced insulation" claims.

IEC 62368-1 — AV & ICT equipment

Hazard-based safety for consumer electronics and networking: classifies energy sources (ES1/ES2/ES3), requires barriers between ES3 (mains) and touchable ES1 circuits, replacing legacy 60950-1/60065.

ISO 26262 / AEC-Q — automotive

Functional safety (ASIL levels) plus AEC-Q102 qualification for optoelectronics: EV traction, chargers, and BMS isolation must carry both the safety rating and automotive stress qualification.

Insulation classes — click to explore

HAZARDOUS 230 VAC mains Basic Suppl. TOUCHABLE SELV circuit / user Double insulation: two independent barriers — either alone protects.

Select an application to load its complete design brief: block diagram, isolation boundary, component selection with calculations, layout and EMC guidance, and the mistakes field engineers actually make.

1. Isolating what doesn't need it

Optocouplers on same-board, same-ground signals add cost, delay, and two failure modes for zero benefit. Isolation is for crossing domains, not decoration.

2. Ignoring CTR degradation

The design that "works" at nominal CTR fails at year seven. Stack the margins: bin minimum × temperature × 0.5 end-of-life.

3. Insufficient isolation voltage

Sizing to the working voltage alone, forgetting surges and the required insulation class. The barrier must survive the site's worst day, not its average one.

4. Poor creepage / clearance

A 5 kV part with 2 mm of PCB between its input and output pads is a 2 mm design. Spacing rules apply to the board, not just the component.

5. Ground plane under the barrier

A pour sneaking beneath the package shorts the creepage distance through the board and adds coupling capacitance — keep the barrier zone copper-free on all layers.

6. LED overdrive

Driving 20 mA "for margin" accelerates the very aging that erodes margin. 2–10 mA is the reliability sweet spot.

7. Ignoring standards until layout is done

Insulation class, creepage, and certified components must be chosen at architecture time. Retrofitting 8 mm spacing into a finished board means a respin.

8. Wrong package for the voltage

A 4-pin mini-flat with 5 mm creepage cannot serve a reinforced 230 V barrier needing 8 mm — wide-body exists for a reason.

9. No derating documentation

If the review can't see your worst-case math (CTR, temperature, aging, tolerances), it didn't happen. Module 9 provides the worksheet.

Answer the questions; the tree recommends optocoupler, an alternative isolation technology, or no isolation at all.

From system working voltage to required barrier ratings (same engine as Module 1, extended with insulation-class context).

Working voltage across barrier (AC rms) 230 V
Min. VIORM
Suggested VISO
Min. impulse withstand

Worst-case CTR stack: bin minimum × temperature × end-of-life, checked against your required output current.

LED current IF 5 mA
Max ambient temperature 60 °C
Required output current 1 mA
Worst-case CTR (EOL, hot)
Worst-case IC
Margin vs requirement

Indicative PCB creepage per IEC 62368-1-style tables (functional/basic & reinforced), by pollution degree and material group.

Working voltage (rms) 230 V
Basic insulation creepage
Reinforced insulation creepage
Note

Educational approximation. Consult the applicable standard's tables and your safety engineer for certification-grade values.

Pick your product category to list the standards that typically govern its isolation design.

Knowledge Check

Five parts: 20 multiple-choice questions, 15 true/false statements, a 10-pair matching exercise, an isolation-boundary identification drill, and 3 engineering scenarios. Score 80% on the multiple choice to pass.

Part A — Multiple Choice

Question 1 of 20  •  Score: 0/20

Part B — True or False

Score: 0/15

Part C — Match Term to Definition

Drag each term onto its definition (or click chip, then target). 10 pairs.

Placed: 0/10

Part D — Place the Isolation Barrier

A grounded MCU controls a mains-driven heater through a driver stage, while a thermocouple amplifier reads the process temperature. Click the connection where the isolation barrier belongs. Then find the fault: click the element that creates a ground loop.

MCU 3.3 V / GND A link A Driver stage mains-referenced link B Heater 230 VAC Thermocouple amp bonded to process link C wire Click a dashed zone: where should the barrier go? Then find the ground-loop culprit.
Task 1: click the link that must become an isolation barrier (optocoupler).

Part E — Engineering Scenarios

Scenario 1 — FPGA meets a noisy 24 V motor controller
"My FPGA communicates with a noisy 24 V motor controller over a 3-wire interface at 1 Mbps, cable length 4 m, separate supplies and grounds." Should an optocoupler be used? Why? What isolation voltage is required, and which package is suitable?
Yes — isolate. Separate grounds + motor PWM noise + 4 m of cable is the textbook ground-loop/common-mode case, and an FPGA bank is the most expensive victim available. At 1 Mbps a phototransistor part is too slow: use high-speed logic-output couplers (6N137/HCPL-2601 class, 10 Mbps) or a digital isolator. Isolation voltage: the working voltage is only 24 V, but surges from the motor environment set the need — 2.5 kVrms VISO (functional/basic) is standard and cheap; an 8-pin DIP or SO-8 package with ≥ 4 mm creepage is ample. Power the field side from the controller's 5 V, keep the FPGA side on its own rail, and route the two grounds to their own connector pins.
Scenario 2 — The sensor that reads garbage every time the crane runs
A 4–20 mA level sensor 80 m away reads perfectly — until the overhead crane operates, when readings jump ±15%. Both the sensor housing and the PLC cabinet are earthed. Diagnose and fix.
Diagnosis: ground loop. Two earth points 80 m apart + crane motor return currents = volts of transient ground difference driving current around the sensor-cable-PLC loop, superimposing error on the 4–20 mA signal. Fix: break the loop — an isolated analog input channel (optocoupler-isolated ADC or isolation amplifier), or an isolated 4–20 mA repeater at the cabinet. Grounding the shield at one end only and twisted-pair cable help, but only isolation removes the galvanic path. Verify by measuring AC voltage between the two earths during crane operation (expect volts).
Scenario 3 — How many MOPP does the pulse oximeter need?
A mains-powered patient monitor connects a fingertip SpO2 sensor (applied part, type BF) to its internal 3.3 V processing board, which is fed from an offline SMPS. Where are the barriers, and what class must each be?
Two barriers. (1) Mains → secondary: the SMPS transformer plus its feedback optocoupler must provide 2×MOPP reinforced insulation (4 kVrms test, 8 mm creepage at 230 V). (2) Processing board → patient-applied part: type BF requires the patient connection to float; the SpO2 interface (LED drive + photodiode signal) crosses a second barrier rated for patient protection with leakage ≤ 100 µA normal condition. A single component failure anywhere must leave one intact means of protection — which is exactly what the MOPP accounting verifies.

Why optocouplers are used — in one table

Driver Mechanism Where it bites
Safety No conductive path from hazardous voltage to people or SELV circuits Mains, medical, industrial, EV
Ground isolation Breaks ground loops; each domain keeps its own reference Long cables, multi-cabinet, battery systems
Noise immunity Common-mode voltage becomes barrier stress, not signal error; CMTI handles fast edges Motor drives, inverters, SiC/GaN bridges
Fault protection Surges and faults stop at the barrier; damage contained per domain Outdoor, lightning-prone, high-energy systems
Mixed voltages 3.3 V logic commands 24–800 V domains without sharing their fate Every industrial interface
Compliance Certified barriers convert engineering into legal market access Every certified product

Design guidelines

  • Isolate at domain boundaries — where grounds, voltages, or safety classes change — and nowhere else.
  • Choose insulation class first (functional/basic/reinforced), from the end-product standard; it drives component and package selection.
  • Size the barrier for the environment's transients (surge class), not just the working voltage.
  • Stack CTR margins: bin minimum × temperature × 0.5 end-of-life ≥ 2× the required output current.
  • Match coupler family to signal speed; use digital isolators above ~10 Mbps.
  • Preserve the barrier on the PCB: creepage, clearance, no copper under the package (Module 11).
  • Document the worst-case math — reviews and auditors both read it.

Downloadable Resources & Checklists

Expand any guide to view it; use your browser's Print → PDF to save a copy.

AI Tutor

Ask about isolation, standards, ground loops, or component selection — or tap a suggested question. (Runs locally on this page's knowledge base.)

Hi! I'm the Module 2 tutor. Ask me why isolation is required, how to pick an isolation voltage, what MOPP/MOOP mean, or when a digital isolator beats an optocoupler.