Electronics Guide

Where the Depth Lives

The two companion articles that develop this subject in detail now sit in the electromagnetic compatibility category, which is where the site's EMC taxonomy is rooted: Radiated Emission Sources covers clock harmonics, switching converters, board structures, heat sinks, connectors, and the coupling paths that turn internal noise into radiated energy, and Radiated Emission Measurement covers test environments, antennas, receivers, pre-compliance methods, correlation between facilities, and measurement uncertainty.

What follows here is the interconnect-facing view: how a trace, a plane, a via, or a cable becomes an unintentional antenna, and what layout and shielding choices change that.

Fundamental Radiation Mechanisms

Radiation occurs when a time-varying current produces a time-varying magnetic field, which in turn induces a time-varying electric field. Beyond a certain distance the two fields sustain each other and detach from the source as a propagating wave. How efficiently a structure performs this conversion depends almost entirely on its physical size relative to the wavelength of the current it carries.

Wavelength and frequency are related by λ = c/f, where c is the speed of light, approximately 3 × 10⁸ m/s. A 100 MHz signal therefore has a wavelength of 3 meters. As a working rule, a structure becomes a consequential radiator once its longest dimension approaches λ/10, which at 100 MHz is 30 centimeters. Cables, board traces, ribbon assemblies, and enclosure panels all fall in that size range across the frequencies of interest, which is why a modern product contains dozens of candidate antennas that no one designed.

At frequencies where the wavelength greatly exceeds the size of the circuit, radiation is weak and the surrounding fields are predominantly reactive: energy sloshes back and forth between the source and the near field rather than escaping. As frequency rises and the wavelength shrinks toward circuit dimensions, radiation efficiency climbs steeply. This is why a product with a modest clock frequency can still fail a radiated emission test. What matters is not the fundamental but the harmonic content produced by fast edges, which extends far above the clock rate.

Near Field and Far Field

The character of the field depends on distance from the source. For an electrically small source, the boundary between the near and far zones falls near r = λ/2π. Inside that radius the electric and magnetic fields are not yet locked into a fixed ratio, and their relative strengths depend on whether the source is a current loop or a voltage-driven conductor. Beyond it the fields settle into a plane wave whose ratio, the wave impedance, approaches the impedance of free space, about 377 ohms.

This distinction has practical consequences. A predominantly magnetic source, such as a small current loop, produces a low-impedance near field that is difficult to shield with thin non-magnetic materials. A predominantly electric source, such as a high-voltage node with little current, produces a high-impedance near field that even a thin conductive layer attenuates easily. At 30 MHz the far-field boundary sits about 1.6 meters from a small source, so a standard three-meter compliance measurement is a far-field measurement across the whole radiated range, whereas a probe held a centimeter above a board is emphatically not.

Differential-Mode Radiation

Differential-mode radiation arises from the loop formed by a signal path and its return path. Current flows out along one conductor and back along the other, and the enclosed area behaves as a small loop antenna. The field radiated by such a loop in the far field is proportional to the loop area, to the current magnitude, and to the square of the frequency, falling off inversely with distance from the source.

The frequency-squared dependence is the dominant term. Doubling the frequency quadruples the radiated field for the same current and geometry, a 12 dB increase, which is why the upper harmonics of a clock rather than its fundamental usually determine whether a board passes. Loop area is the one factor a designer controls directly and completely, and it is therefore the primary lever.

Minimizing differential-mode radiation means minimizing enclosed area. Route signals directly over a continuous reference plane so that return current can follow immediately beneath the trace, which is the path of least inductance above a few megahertz. Keep power and ground plane pairs closely spaced in the stackup. Avoid splits, slots, and voids in reference planes, since a return current forced to detour around an obstruction encloses a far larger area than the trace geometry suggests. Where a signal must cross a plane discontinuity, provide a deliberate return path such as a stitching capacitor or a via to an adjacent intact plane.

Tightly coupled differential pairs help for the same geometric reason: the two currents are equal and opposite and their fields cancel in proportion to how closely the conductors track each other. Cancellation is never perfect, and any length or impedance mismatch between the two halves converts part of the differential signal into common mode, which radiates far more efficiently.

The radiation pattern of a small loop is worth understanding, because compliance testing measures both polarizations and rotates the equipment through a full circle to find the worst case. A small loop radiates most strongly in the plane of the loop and has a null along its axis. A loop lying flat in the plane of a board therefore radiates outward toward the board edges, with a null directly above and below it, and the field it produces is polarized parallel to the board. A loop standing perpendicular to the board radiates most strongly in its own vertical plane. Real boards contain loops in every orientation, so no single test geometry can be assumed to be benign.

Common-Mode Radiation

Common-mode radiation dominates the radiated emissions of most real products, particularly any product with external cables. Common-mode current flows in the same direction on every conductor of a cable or bundle and returns through the chassis, the earth connection, or stray capacitance to the surroundings. Rather than the small, deliberate loop of differential-mode current, its return path encloses the entire physical extent of the system, and the cable becomes an efficient monopole driven against the product as a counterpoise.

For an electrically short radiating element, the far field produced by common-mode current is proportional to the current, to the length of the element, and to the first power of frequency. That is a gentler frequency dependence than the frequency-squared law of a small loop, but the effective length of a cable is one hundred to one thousand times the effective dimension of a board-level loop, and the length term wins decisively in practice.

The currents involved are startlingly small. On a one-meter cable measured at three meters, roughly 8 microamperes of common-mode current at 30 MHz produces a field of about 100 microvolts per meter, which is precisely the FCC Class B limit in that band. At higher frequencies still less current suffices, because the radiated field rises with frequency faster than the limit line does. Expressed as power, a product need only radiate on the order of a nanowatt to fail. No practical measurement of a signal will ever reveal a few microamperes of stray current, which is why common-mode problems are diagnosed with current probes and near-field probes rather than with an oscilloscope on the signal.

Common-mode current has many origins, and most of them are asymmetries rather than deliberate signals. Imbalance between the two halves of a differential pair converts differential current to common mode. Voltage developed across the finite inductance of a ground structure drives the chassis relative to the cable. Capacitive coupling from a switching node, a heat sink, or a fast bus couples displacement current into anything nearby. Ground potential differences between separately referenced subsystems push current along the interconnect between them.

Cable resonance sharpens the problem. A cable radiates most efficiently when its length is an odd multiple of a quarter wavelength, at which point it presents a low driving-point impedance and even a weak source can push substantial current onto it. A one-meter cable resonates near 75 MHz and again near 225 MHz, 375 MHz, and so on. A radiated emission plot that shows tall, narrow peaks marching up the spectrum at a regular interval is very often a cable resonance rather than a source problem.

Control begins with prevention: keep common-mode current off the cable in the first place. Balanced drivers and receivers with good common-mode rejection reduce the amount generated. Continuous reference planes without slots or splits keep return currents tight and hold the board's reference at a uniform potential. Where prevention is incomplete, block the current at the boundary. Common-mode chokes present high impedance to common-mode current while passing the differential signal untouched, and filtered or feed-through connectors shunt high-frequency common-mode current to the chassis at the point where the cable leaves the shielded volume. The location of the shield bond matters as much as its existence: a shield grounded at the wrong point can raise emissions rather than lower them by giving common-mode current a more efficient path onto the cable exterior.

Unintentional Antennas in Electronic Systems

Recognizing which structures in a product are capable of radiating is the first step in predicting and controlling emissions. Three families account for most failures: conductors on the board, cables leaving the enclosure, and openings in the shield.

Board Traces and Planes

A trace routed over a continuous reference plane is a transmission line, not an antenna. Its return current flows in the plane directly beneath it, the enclosed loop is very small, and the structure radiates weakly. Radiation from board conductors therefore signals that something has broken this arrangement: a plane split beneath a high-speed trace, a via transition that changes reference planes without a nearby return via, a trace routed near a board edge where the plane no longer extends beneath it, or a signal with no adjacent plane at all. In each case the return current is forced onto a longer path, the enclosed area grows, and the trace begins to drive the board itself as a common-mode source.

Clock traces deserve particular attention because they carry a repetitive waveform with fast edges and therefore concentrate energy into discrete, narrow harmonics that show up cleanly against the noise floor of a receiver. The spectral envelope of a trapezoidal pulse falls at 20 dB per decade above a breakpoint set by the pulse width and at 40 dB per decade above a second breakpoint near 1/(πt_r), where t_r is the edge transition time. A 1 ns edge places that second breakpoint around 318 MHz, so significant harmonic energy persists well into the hundreds of megahertz regardless of how modest the clock frequency appears.

Cables

External cables are the most efficient radiators in most systems, both because they are long and because they extend outside whatever shielding the enclosure provides. Unshielded cables radiate directly from common-mode current on their conductors. Shielded cables radiate from current flowing on the outer surface of the shield, which is excited by imperfect shield termination, by asymmetry in the shield current distribution, and by the finite transfer impedance of the shield itself. A shield does not make a cable quiet; it moves the problem to the shield surface and to the quality of the bond at each end.

Heat Sinks and Internal Metalwork

Any isolated conductor of appreciable size inside a product can be driven as an antenna. A heat sink mounted on a fast processor is capacitively coupled to the die and package, floats at radio frequency unless deliberately bonded, and has dimensions that put its resonances squarely in the range covered by compliance testing. The same applies to unbonded shield cans, mounting brackets, display bezels, and internal cable routing that runs the length of the chassis. Bonding such structures to the board reference or the chassis at multiple points, with short and low-inductance connections, converts them from radiators into part of the shield.

Apertures and Slots

An opening in a conductive surface is the complement of a wire antenna and radiates by the same physics. Ventilation holes, display windows, panel seams, and connector cutouts all allow internal fields to escape. What governs leakage is the longest dimension of the opening, not its area: a slot reaches half-wave resonance and radiates most efficiently when its length approaches λ/2, which for a 10 centimeter slot occurs near 1.5 GHz. A hairline gap along a 15 centimeter seam is therefore a far worse radiator than a 5 millimeter round vent hole, even though the vent hole removes more metal.

Slots cut in reference planes are doubly harmful. They radiate as slot antennas, and they force signal return current to detour around them, enlarging the loop area of every trace that crosses the slot. A long slot in a ground plane, whether created by a row of through-hole connector pins or by a deliberate analog and digital partition, is one of the most reliable ways to turn a well-behaved board into a failing one.

Enclosure Shielding

A conductive enclosure is the last line of defense against radiated emissions and the first line of defense against external interference. Its theoretical performance is easy to achieve on paper and very difficult to achieve in a product, because real enclosures must open, ventilate, display information, and admit cables.

Shielding Mechanisms

Three mechanisms contribute to shielding effectiveness, the ratio in decibels between the field present without the shield and the field present with it. Reflection at the air-to-metal boundary rejects energy because of the impedance mismatch between free space and a good conductor. Absorption dissipates the energy that does enter the metal as resistive and magnetic loss. Multiple internal reflections within a thin shield modify the result, generally reducing effectiveness for thin barriers at low frequencies.

Which mechanism dominates depends on frequency and field type. Electric fields and plane waves reflect strongly from any conductive surface, so even thin aluminum foil is an excellent electric-field shield. Low-frequency magnetic fields reflect poorly and must be handled by absorption or by redirection through a high-permeability path, which is why magnetic shielding calls for steel, mu-metal, or permalloy rather than aluminum. Absorption is governed by the skin depth δ = √(2/ωμσ), where ω is angular frequency, μ is permeability, and σ is conductivity. One skin depth of material contributes roughly 8.7 dB of absorption loss, so a barrier three to five skin depths thick attenuates strongly by absorption alone.

Material Selection

Aluminum combines good conductivity, low weight, corrosion resistance, and low cost, which makes it the default for electronic enclosures. Copper conducts better but is heavier and more expensive. Steel offers less conductivity but far higher permeability, which makes it the better choice where low-frequency magnetic fields are the concern. Conductive coatings on plastic housings, whether nickel- or copper-filled paint, vacuum-deposited metal, or electroless plating, can be effective, but they demand careful control of coating continuity, thickness, and the treatment of mating surfaces.

Above a few megahertz, material choice is rarely the limiting factor. A one-millimeter aluminum enclosure offers well over 100 dB of theoretical shielding effectiveness, an amount no product ever realizes. The achievable figure is set by the discontinuities, and a single untreated seam or oversized aperture can pull an enclosure from a nominal 100 dB down below 30 dB. A shield is only as good as its worst opening.

Seams and Joints

Seams between panels are the most common shielding failure. A seam is a long, narrow slot, and it radiates whenever current must cross it or whenever internal fields excite it. Because leakage scales with the length of the opening rather than its width, a seam that is mechanically tight but electrically discontinuous performs badly: paint, anodizing, and oxide films are all insulators, and a painted joint that looks flush can behave as an open slot at radio frequency.

Effective seams require continuous, low-impedance metal-to-metal contact along their full length. Conductive gaskets, whether beryllium-copper fingerstock, knitted wire mesh, or conductive elastomer, maintain contact across mechanical tolerance and surface irregularity. Where gaskets are not used, fastener or contact spacing must be close enough that the unsupported span never approaches a resonant length. Practice commonly calls for contact points no farther apart than one-twentieth to one-tenth of a wavelength at the highest frequency of concern; at 1 GHz, where the wavelength is 300 millimeters, this means contact every 15 to 30 millimeters. Mask paint from mating surfaces, or specify a conductive finish such as chemical conversion coating that provides corrosion protection without an insulating layer.

Apertures and Ventilation

Since enclosures must breathe and display information, aperture management is unavoidable. A useful first-order estimate places the shielding effectiveness of a single slot at 20 log₁₀(λ/2L), where L is the longest dimension of the opening. That expression yields 20 dB when the slot is one-twentieth of a wavelength long and falls to zero at half-wave resonance, which is the basis for the common design guideline of keeping the largest aperture dimension below λ/20 at the highest frequency requiring control. At 1 GHz that limit is 15 millimeters. Multiple nearby apertures degrade the result further, by roughly 10 log₁₀(n) for n similar openings.

For a given total open area, many small holes leak far less than a few large ones, because leakage is governed by the longest dimension of each opening rather than by area. Perforated panels and honeycomb vent panels exploit this directly. A honeycomb panel goes further by making each cell a waveguide operating well below its cutoff frequency, where attenuation is roughly 30 dB for every increment of depth equal to one cell diameter. A honeycomb three cell-diameters deep can therefore contribute on the order of 100 dB while passing air freely.

Display windows require either a fine conductive mesh or a transparent conductive coating, bonded continuously around the entire perimeter to the enclosure. Touch screens compound the difficulty, since the sensing layer must remain both optically transparent and electrically functional while the shield around it stays continuous. In every case the perimeter bond, not the shielding material itself, determines the result.

Where a large opening cannot be avoided, break up its long dimension. A rectangular cutout partitioned by internal conductive dividers behaves as several short slots rather than one long one, provided the dividers make solid electrical contact at both ends. Relocating an aperture away from a region of intense internal field is often cheaper and more effective than shrinking it.

Connectors and Cable Penetrations

Every cable that crosses the shield boundary is a potential bypass around all the effort spent on panels and seams. Connector shells must bond to the enclosure with 360-degree contact, not through a wire, a mounting screw, or a pigtail. Filtered connectors integrate a feed-through capacitor on each pin, shunting high-frequency energy to the chassis exactly at the boundary before it can reach the external cable. They are costly, but for a product with many I/O lines they frequently solve a problem that no amount of internal filtering will.

Enclosure Bonding

Older guidance recommending a single grounding point for an enclosure applies to low-frequency noise and safety practice, not to shielding. At the frequencies covered by radiated emission testing, any conductor long enough to matter has substantial inductance, and a single connection cannot hold a large structure at a uniform potential. Effective practice bonds enclosure panels, internal shields, board reference planes, and connector shells to one another at many points with short, wide, low-inductance connections, so that the shielded volume behaves as one equipotential surface at radio frequency. Low-frequency circulating-current concerns, where they exist, are better addressed by circuit isolation than by deliberately breaking the radio-frequency bond.

Cable Radiation and Control

Because cables dominate radiated emissions in most systems, cable treatment is often the difference between passing and failing. Three measures work together: shielding the cable, blocking common-mode current with impedance, and choosing routing and length that reduce antenna efficiency.

Shielded Cable Design

A shield provides a controlled return path for signal current and confines the fields of the pair or coax inside it. Its effectiveness is quantified by transfer impedance, the voltage developed along the inside of the shield per ampere of current flowing on the outside. Lower transfer impedance means better isolation between the internal signal and the external environment, and transfer impedance generally rises with frequency as the shield's imperfections become electrically significant.

Construction determines the result. A single braid provides good flexibility and low DC resistance but has optical coverage below 100 percent, and its weave allows magnetic field leakage that worsens with frequency. Foil provides complete coverage but is fragile and has high resistance along its length. Foil-plus-braid construction combines the coverage of foil with the mechanical durability and low resistance of braid and is the usual choice for demanding applications. Multiple braid layers or solid tube shields perform better still at correspondingly higher cost and stiffness.

Termination is more important than construction. A pigtail, in which the shield is gathered into a short wire before connection, defeats the shield: it destroys the coaxial geometry, inserts inductance in series with the shield current, and creates a loop that radiates. Terminate shields with a full 360-degree bond using a connector backshell, conductive cable gland, or EMI grommet that clamps the entire circumference to the chassis. A few centimeters of pigtail can cost tens of decibels of shielding effectiveness at ultra-high frequencies.

Common-Mode Chokes

A common-mode choke presents high impedance to current flowing in the same direction on all conductors while presenting almost none to the differential signal, because the differential fluxes cancel in the core. A ferrite sleeve slipped over a cable is the simplest form and is often adequate; a wound choke, in which the cable passes through the core several times, raises the impedance in proportion to the square of the turns count.

More turns is not universally better. Interwinding capacitance forms a parallel resonance with the winding inductance, above which the choke's impedance falls again, so a heavily wound choke may be less effective at the frequency that actually fails. Material choice matters for the same reason: manganese-zinc ferrites reach peak impedance in the low megahertz region, while nickel-zinc ferrites remain useful into the hundreds of megahertz. Broadband problems sometimes need two cores of different materials in series.

Placement is critical. Position the choke as close as possible to the point where the cable leaves the shielded enclosure. Any length of cable between the noise source and the choke remains free to radiate, and a choke installed a few centimeters inside the enclosure protects only the cable beyond it. Finally, verify that the added common-mode inductance does not degrade the signal it protects; high-rate data links can lose timing margin to a choke chosen for its low-frequency impedance alone.

Cable Routing and Length

Length sets both the antenna size and the resonant frequencies, so shorter cables are quieter cables. Where length is fixed, routing still helps. Running a cable close against a chassis wall or a ground plane provides a nearby return path, reducing loop area and lowering the impedance the common-mode source must drive. Bundling cables that exit through a single penetration keeps them mutually coupled and concentrates the filtering effort at one boundary. Cables that leave from opposite ends of a chassis, by contrast, form a dipole with the chassis as its center and are among the worst configurations available.

Avoid routing a cable across an aperture or along a seam, where the fields leaking out are strongest and the coupling into the cable is highest. Keep noisy cables, such as motor leads and switching-converter outputs, physically separated from quiet ones, and cross them at right angles where they must meet.

Measurement and Verification

Compliance is demonstrated by measurement, and design decisions are validated the same way. The radiated emission measurement subcategory treats test facilities, antennas, receivers, and uncertainty in detail; what follows is the overview an engineer needs in order to interpret a test report and plan a debugging session.

Test Environments

The open area test site is the historical reference: an outdoor conductive ground plane with the equipment on a non-conductive turntable and a calibrated antenna at 3 or 10 meters. It is faithful but exposed to weather and to ambient broadcast signals, and it has been largely displaced by the semi-anechoic chamber, which lines walls and ceiling with absorber while retaining a conductive floor. The chamber reproduces the same single-reflection geometry indoors, free of ambient signals and available on demand. Above 1 GHz the floor reflection is suppressed with additional absorber, effectively making a fully anechoic environment. Either facility must be validated against the site requirements of CISPR 16-1-4, using normalized site attenuation below 1 GHz and site voltage standing wave ratio above it.

Smaller facilities serve development rather than certification. A GTEM cell is a tapered, terminated transmission-line structure that creates a uniform field region large enough for a small product, converting emissions into a measurable cell voltage; a defined procedure correlates the result with open-site data. GTEM cells occupy a fraction of the floor space of a chamber, work over a very wide frequency range, and give fast, repeatable comparisons, which makes them well suited to before-and-after evaluation of design changes.

Procedure, Detectors, and Bandwidths

A standard measurement rotates the equipment under test through a full 360 degrees of azimuth while the receiving antenna is scanned in height, conventionally from 1 to 4 meters, in both horizontal and vertical polarization. The purpose of this search is to find the maximum field at each frequency, because the pattern of an accidental antenna is unpredictable. Cables are arranged and dressed according to the standard, since their position materially changes the result.

Receivers follow CISPR 16-1-1. Below 1 GHz the resolution bandwidth is 120 kHz, and above 1 GHz it is 1 MHz. A fast peak-detector prescan identifies frequencies of concern; those frequencies are then remeasured with the quasi-peak and average detectors against which the limits are actually written. Quasi-peak weighting penalizes repetitive impulses more than infrequent ones, so a broadband, low-repetition-rate source may pass quasi-peak despite a high peak reading. Comparing peak, quasi-peak, and average readings at a failing frequency is itself diagnostic: a narrow emission with all three readings equal is a continuous-wave source such as a clock harmonic, whereas a large peak-to-average difference indicates impulsive switching.

Near-Field Scanning and Pre-Compliance

Near-field probes, small magnetic loops or short electric monopoles, measure fields within millimeters of a board, a cable, or a seam. They do not predict far-field levels, since the near-field relationships differ from those in the far zone and probe coupling is uncalibrated, but they localize a source to a specific component, trace, or opening better than any other technique. Automated scanners assemble probe readings into field maps overlaid on the board layout, which turns an abstract spectrum peak into a physical location.

A clamp-on radio-frequency current probe on a cable is equally valuable and often more directly actionable, because common-mode current on cables correlates well with far-field emissions. Measuring current on each cable in turn identifies which one drives the failure, and re-measuring after a ferrite is added quantifies the improvement immediately. Combining a pre-compliance receiver, a set of near-field probes, and a current probe on the bench catches most problems long before a chamber is booked.

Regulatory Standards and Limits

Radiated emission limits vary with product category, intended environment, and market. Two distinctions run through nearly all of them: the separation of equipment into a stricter residential class and a more permissive commercial or industrial class, and the specification of a measurement distance that must be respected when comparing figures between standards.

FCC Part 15 (United States)

The Federal Communications Commission regulates unintentional radiators under Title 47, Part 15, Subpart B. Section 15.109 sets Class B limits, for equipment marketed for use in a residential environment, at 100 microvolts per meter from 30 to 88 MHz, 150 microvolts per meter from 88 to 216 MHz, 200 microvolts per meter from 216 to 960 MHz, and 500 microvolts per meter above 960 MHz, all measured at 3 meters. Class A limits, for equipment marketed for commercial, industrial, or business use, are 90, 150, 210, and 300 microvolts per meter across the same bands, measured at 10 meters. Where bands overlap at an edge, the tighter limit applies.

The difference between the classes is larger than the raw numbers suggest, because the measurement distances differ. Referring the Class A values to 3 meters adds roughly 10 dB, so Class B is about 10 dB stricter than Class A through most of the spectrum, a factor of three in field strength. The upper frequency of the scan is not fixed but depends on the highest frequency generated or used within the device.

CISPR and European Standards

The International Special Committee on Radio Interference, known by its French initials CISPR and operating under the IEC, develops the emission standards adopted across most of the world. CISPR 32 specifies emission requirements for multimedia equipment, a category that spans information technology, audio, video, and broadcast-receiving equipment. It replaced the older CISPR 22 for information technology equipment and CISPR 13 for broadcast receivers, both of which were withdrawn in 2017. In Europe it is implemented as the harmonized standard EN 55032, which together with the immunity standard EN 55035 supports the EMC portion of CE marking under the EMC Directive.

Below 1 GHz, CISPR 32 sets the Class B radiated limit at 40 dB(µV/m) from 30 to 230 MHz and 47 dB(µV/m) from 230 to 1000 MHz at a 3-meter distance. The corresponding Class A limits are specified at 10 meters and, when referred to the same distance, sit roughly 10 dB higher. Above 1 GHz the separation between the classes narrows to about 6 dB, and limits are stated with both peak and average detectors. As with the FCC scheme, the upper frequency of the measurement depends on the highest internal frequency of the product, reaching 6 GHz for the fastest equipment.

Sector-Specific Standards

Many industries impose their own requirements, generally stricter than the commercial baseline because the electromagnetic environment is harsher or the consequences of interference are more severe. Automotive components are tested to CISPR 25 in an absorber-lined shielded enclosure at a one-meter antenna distance, with limits graded into severity classes so that a vehicle manufacturer can select the level appropriate to a given installation. Medical electrical equipment is covered by IEC 60601-1-2, which invokes CISPR 11 for emissions and adds immunity requirements reflecting the clinical environment. Military and aerospace equipment falls under MIL-STD-461, whose RE102 test measures radiated electric field emissions from 10 kHz to 18 GHz at a one-meter distance, and under RTCA DO-160 for airborne equipment. The wider frequency range and closer measurement distance of these standards make them considerably more demanding than commercial limits.

Design Strategies for Low Emissions

Emission control is inexpensive when it is designed in and expensive when it is added on. A layout choice costs nothing at the schematic stage; the same problem discovered in a compliance chamber may force a board respin, an enclosure change, or an added filter that eats the product's margin. The measures below are ordered by leverage, from suppressing the source to containing what escapes.

Source Control

The cheapest emission is the one never generated. Use the slowest edge rate the timing budget allows. Because the spectral envelope above the edge-rate breakpoint scales inversely with transition time, relaxing an edge from 1 nanosecond to 5 nanoseconds lowers that breakpoint from roughly 318 MHz to roughly 64 MHz and reduces the high-frequency envelope by about 14 dB. Many logic families offer controlled-slew variants, and a small series resistor at the driver both slows the edge and damps reflections.

Spread-spectrum clocking modulates the clock frequency slightly, spreading each harmonic across a band rather than concentrating it in one line. Because the receiver integrates over a fixed 120 kHz bandwidth, spreading a harmonic across a wider band reduces the reading. The benefit grows with harmonic number, since a fixed percentage deviation covers proportionally more absolute bandwidth at higher frequencies, and reported reductions commonly fall between about 5 and 15 dB. The technique has real limits: it redistributes energy rather than eliminating it, it adds tracking jitter that some serial links and clock-recovery circuits cannot tolerate, and specifications for certain interfaces restrict or forbid it. Treat it as margin, not as a substitute for sound layout.

Beyond timing, choose components deliberately. Switching converters with soft-switching topologies, controlled gate drive, or integrated input filters emit far less than the same power stage designed for efficiency alone. Enable clocks only in the subsystems that need them, and gate off unused oscillators rather than leaving them running.

Board Layout

Layout determines both how much noise a board generates and how much of it reaches an antenna. Provide solid, unbroken reference planes adjacent to every signal layer so that return current has a direct path. Route high-speed signals on internal layers sandwiched between planes, where the surrounding copper both confines fields and shields the trace. Keep traces away from board edges, where the reference plane no longer extends beyond them and fields fringe into free space; pulling high-speed routing back from the edge and adding a stitched ground guard along the perimeter both help.

Partition the board by function rather than by cutting up the reference. Place noisy switching regulators, clock generators, and high-speed buses away from I/O connectors and analog sections, and keep their current loops physically compact. Resist the impulse to split ground planes: a continuous plane with careful component placement almost always outperforms a partitioned plane with a bridging connection, because the split forces return currents into detours that radiate. Place decoupling capacitors immediately at the power pins they serve, with short, wide connections and vias directly into the planes, since the loop formed by the capacitor, its vias, and the die is itself a radiator at high frequency.

Treat the I/O region as a controlled boundary. Group connectors along one edge, place filtering and common-mode chokes in the connector area, and provide a quiet chassis-referenced region beneath them so that filtered lines do not immediately recouple to board noise on the far side of the filter.

Filtering and Containment

Filtering intercepts energy before it reaches a radiating structure. Apply it at the boundary a signal crosses, whether that is a power inlet, an I/O connector, or the wall between a noisy and a quiet domain. Combine capacitor values so that the aggregate impedance stays low across the range of concern, and place a low-inductance path to the reference at every filter return, since a filter is only as good as its ground connection. Feed-through capacitors mounted in the chassis wall outperform ordinary components on a board because they place the shunt element exactly at the shield boundary and carry no lead inductance across it.

Where source control, layout, and filtering leave a residue, containment remains. Internal shield cans over an offending module, compartmentalization within a larger enclosure, and local shielding of a cable run all confine fields close to their origin, where the structures available to radiate them are small.

Troubleshooting Radiated Emission Failures

A failing scan reports a frequency and a level, not a cause. Systematic troubleshooting works backward from the symptom to the source, the coupling path, and the antenna, and it tests one variable at a time.

Identifying the Source

Start with the frequency itself. Divide the failing frequency by the candidate clock frequencies in the product; an exact integer ratio identifies the oscillator immediately. Switching converters produce a fundamental in the tens or hundreds of kilohertz with harmonics extending far higher, often as a comb of closely spaced lines. Emissions that shift when a subsystem changes operating mode belong to that subsystem.

Next identify the antenna. Disconnect cables one at a time, or replace them with short substitutes, and watch the level at the failing frequency. A large drop when a cable is removed identifies a common-mode cable problem, which is by a wide margin the most common finding. If the level is unaffected by every cable, the enclosure or the board itself is radiating, and near-field probing across the surface and along the seams will localize it. Temporary copper tape over a suspect seam or vent is a fast confirmation.

Finally, read the spectrum's shape. Narrow, stable peaks indicate a continuous-wave source such as a clock harmonic or a cable resonance. Broad elevated regions indicate fast edges or impulsive switching. Peaks at regular intervals that do not correspond to any clock usually indicate a resonant structure being excited by broadband noise, in which case detuning or damping the structure may be easier than quieting the source.

Corrective Actions

Cable-driven emissions respond to common-mode chokes, better shield termination, and improved filtering at the connector. A snap-on ferrite is the standard first experiment: if it helps, a designed-in choke or filtered connector will solve the problem permanently, and the ferrite has already quantified the improvement available. Check every shield termination for pigtails, which are among the most frequent and most easily corrected faults.

Aperture and seam emissions respond to gasketing, closer fastener spacing, smaller or subdivided openings, honeycomb vent panels, and relocation of the opening away from intense internal fields. Where the enclosure cannot be changed, an internal shield around the offending module addresses the same problem from the other direction.

Board-level emissions call for improved return-path continuity, additional or relocated decoupling, series damping at fast drivers, and in stubborn cases a stackup or routing revision. Late-stage fixes at this level are the most expensive of all, which is the strongest practical argument for treating emission control as a design input rather than a verification result.

Whatever the fix, change one thing at a time and re-measure. Stacked modifications conceal which one worked, and a change that helps at one frequency may worsen another. Record the configuration precisely, since cable dress and enclosure closure alone can move a reading by several decibels.

Conclusion

Radiated emissions combine electromagnetic theory with mechanical design, board layout, and component selection, and they respond only to a systematic approach. The underlying physics is compact: a structure radiates in proportion to how large it is relative to the wavelength and how much high-frequency current it carries. Everything else follows from that. Differential-mode loops are controlled by geometry, common-mode currents are controlled by balance and by impedance at the boundary, and whatever remains is controlled by the integrity of the shield.

The recurring lesson is that emission control cannot be retrofitted cheaply. Stackup, plane continuity, connector placement, enclosure seams, and cable strategy are all fixed early, and each of them determines a product's emission behavior far more than any filter added afterward. Engineers who plan the return current path, keep cables free of common-mode current, and treat every enclosure opening as an antenna will meet regulatory limits with margin and will spend far less time in a chamber discovering why they did not.

Related Topics