You have been designing circuits.
You have not been designing for the real world.
EMI/EMC is not a compliance checkbox. It is the discipline of understanding what electromagnetic energy does to your circuit when nobody is watching — and designing so the answer is nothing.
Most hardware engineers encounter EMI/EMC late — at the test lab, during a field failure, or when a product gets returned. This module exists to reframe the problem before that happens. Read the four sections in order. Each one closes a gap you did not know you had.
What is EMI/EMC
The coupling triangle. Conducted vs radiated. Emissions vs immunity. The vocabulary that makes the rest of this bootcamp possible.
Why Products Fail
Four products that passed every bench test and failed in the field. The root cause in each case was not a mistake — it was a gap in the mental model.
Why Standards Exist
Standards are not arbitrary checklists. Each one was written because something failed and someone was harmed. Understanding the origin changes how you read them.
The Engineering Mindset
The cost of finding EMC problems at architecture stage vs at the test lab. The three levers every EMC engineer uses. The mindset shift this bootcamp is built on.
Electromagnetic Interference (EMI) is unwanted electromagnetic energy that disrupts the operation of a circuit. Electromagnetic Compatibility (EMC) is the discipline of designing systems so that they neither generate disruptive EMI nor are affected by it. Every EMC standard, every test, every design rule in this bootcamp traces back to these two definitions.
But definitions are not understanding. Here is the physical reality underneath them.
The Coupling Triangle
Every EMI problem — without exception — has exactly three components. Remove any one and the problem disappears.
Something generating electromagnetic energy
A switching MOSFET. A clock oscillator. A motor drive. A microcontroller toggling a GPIO. Any circuit element where current or voltage changes with time is an EMI source.
A mechanism that carries the energy to the victim
Conducted (through shared wires, planes, or supply rails) or radiated (through space as electromagnetic waves). The path is often invisible — a parasitic capacitance, a shared ground impedance, a cable acting as an antenna.
Something that is disrupted by the energy
A sensitive analog input. A radio receiver. Another product nearby. Your own microcontroller. The receptor does not have to be fragile — it only has to be sensitive enough at the frequency the source is generating.
Conducted vs Radiated — and Why It Matters
⚡ Conducted
- Energy travels through electrical connections — power cables, signal cables, shared supply rails
- Measured at the power port using a LISN — a defined test network
- Dominant at lower frequencies — typically below 30 MHz
- Primary source: switching power supplies, motor drives, anything with large dI/dt on a shared supply
📡 Radiated
- Energy travels through space as electromagnetic waves — no wire needed
- Measured in a chamber at 3m or 10m from the product
- Dominant at higher frequencies — typically above 30 MHz
- Primary source: cables acting as antennas, PCB loops, clock harmonics, switching node radiation
The bench is a controlled environment. Clean supply, short cables, no adjacent equipment, no RF fields, no switching machinery nearby. Your product is designed on the bench and tested in a chamber — but it lives in neither. It lives in the field, where all of these assumptions break down simultaneously.
Expand each case. The symptom appears first — exactly as an engineer would encounter it. Then the root cause. Then what needed to change.
What Happened
The 32 MHz system clock had a 3rd harmonic at 96 MHz. The USB cable attached to the hub during testing picked up common-mode current from the PCB ground and re-radiated it efficiently — a 1m USB cable is a quarter-wave antenna at 75 MHz. Close enough to resonate at 96 MHz.
What Was Missed
The pre-compliance scan was done without the USB cable attached. The board alone passed. Adding the cable — which is part of the standard test configuration — made the cable the antenna. The PCB was not the radiating structure. The cable was.
What Fixed It
A common-mode choke on the USB data lines reduced the common-mode current driven onto the cable. The board layout was not changed. A single component in the right place reduced the 96 MHz emission by 11 dB.
What Happened
The camera module's LVDS serialiser ran at 400 MHz. Its 3rd harmonic at 1.2 GHz fell within the FM broadcast band. The coaxial video cable — routed near the roof antenna — coupled the interference into the antenna cable through near-field inductive coupling.
What Was Missed
The cable routing was finalised after EMC testing. The system was tested on a bench with cables bundled away from each other. In the vehicle installation, the video cable ran within 30mm of the antenna feeder for 400mm. That proximity created enough mutual inductance to transfer the interference.
What Fixed It
Rerouting the video cable away from the antenna feeder eliminated the coupling path. No component change. No board respin. The fix was mechanical — but only visible if you understood the coupling mechanism first.
What Happened
The EFT burst injected fast transients onto the power cable. The transients coupled capacitively through the PCB's power supply filter into the digital ground plane. The ground plane voltage shifted transiently, causing the reset pin of the microcontroller — referenced to that ground — to see a false low.
What Was Missed
The filter components were correctly specified for conducted emission. But the layout placed the filter on the wrong side of the ground plane split — so the transient energy that the filter was supposed to block was actually being injected directly into the digital section's reference plane.
What Fixed It
Moving the filter components to the correct physical location — before the ground plane split — with a short, direct path to chassis ground. Same components. Different placement. The reset pin never saw the transient again.
What Happened
The 433 MHz RF field induced a common-mode voltage on the sensor cable. The analog front-end amplifier had a common-mode rejection ratio (CMRR) of 60 dB at DC — but only 20 dB at 433 MHz, because CMRR degrades rapidly with frequency in most op-amp topologies. The induced noise was well within the amplifier's output range and appeared as a valid sensor signal.
What Was Missed
The designer specified the amplifier's DC CMRR — which was excellent. The amplifier's CMRR at RF frequencies was never checked. The datasheet showed it in a graph on page 12. It was never read. The gap between DC CMRR and RF CMRR is where this product failed.
What Fixed It
A common-mode filter on the sensor cable input, with a bypass capacitor to chassis ground at the connector, reduced the RF common-mode voltage by 26 dB before it reached the amplifier. The amplifier's degraded CMRR at RF no longer mattered because the RF energy was removed upstream.
A compliance standard looks like a list of test levels, measurement methods, and pass/fail criteria. That is how most engineers read it — as a bureaucratic hurdle to clear before shipping. That reading is technically correct and practically useless.
Every test level in every standard was derived from a real measurement of a real threat in a real environment. When a standard says "test at 3 V/m RF field," that field level was chosen because 3 V/m is what a walkie-talkie produces at 1 metre. When a standard says "test with a 2 kV surge," that 2 kV was measured on real industrial power supplies during real switching events. The numbers are not arbitrary. They are the documented physics of the environment your product will live in.
Standards you will encounter — and why they were written
This is not a complete list. It is the minimum you need to understand what you are designing for.
| Standard | What it governs | The incident behind it |
|---|---|---|
| CISPR 32 | Emissions from multimedia and consumer electronics — the limit your product must stay under | Proliferation of switching power supplies in the 1980s caused measurable degradation of AM and FM broadcast reception in residential areas. CISPR 32 codified the emission limits that prevent it. |
| FCC Part 15B | US market emission limits — Class A (commercial) and Class B (residential) | The FCC received tens of thousands of interference complaints in the 1970s as personal computers entered homes. Part 15 established the limits that made coexistence with broadcast radio possible. |
| IEC 61000-4-4 | EFT/burst immunity — how your product must survive fast transients on power and signal ports | Industrial contactor switching on shared supply rails was causing systematic control system failures in manufacturing plants. The burst waveform was derived from measured transients on real industrial panels. |
| ISO 7637-2 | Automotive supply transients — load dump, inductive switching, ignition events | Alternator field collapse events on early automotive electrical systems were destroying unprotected electronic modules. Each pulse waveform in ISO 7637 was measured on a real vehicle electrical system. |
| IEC 60601-1-2 | Medical equipment EMC — emissions and immunity for equipment near patients | Documented interference to active implantable devices (pacemakers) from electronic equipment in clinical settings. The immunity test levels reflect the field strengths generated by clinical equipment at patient proximity. |
| MIL-STD-461 | Military electronics — emissions and susceptibility across the full frequency range | Electromagnetic interference between co-located military systems — radar, communications, weapons — on shared platforms caused operational failures. MIL-STD-461 defines the separation requirements between systems on the same platform. |
There is a specific moment in most hardware projects where EMC becomes the problem. It is not during architecture. It is not during schematic. It is not during layout. It is in the test lab, four weeks before ship date, when the product fails radiated emissions by 8 dB and the options are a board respin, a metal enclosure nobody budgeted for, or a costly retest cycle.
That moment is expensive not because EMC is hard. It is expensive because all the decisions that determine EMC performance were already made — in the stackup, in the schematic, in the layout — by people who were not thinking about EMC when they made them.
The Cost of Finding Problems Late
The Three Levers
Every EMC solution — every ferrite bead, every ground plane, every shielded cable, every filtering capacitor — is an application of one of three levers. Understanding which lever you are pulling, and why, is what separates reactive EMC debugging from proactive EMC design.
Suppress at the Source
Reduce the amplitude or rate of change of the interfering signal at its origin — slower switching edges, lower switching frequency harmonics, smaller loop area for the switching current. The most effective lever. Applied at architecture and layout stage.
Break the Coupling Path
Interrupt the mechanism that carries energy from source to receptor — filtering on power ports, isolation on signal interfaces, shielding around radiating structures, ground plane continuity that forces return current to follow a controlled path.
Harden the Receptor
Make the victim circuit less sensitive to the interference — differential signalling with wide common-mode range, decoupled supply rails, input filtering on sensitive analog inputs, firmware-level noise rejection. The last resort, but sometimes the only option.
The bootcamp follows the real hardware engineering process — from building the mental model, through the full design flow, into domain-specific deep dives, and out through testing and certification. Each module delivers one complete piece of the engineering capability.
Ready to build this capability?
Self-paced access · All 15 modules · Design checklists · Scenario-based quizzes · Built by a practicing hardware engineer
These are not definition questions. They are engineering scenarios — the kind you encounter on real projects. Answer based on what you read in this module.
The rest of the bootcamp goes deeper on every answer above.
15 modules · Full design flow · Pre-compliance through chamber · Built by a practicing hardware engineer