| Episodes | 32 |
| Mentions | 44 |
| Cited here | 24 |
| First — last | #9 — #704 |
| Top guests | Henry Ott, Alan Wolke, Hank Zumbahlen |
| Related | radio frequency |
Electromagnetic interference is unwanted electromagnetic energy that couples from one circuit or product into another, and electromagnetic compatibility is the paired requirement that a product not radiate harmful interference capable of stopping other equipment from working, aircraft being the canonical example, as well as the converse requirement that the product itself keep working in the presence of others.[523] Radiated emission is governed principally by loop area, so long traces that enclose area with their return path are the structures that cause emissions problems, and reviewing a layout for large loops is the fastest way to predict which circuits will fail a scan.[183] Noise and interference control is largely absent from electrical engineering curricula, so practitioners who need it typically teach themselves on a live problem; the field’s early literature was thin enough that an engineer solving noise coupling problems on the job could become his organisation’s resident expert on the strength of that self-teaching alone.[165]
Physical mechanisms
The boundary between low-frequency and high-frequency circuit behaviour is not sharply defined but sits somewhere between 10 and 100 MHz, with design below about a megahertz comparatively simple and circuits above 100 MHz needing different treatment; that boundary is roughly the same one at which a circuit starts to radiate appreciably.[185] Physically small boards have inherently small current loops, which limits both radiated emission and field pickup, and on a product of roughly two by one centimetres with reasonable layout, EMC did not become a design problem at all while antenna performance did.[175] The same reasoning distinguishes integration levels: digital isolators that integrate their transformer on-chip are relatively immune to external magnetic fields because the enclosed loop is tiny, whereas an outboard transformer is larger and adds the traces running out to it, enclosing more area and making the same function more susceptible.[185]
Fast edges are the usual source. On audio hardware the recurring offender on radiated emission scans is the I2S master clock, precisely because its edge rate is deliberately made fast to meet the rise-time specification of the DAC or ADC it drives, so the requirement that makes the clock work is the requirement that makes it radiate.[474] Such a clock shows up on a scan not at its third or fifth overtone but at much higher-order overtones, which for typical audio rates land near 100 MHz, the region where the cable interconnects running between boards inside the product are efficient antennas, so the emission is radiated by the internal wiring rather than by the clock trace.[474] On USB and similar interfaces the radiating structure is likewise often not the cable itself but the termination at its end, so an expensive well-built cable plugged into a poor jack can still cost a product its compliance result.[183] A product built with two isolation domains is difficult to take through type approval for a related reason, because the interconnecting cables have very little return path available and consequently carry and radiate harmonics that a single-domain design would not present.[640]
Phase-cut mains dimmers work by chopping the AC waveform, and the resulting fast discontinuity in the current makes them a strong broadband emissions source, with the hard part being to solve the same dimming function cleanly at low cost and in volume.[56] Transient emissions from a product often do not originate in a single source at all, frequently being a mixing or intermodulation product of several subsystems, so the offending emission appears only when two or more buses happen to line up, which is why such failures present as intermittent and resist single-cause debugging.[117]
Susceptibility
Safety-critical actuators in vehicles were an immunity problem from the moment they were introduced. Early anti-lock braking systems, first fitted to heavy trucks because a locked-wheel jackknife is so costly, proved susceptible to nearby transmitters, and a radio keyed from a vehicle alongside could apply all the brakes on the rig.[77]
Undefined logic states are a classic vulnerability: a bus line left undriven during a read, without an adequate pull-up to define its state, returns whatever RF the node has picked up rather than a defined logic level, turning an ordinary input into an antenna and producing data-dependent intermittent faults.[222] Running a single-ended, ground-referenced data signal off a board on a cable is a standing immunity failure in a hostile electromagnetic environment, because it provides no common-mode rejection of the noise the cable picks up, and differential signalling with error detection on the link is the corresponding remedy; a medical device that failed regulatory testing showed exactly this pattern, together with no ESD protection at all on its external interfaces.[704]
Very high gain analogue instruments are commonly built inside a fully enclosed metal box, because at high gain any coupled interference appears at the output as signal, and a field photometer built around an ultra-high-gain op-amp stage was completely enclosed in aluminium for that reason.[171] Enclosures can also create problems: closing a module into its metal housing can form a resonant cavity that produces oscillations or spurious behaviour absent when the unit is open, an effect diagnosed by the symptom disappearing whenever the lid comes off and remedied by packing conductive foam inside the enclosure.[117]
Interference sources in the environment
A GSM handset transmits in high-current bursts, on the order of an amp for roughly twenty milliseconds, which makes a phone an unusually severe local interference source compared with continuous-carrier transmitters.[319] A handheld multimeter that locked up near such a phone was traced to a PCB trace on the MSP430 programming line coupling energy at the GSM band and driving the processor into a programming mode, and the failure was frequency-specific, since handsets on other cellular standards in other parts of the spectrum did not reproduce it.[319]
Utilities inject control signalling onto the mains supply, and that signalling has been known to disturb connected equipment, with the common remedy being to fit a reasonably capable mains filter to susceptible products so the provider’s control signals are attenuated at the input.[477] Electric fence energisers, and lightning during storms, disturb nearby digital television receivers, because the impulsive field swamps the front end so packets are lost and the picture freezes while error correction reruns; digital modulation changes the symptom from a visible analogue smear to a hard dropout, which makes the interference harder for the user to attribute.[481] A single non-compliant legacy appliance can disrupt shared infrastructure in the same way: an old television emitted a single high-level impulse noise event at switch-on that took out a village’s broadband connection, and its strict 7 a.m. regularity was the clue that identified it, because a fault repeating at a fixed hour points at someone’s daily routine rather than at the network.[510]
Bench equipment can also be the interference source in the fault being investigated. A hot air gun brought near a board induces disturbance through its own field, so the symptom is wrongly attributed to temperature when the actual stimulus is the large coil energised inches from the hardware.[494]
Design practices
Noise, EMI and ESD contingencies are cheapest when built into the architecture and board layout from the outset, with the protection, the filtering and the fallback options designed in before there is a symptom, rather than retrofitted after a failed compliance run, since by then the architecture and layout decisions that caused the problem are fixed.[704] Emissions remedies also redistribute energy rather than remove it, so a patch such as copper tape over one leak commonly makes the noise escape somewhere else, the way pressing on a balloon makes it bulge elsewhere, though that behaviour is not a reason to skip the first fix because each leak can be closed in turn.[165]
Several remedies operate on the source rather than the path. Slowing the configured slew rate on a device’s outputs reduces the harmonic content of its switching edges and can turn a failed EMC scan into a pass, at the cost of the timing margin the faster edges were providing.[472] DC-DC converters that sit near a radio have their switching frequency chosen to fall outside the receiver’s sensitive bands, and parts switching at 2 MHz were introduced so that the fundamental sat clear of the AM band, after which choosing switching frequency by the radio’s bands rather than by converter efficiency alone became established practice.[635] Deliberately modulating a converter’s switching frequency spreads its emitted energy over a band instead of concentrating it in narrow harmonics, lowering the peak the customer’s EMC scan measures without changing the total energy emitted.[635]
Protection at the supply input is a separate concern: a lightning strike near the installation or an ungrounded operator handling the hardware injects a large surge current that can destroy the switching converter electronics, so the input needs dedicated protection such as TVS diodes rather than reliance on the converter’s own ratings.[401] Filtering cannot simply be applied everywhere on a mass- and vibration-constrained platform, however, because filter components make the design physically larger and heavier and a heavy, bulky assembly survives launch vibration and shock less well; on a satellite power bus shared with transmitters, computers, actuators and motors, how much to filter is a mass-versus-immunity judgement rather than a default.[401]
Testing
Most product companies have no in-house EMI compliance chamber and buy chamber time from an external laboratory, so every emissions question costs both money and calendar time and cannot be answered on the bench; an intermittent result is the worst case, since a unit that fails one run in five leaves the team with no clear pass and no clear defect.[117] Precompliance emissions work on the bench is therefore done by sweeping a near-field probe over the powered assembly looking for spurs and anything else coming out of it, which localises the radiating structure before any chamber time is bought.[436] Such scanning is deferred to the final prototype before production rather than run on early iterations, on the same reasoning that a test fixture is not built until the design is settled, since earlier boards will still change and measurements made on them do not carry forward.[436]
Holding a mobile phone next to a powered product is a standard informal bench check for susceptibility, since the handset radiates strongly across its band, and keying nearby VHF radios has been used the same way; it gives no compliance data but reveals gross immunity problems before any chamber booking.[319] An immunity fix can equally be demonstrated by deliberately exposing the repaired hardware to a worst-case field rather than by retesting in the normal environment, as when fitting a six-foot cable and wrapping it around an uncased CRT deflection yoke proved a circuit was no longer sensitive.[222]
Compliance-only components complicate production test. A pull-up or bias resistor placed across a microcontroller’s crystal may be present for emissions reasons rather than for oscillator function; on an AVR-based board the part was not required for the oscillator to start, and its likely purpose was to control high-frequency noise so the board would pass FCC testing.[11] Because such a component is not needed for the product to work, a board missing it powers up and behaves normally, so the unit ships and the defect surfaces only in the field or at retest.[11]
Regulatory context
A connected product sold internationally is qualified for emissions against several regional regimes at once, including FCC in the United States, the Canadian regulator renamed from Industry Canada to ISED, and CE marking for Europe, so the board must be shown free of EMI issues in each region rather than in one.[509] European requirements place more weight on immunity, meaning a product’s ability to keep working amid interference from other products, than United States requirements do, which are weighted toward limiting what a product emits.[472]
Automotive EMC requirements began as manufacturers’ self-imposed restrictions rather than as external regulation, because the consequences of interference in a car, such as an airbag firing, an engine control circuit disturbed or cruise control misbehaving, followed directly from what the product was.[165] Regulatory testing for medical devices is more demanding than consumer compliance, with applied EMI, applied ESD and magnetic field exposure all imposed and the system required to continue functioning throughout, no reset or crash being permitted during an ESD pulse; the priority ordering resembles aerospace, with reliability first and cost a secondary consideration.[704]
The ESD immunity requirement for a consumer product is 4 kV contact discharge and 8 kV air discharge, representing the spark that reaches the device from a held metal object or across the gap from a finger to a front panel, while real discharges of 15 kV are entirely ordinary on a dry day, so the standard is a floor rather than a description of the service environment.[472] Regulatory EMC limits function as a minimum bar rather than a design target more generally, and companies making products that are handled daily across wide temperature, humidity and RF conditions generally hold internal requirements that exceed them, with the appropriate level scaling to the product, since a giveaway novelty and a cell phone do not warrant the same standard.[472]