EMC pre-compliance testing
The Test That Finds the Failure Before the Lab Does
About half of products fail EMC testing on the first attempt, and each failure costs $5,000–$30,000 and 2–6 weeks. A pre-compliance scan on your own bench finds the failure before the lab does, for $300–$800 a scan.
The board works on the bench. The radio connects. The sensor reads. The firmware runs. The team books the lab, ships the samples, and waits four to six weeks for the chamber slot, and then the product fails radiated emissions by 6 dB at one frequency, and the launch date becomes the retest date.
That is not bad luck. About half of products fail EMC testing on the first attempt, that number comes from Intertek Labs, and it lines up with what every EMC test lab reports. Each retest cycle adds $5,000 to $30,000 and 2 to 6 weeks to the timeline.
And a failed cycle means waiting four to six weeks for a chamber slot, paying for a design review, spinning up a new board, and building fresh samples. By the time the team is back in front of the chamber, 90 days have gone by.
The test that finds the failure before the lab does is called pre-compliance testing, and it is not a smaller version of the lab test. It is a different kind of test with a different job. The lab test is a verdict: pass or fail, and the report goes to a certification body.
Pre-compliance is a diagnostic: it shows how close to the boundary the product is, where it is leaking, and what design change will bring it under the limit. It is the difference between one clean cycle and three painful ones.
The simple idea first
Think of the lab as an exam and pre-compliance as the practice paper. The exam is expensive, it is booked weeks ahead, and you get one shot at it before the retake costs you another month. The practice paper is cheap, you can sit it as many times as you like, and it tells you which questions you will get wrong, so you can fix them before the exam.
Or think of it as a smoke alarm and a fire inspection. The lab is the inspection: it decides whether the building can open. Pre-compliance is the smoke alarm: it goes off while the fire is still small enough to put out with a glass of water. Nobody skips the smoke alarm because the inspection is coming.
The technical idea is simple too. Every digital product is a small radio station that does not mean to be one. A clock oscillator switches 48 million times a second, and every switch sends a little radio energy out through the traces, the cables, and the seams of the enclosure.
The lab measures that energy at a calibrated distance, 3 metres for the FCC, 10 metres for CISPR 32, and compares it against a limit line. Pre-compliance measures the same energy on your own bench, with a spectrum analyser and a near-field probe, and tells you which frequencies are too loud before the lab does.
The rest of this article is about what the bench can see, what it cannot see, and what the numbers say about the cost of skipping it.
The coin flip
Start with the number that decides everything else. About half of products fail EMC testing on the first attempt. That is Intertek's figure, and it lines up with what every EMC test lab reports.
Rohde & Schwarz, which sells the test equipment, puts the failure rate higher, roughly 70 to 90%, for products that arrive at compliance testing without pre-compliance work. Ideambox, in its EMC design and pre-compliance reference, frames it differently again: EMC failures account for about 40% of first-pass lab failures in consumer electronics. Three sources, three numbers, one direction: failing first is normal, not exceptional.
The second round is where the numbers turn. Intertek reports failure rates of 5 to 7% on retests, and by the second round success is 93 to 97%. The product that fails once almost always passes the second time, because the first failure told the team exactly what to fix. The question is whether the team pays for that lesson in the lab, or on its own bench.
Here is what the lab costs. Emissions testing, the radiated and conducted scans, runs $2,000 to $5,000 and takes 1 to 2 days. Immunity testing, which the EU requires and the FCC does not, adds $3,000 to $8,000 and 1 to 4 days.
By device type, a simple digital device with no radio is $2,000 to $5,000; a device using a pre-certified RF module is $5,000 to $10,000; a custom Wi-Fi or Bluetooth transmitter is $8,000 to $15,000; a multi-radio device is $12,000 to $25,000. Chamber time runs $1,500 to $2,500 a day. In high-reliability markets, automotive, medical, defence, a cycle rises to $13,600 to $34,000, and most teams carry a further 10 to 30% buffer for design changes after a failure.
And here is what the failure costs. Each retest cycle adds $5,000 to $30,000 and 2 to 6 weeks. The fix itself can be cheap, adding a ferrite bead costs pennies per unit and takes a day, or expensive: redesigning the power supply filter means a board respin at $2,000 to $15,000 and 2 to 4 weeks.
The bottleneck is rarely the testing itself. Lab queue time, documentation preparation, and deficiency rounds during certification review eat more calendar time than the measurements. During peak seasons, after CES in the first quarter, and before holiday launches in the third, lab booking lead times can stretch to 6 weeks or longer.
Book 4 to 6 weeks ahead during these periods, and ask about retest booking before you commit, because a 6-week wait for a retest slot after a failure can destroy a product launch timeline.
The total, for a simple FCC-only campaign, is 3 to 6 weeks. For FCC plus CE plus ISED, 4 to 10 weeks. And if the product fails, add 90 days to get back in front of the chamber.
The test, in two halves
EMC testing has two halves, and they ask opposite questions.
Emissions testing asks: does this device interfere with other equipment? Every market requires it. The FCC, CE, and ISED all mandate emissions testing before a product can be sold. If the switching power supply pumps noise onto the AC line, or the clock oscillator radiates radio energy from its traces, emissions testing will find it.
Immunity testing asks: can this device withstand interference from the outside world? The EU requires immunity testing under the EMC Directive and the Radio Equipment Directive. The FCC and ISED do not. This gap between US-only and international campaigns surprises a lot of teams when they expand to Europe for the first time. A product designed and tested only for the FCC has never been exposed to ESD, surge, EFT, or radiated immunity testing. When that product goes through CE testing for the first time, immunity failures are common, because nobody thought about susceptibility during the design.
The emissions half has two tests. Radiated emissions measures the radio energy radiating from the device and its cables, from 30 MHz to 6 GHz and beyond. It is where most products fail.
The lab places the device on a turntable in a semi-anechoic chamber, rotates it through 360 degrees, and scans an antenna across the full frequency range while varying its height from 1 to 4 metres to find the maximum emission at each frequency, in both horizontal and vertical polarisation. Results are compared against the limit lines using a quasi-peak detector below 1 GHz and peak or average detectors above it.
A typical failure shows 1 to 3 discrete frequency spikes exceeding the limit by 3 to 15 dB, usually at exact harmonics of a clock or switching frequency.
Conducted emissions measures the radio noise the device pushes back onto its AC power cord, from 150 kHz to 30 MHz. A Line Impedance Stabilisation Network, a LISN, presents a standardised 50-ohm impedance to the power input while the lab measures the noise voltage on each conductor, line and neutral. Switching power supply harmonics are the dominant source of conducted emissions failures. The fundamental switching frequency, typically 65 kHz to 500 kHz, and its harmonics appear directly in the measurement. A missing or undersized EMI input filter is the most frequent root cause.
The immunity half is a suite of disturbances, each with its own standard: ESD at ±4 kV contact and ±8 kV air (IEC 61000-4-2); radiated immunity at 3 to 10 V/m from 80 MHz to 6 GHz (IEC 61000-4-3); EFT and burst at ±1 kV on power and ±0.5 kV on signal lines (IEC 61000-4-4); surge at ±1 kV line-to-line and ±2 kV line-to-ground (IEC 61000-4-5); conducted immunity at 3 V from 150 kHz to 80 MHz (IEC 61000-4-6); voltage dips and interruptions at 0%, 40%, and 70% of nominal (IEC 61000-4-11); and power-frequency magnetic field at 3 A/m (IEC 61000-4-8).
Each test applies a calibrated disturbance while the lab watches whether the device keeps working. The results are judged against four criteria: A, normal operation during and after the test, no degradation; B, temporary degradation allowed during the test, but the device recovers by itself; C, temporary loss of function needing an operator to restart it, still a pass at some test levels; D, non-recoverable damage or data loss, always a fail.
ESD is the immunity test that causes the most failures. The lab applies calibrated discharges, up to ±8 kV contact and ±15 kV air, to every user-accessible surface and connector. The typical failure is that the device resets, locks up, or loses communication when discharged near an I/O port. The root causes are unprotected connectors with no TVS diodes, ESD protection devices placed too far from the connector pin, they need to be within 5 to 10 mm, and no low-impedance path from the connector shell to chassis ground.
One more thing the lab measures, and the bench cannot: the margin. Results are plotted against the limit lines, and a pass is not enough. A measurement 1 dB below the limit is technically passing but leaves no room for production variation, temperature drift, or measurement uncertainty. Most labs will flag anything within 3 dB of the limit as a concern. The product needs margin, not just a pass, and the bench is where the margin is bought.
The class: A or B, and why it matters more than the test
Before any test is run, one decision sets the limit line the product is measured against. FCC Part 15 and CISPR 32 both split devices into two classes, and the choice is made by where the product is used, not by how good it is, and not by how much the test costs.
Class B is for residential environments. Stricter limits. Required for any consumer product, anything that could end up in a home, an apartment, a hotel room, anywhere a television or radio receiver sits within a few metres.
Class A is for commercial and industrial environments. The limits are roughly 10 dB more relaxed, and the product carries a warning label stating it may cause interference in residential settings.
The gap is about 10 dB, a factor of roughly three in field strength, but it varies by band. Compared at the same 3-metre distance, Class B radiated limits are about 6 to 11 dB tighter than Class A, and Class B conducted limits are 13 to 23 dB tighter. On conducted emissions at 0.5 to 5 MHz, Class A allows 73 dBµV quasi-peak and Class B allows only 56, a 17 dB tightening.
On radiated emissions at 30 to 88 MHz, Class A is 90 µV/m at 10 metres and Class B is 100 µV/m at 3 metres, which works out to roughly 9.5 dB tighter at the same distance. Above 960 MHz the radiated gap narrows to about 6 dB.
The 10 dB gap is significant, and it is the trap. A product that passes Class A with 6 dB of margin will fail Class B. That is MarkReady's sentence, and it is worth reading twice: the product did not get worse. The limit moved.
There is a real case of exactly this. A 65 W desktop adapter passed its emissions test at 10 metres with about 5 dB of margin, and the report was issued against Class A limits. Eight months later the same adapter was quoted to a US retail chain.
Their compliance team asked for Class B evidence and sent the file back the same week. Nothing in the report was wrong. The laboratory measured against the limit set it was given.
The decision that caused the problem was made much earlier, when the class was chosen to fit the test budget rather than the channel the product would eventually be sold through.
The rule is simple: if there is any chance the product ends up in someone's home, it needs to meet Class B. And the class is not a certificate or an approval issued by an authority. It is a limit set inside a standard, recorded on the test report, and in the United States mirrored by a statement that must appear in the user manual. Two products can be identical in hardware and still be assessed to different classes, one of them will simply fail the tighter limit.
There is one more gap between markets, and it is bigger than the class gap. A product that passes CISPR 32 Class B emissions will almost always pass FCC Part 15B as well, since the limits are virtually identical for conducted emissions and close for radiated, and the reverse is also generally true. But test reports are not directly transferable, because the measurement methods and documentation formats differ.
The real gap is immunity. A product designed and tested only for the FCC has never been exposed to ESD, surge, EFT, or radiated immunity testing, and when it goes through CE testing for the first time, immunity failures are common. The emissions half travels well between markets.
The immunity half does not travel at all.
The bench: what it can see, and what it cannot
Pre-compliance testing is measuring the product's emissions on your own bench before an accredited lab does: near-field probes to find the sources, a LISN and a spectrum analyser for conducted emissions, and an antenna for a radiated pre-scan. The goal is not to pass. The goal is to find and fix every problem so that when the team enters the formal lab, it passes on the first attempt.
Here is what the bench can catch, and how well each measurement correlates with the formal result. Conducted emissions, with a LISN and a spectrum analyser, correlate excellently, within 3 to 6 dB of the formal result. Radiated emissions measured relatively, with near-field probes, identifies the sources but not the absolute levels, the probes tell you where the energy is coming from and let you compare before and after, but they do not give calibrated field-strength numbers.
Radiated emissions measured approximately, with an antenna and an outdoor space or a GTEM cell, lands within 6 to 15 dB depending on the environment.
And here is what the bench cannot catch: immunity performance, ESD, surge, and radiated immunity all need specialised generators and a controlled environment; edge cases within 3 to 6 dB of the limit line; and spurious emissions above the analyser's frequency range. The bench is a diagnostic, not a certificate. It finds the peaks. The chamber decides whether they pass.
The equipment is not exotic, and it is not expensive, not compared with the lab. A basic pre-compliance setup, a spectrum analyser, a LISN, and near-field probes, costs $3,000 to $8,000 to buy. A budget setup of about $8,000 to $15,000 catches about 70% of the issues that would otherwise fail formal lab testing.
Renting a pre-compliance bench for $500 to $2,000 a day pays for itself after one avoided retest. A pre-compliance scan at a lab's reduced rate costs $300 to $800, against $3,000 to $8,000 for a failed lab booking.
The shopping list, with Ideambox's prices: a spectrum analyser covering 9 kHz to 6 GHz, $5,000 to $25,000; a near-field probe set with H-field and E-field probes, $300 to $1,000; a clamp-on current probe, $500 to $2,000; an RF turntable or a test chamber, $5,000 to $50,000; an ESD gun at 8 kV contact and 15 kV air, $2,000 to $5,000; a surge generator, $5,000 to $15,000; and spectrum analyser software with a compliance limit overlay, $1,000 to $5,000. At the bottom of the range, a working pre-compliance setup can be built for under $200, a software-defined radio dongle, a set of near-field probes, a 20 dB preamplifier, and some coax, and it has caught problems on every project it was pointed at.
The probes are the cheapest item on the list and the most useful: they are the stethoscope, and the analyser is the ear.
There are three measurements on the bench, and they answer three different questions.
The near-field probe answers: where is the energy coming from? An H-field loop picks up magnetic field and current flow, noisy traces, IC activity, and a small 5 mm loop finds localised hotspots at pin level. An E-field tip picks up electric field and voltage nodes, connector emissions, slot antennas. The reading is relative, in dBµV into 50 ohms, not an absolute far-field level. But the correlation rules are practical: if a near-field H-probe reading exceeds −10 dBm at 3 mm distance, the associated trace or IC will likely contribute to a far-field limit exceedance; a 20 dB reduction in the near-field reading typically gives 15 to 20 dB of far-field improvement; and the five highest peaks in a near-field scan are the ones that correlate with formal test failures. The probes do not tell you the number. They tell you where to look, and they tell you, in real time, whether the fix worked.
The LISN answers: how much noise is the product pushing back onto its power line? A LISN presents a standardised 50-ohm impedance to the power input and gives the analyser a clean port to measure the noise voltage on each conductor, from 150 kHz to 30 MHz. It is the measurement with the best correlation to the formal result, within 3 to 6 dB, because the setup is the same in both places: the same impedance, the same frequency range, the same detector. For a mains-powered product, the LISN scan is the closest thing to a rehearsal the bench offers. For a battery-powered product with no mains port, there is no conducted test at all, the scan that matters is radiated.
The antenna, or the GTEM cell, answers: how loud is the product, roughly, at a distance? A simple antenna a metre away from the product on a workbench, in a conference room or a basement where ambient signals are low, gives a rough estimate of pass or fail. The measurement is no longer calibrated, so the team establishes a baseline trace and compares every fix against it. Henry Ott's method adjusts the FCC and CISPR limits to a 1-metre test distance and adds a further 6 dB on top of the extrapolated limits, because the measured frequencies below 150 MHz may be in the near field. A GTEM cell, a compact, tapered chamber that sits on the bench, gives a better correlation, well characterised against chamber results up to about 1 GHz, but it is limited by the size of the product that fits inside it. Either way, the rule is the same: give yourself about 6 dB of margin, to lower the risk of failing in a real chamber. Correlation to formal testing is typically ±6 to 10 dB for frequencies above 100 MHz, which is why the margin matters more than the number.
The 6 dB margin is the number that keeps coming up. The lab flags anything within 3 dB of the limit. The pre-compliance guides say to keep every peak more than 6 dB below the limit before booking the lab.
The peak-scan rule says to switch to a slow quasi-peak measurement for every peak within 6 dB of the limit. It is not a standard. It is a rule of thumb, from several sources at once, and it is the difference between a bench result the team can trust and one it cannot.
The bench also has a rhythm the lab does not. A bench scan costs no lab time, so it can be repeated after every layout change, every ferrite, and every firmware build, as long as every scan is taken the same way. The same setup, the same distance, the same detector, the same baseline trace saved on the analyser.
The scan is a habit, not an event. That is the whole point: the lab is visited once, at the end, and the bench is used a hundred times along the way.
What fails, and where it comes from
Failures rank by frequency roughly as follows, based on aggregated data from EMC test labs. Radiated emissions is first, at 40 to 50% of failures, partially detectable in pre-compliance, because near-field probes identify the sources. Conducted emissions is second, at 20 to 25%, detectable, with a LISN and a spectrum analyser, and good correlation.
ESD immunity is third, at 10 to 15%, partially detectable, with an ESD gun on the bench. Radiated immunity is fourth, at 5 to 10%, not detectable without an anechoic chamber. EFT and burst is fifth, at 3 to 5%, it needs an EFT generator.
Surge immunity is sixth, at 2 to 4%, it needs a surge generator. The first two, radiated and conducted emissions, are two-thirds of all failures, and both are visible on the bench. The last three are the ones the bench cannot see, and they are the reason the lab still exists.
Each failure has a signature, and the signature names the source. A clock oscillator's harmonics appear at exact multiples of the crystal frequency, 25, 50, 75, 100 MHz. A USB clock shows at 48 MHz and its harmonics.
An Ethernet PHY shows at 125 MHz and its harmonics. A switching power supply shows at 100 to 500 kHz and its harmonics, on the conducted measurement. A broadband hump at 50 to 200 MHz is switch-node ringing in the SMPS.
Emissions that appear only when the display is active point at the display interface, LVDS, eDP, HDMI. Emissions that change with cable length mean the cable is acting as the antenna. Emissions that change with the device's orientation on the turntable mean the radiation is coming from an enclosure aperture.
Below 200 MHz, cables are the dominant radiator. Common-mode current on I/O cables, USB, HDMI, Ethernet, the power cord, couples from internal circuits through inadequate filtering at the connector. Missing or poorly grounded cable shields make it worse, and a shield grounded at only one end provides zero common-mode suppression.
From Article #19: common-mode radiation accounts for 80 to 90% of emission failures, and 5 µA of common-mode current at 300 MHz on a 1-metre cable produces 629 µV/m, more than three times the 200 µV/m Class B limit at that frequency, about 10 dB over it. Five microamps. That is the scale of the problem: a current too small to measure on a multimeter, on a cable too short to notice, is enough to fail the test.
Above 200 MHz, enclosure apertures and PCB slots dominate. Ventilation slots, seams between enclosure halves, and display cutouts radiate as slot antennas. A 15 cm slot resonates at roughly 1 GHz. The slot does not need to be long, it needs to be the right length for the frequency, and at the frequencies a 48 MHz clock produces, a 15 cm slot is exactly the right length.
Clock harmonics are the most predictable failure of all. A 25 MHz crystal with a 1 ns rise time generates harmonics past 300 MHz. USB clocks at 48 MHz, Ethernet PHY clocks at 125 MHz, and DDR clocks at 133 to 200 MHz are frequent offenders. The team knows its clock frequencies before it ever goes to the lab, it can calculate the harmonics and check whether any of them land in bands where the margin is thin. That calculation is the demo below.
The fix ladder
When a failure is found, the fixes come in three rungs, in order of cost, and the rung matters more than the fix.
Component-level fixes cost $0.01 to $5 per unit and take 1 to 3 days. A ferrite bead on an I/O line gives 5 to 15 dB above 30 MHz. A snap-on ferrite core on an external cable gives 3 to 10 dB. Adding or increasing an X capacitor gives 10 to 20 dB at the switching fundamental; adding or increasing Y capacitors gives 10 to 20 dB of common-mode reduction, limited by leakage current requirements. A TVS diode at a connector is the difference between passing and failing ESD. A series resistor on a clock output gives 3 to 8 dB. A bypass capacitor of 100 pF to 1 nF at a connector gives 5 to 15 dB above 100 MHz. These are the first things to try, and they can be applied to existing boards with rework or minor BOM changes. A ferrite bead costs pennies and takes a day.
Board-level fixes cost $2,000 to $15,000 for a respin and take 2 to 4 weeks. A continuous ground plane, with no splits or gaps, gives 10 to 20 dB against a fragmented plane. Minimising high-speed signal loop area gives 6 to 15 dB per halving of the loop area, and from Article #19, the field from a differential-mode loop grows with the square of the frequency, so the loop area is the lever that matters most at the harmonics. Moving the EMI filter to the connector edge of the board gives 10 to 20 dB. Ground stitching vias around high-speed traces give 3 to 10 dB. A spread-spectrum clock generator gives 6 to 10 dB of peak reduction. These are the fixes that need a new board, and a new board needs new samples, and new samples need a new lab booking.
Enclosure-level fixes cost $1,000 to $10,000 in tooling and take 2 to 6 weeks. A conductive gasket on the enclosure seams gives 10 to 30 dB. A shield can over a noisy IC gives 15 to 30 dB if it is properly grounded. Conductive paint on the inside of a plastic enclosure gives 20 to 40 dB. Multiple small ventilation holes instead of long slots give 10 to 20 dB. These are the fixes that touch the tooling, and tooling is the slowest thing on the list.
The ladder is the point. A failure found on the bench is usually fixed on the first rung, pennies and a day. The same failure found in the lab is usually fixed on the second or third rung, a respin or new tooling, weeks, and a retest slot four to six weeks away. The fix does not get more expensive because the failure got worse. It gets more expensive because the failure was found later.
There is also a fix that costs nothing at all, and it is the cheapest item in the whole article: design the EMI filter first, not last. Every product with a switching power supply needs a properly designed input filter at the AC entry point, a common-mode choke, X capacitors, Y capacitors, placed right at the connector. Do not leave it for cleanup after the first test failure.
And use a continuous ground plane: one layout practice, 10 to 20 dB of radiated emissions margin compared with split or fragmented planes.
The three builds
There are three ways to get to the chamber, and they differ in one thing: when the failure is found.
The walk-in. The team designs the board, builds the samples, and books the lab. No pre-compliance scan, no bench, no rehearsal. The lab is the first time anyone measures the emissions. About half the time, the product fails, and each failure costs $5,000 to $30,000 and 2 to 6 weeks, plus 90 days to get back in front of the chamber. The launch date becomes the retest date. The walk-in is the most common build, and it is the most expensive one.
The rented bench. The team books a half-day pre-scan at a lab's reduced pre-compliance rate, or rents a bench for $500 to $2,000 a day. The scan finds the major issues, the clock harmonics, the cable, the conducted noise, and the fixes are component-level: a ferrite bead, a capacitor, a resistor, a day of rework. Pre-compliance drops first-time failure rates from roughly 50% to under 10%. The $500 to $2,000 spent here saves $5,000 to $30,000 in retest costs. This is the build for a team that has one product and one launch.
The owned bench. The team buys the setup, $3,000 to $8,000 for the basic bench, $8,000 to $15,000 for the budget setup that catches about 70% of the issues that would otherwise fail formal testing, and runs the scan after every layout change, every ferrite, and every firmware build. The scan is a habit, not an event. The deliverable is not a certificate. It is one Integrated Change List: filter values, ground-plane revisions, cable shielding, firmware mitigations, mechanical reinforcement, each item tied to a measured margin, so the team knows what to change and how much headroom it buys. This is the build for a team that ships more than one product, because the bench is bought once and the lab is paid for every time.
The three builds are not three budgets. They are three answers to the same question: when do you want to find out? The walk-in finds out in the chamber, at $5,000 to $30,000 a lesson. The rented bench finds out for $500 to $2,000. The owned bench finds out for the price of the setup, and then for free, forever.
Example Labs is a made-up company we use for worked examples.
DEMO
The hub is the field sensor hub from the cluster: a 100 × 60 mm, 4-layer board with 120 nets and 114 test points, an STM32H7 running at 48 MHz, a BLE radio at 2.4 GHz, a sensor on a 3.0 V rail, in a 100 × 60 × 40 mm enclosure with an aluminium base, a polycarbonate lid, and 8 × M4 screws. It is battery-powered, on a CR2032, so it has no mains port and no conducted emissions test, the scan that matters is radiated.
The hub is going to the lab for FCC Part 15B and CE, and the team has to decide whether to scan it first.
Step 1, the harmonics the team already knows. The clock is 48 MHz, so its harmonics are 48, 96, 144, 192, 240, 288, 336 MHz, and on past the knee. The knee frequency, the frequency above which the harmonics start falling fast, is about 0.35 divided by the rise time. For a 1 ns rise time, the knee is 350 MHz. So the strong harmonics are the ones up to 336 MHz: seven of them, and the 7th harmonic at 336 MHz is the last strong one.
The FCC Class B limit at 3 metres steps as the frequency rises: 40 dBµV/m from 30 to 88 MHz, 43.5 dBµV/m from 88 to 216 MHz, and 46 dBµV/m from 216 to 960 MHz. The limit rises by 6 dB from the fundamental's band to the 7th harmonic's band. But the field from a differential-mode loop grows with the square of the frequency, and for a square-wave clock, the harmonic current falls as 1 over the harmonic number, so the field at the nth harmonic is n times the field at the fundamental.
The 3rd harmonic is 3 times the fundamental, or 9.5 dB louder. The 5th is 5 times, or 14.0 dB louder. The 7th is 7 times, or 16.9 dB louder.
The limit rises 6 dB. The field rises 16.9 dB. The margin shrinks by about 11 dB across the harmonics, and the harmonics are where the product fails.
Step 2, the peak. Suppose the fundamental sits 5 dB below its limit, at 35 dBµV/m against a 40 dBµV/m limit. That is the article's own worked example, labelled, a board that looks fine at its clock frequency. Then the 3rd harmonic at 144 MHz sits at 44.5 dBµV/m against a 43.5 limit, 1.0 dB over. The 5th at 240 MHz sits at 49.0 against 46, 3.0 dB over. The 7th at 336 MHz sits at 51.9 against 46, 5.9 dB over, about 6 dB over the limit. The fundamental passes. The 7th harmonic fails by 6 dB. And 336 MHz is above 200 MHz, where enclosure apertures and PCB slots dominate, so the frequency-to-source guide says to check the enclosure first: the seam between the aluminium base and the polycarbonate lid, and the lid's fastener spacing, before touching the board. If the emissions change when the device is rotated on the turntable, the radiation is coming from an aperture, and the fix is in the enclosure, not the circuit.
Step 3, what the bench sees. The near-field probe finds the source in minutes: the 48 MHz clock trace and its return, and the 336 MHz peak on the probe at the lid seam. The probe does not give the absolute level, it gives the relative reading, and the five highest peaks are the ones that correlate with formal failures. The radiated pre-scan, with an antenna a metre away in a quiet room, gives the approximate level, within 6 to 15 dB. The rule is to keep every peak more than 6 dB below the limit before booking the lab, and the 336 MHz peak is 6 dB over it, so the bench has done its job: it found the failure, and it found it while the fix is still cheap.
Step 4, the fix. The first rung is a ferrite bead on the I/O line, 5 to 15 dB above 30 MHz, pennies, a day, or a series resistor on the clock output, 3 to 8 dB. Suppose the bead gives 12 dB at 336 MHz, the same reduction a real pre-compliance case achieved on a 120 MHz peak that was 6 dB over the Class B limit. The 7th harmonic falls from 51.9 to 39.9 dBµV/m against the 46 limit, 6.1 dB below it. That is exactly the margin the lab wants: more than the 3 dB it flags as a concern, and the 6 dB the pre-compliance guides ask for. The fix cost pennies and a day. The alternative, found in the lab instead of on the bench, would have been a board respin at $2,000 to $15,000 and 2 to 4 weeks, or a conductive gasket on the lid seam at $1,000 to $10,000 in tooling and 2 to 6 weeks.
Step 5, the total. The formal test for the hub, using a pre-certified BLE module, runs $5,000 to $10,000; a custom transmitter would run $8,000 to $15,000. CE immunity adds $3,000 to $8,000. The expected cost of the retest, without pre-compliance, is the 50% failure rate times the $17,500 midpoint of the $5,000 to $30,000 retest range: $8,750. With pre-compliance, the failure rate falls under 10%, so the expected retest cost is $1,750, plus a $300 to $800 scan, for a total of $2,300, a saving of $6,450; or plus a rented bench at $500 to $2,000 a day, for a total of $3,000, a saving of $5,750. And the schedule: a failed cycle costs 90 days to get back in front of the chamber. A pre-compliance scan costs 1 to 5 days. The bench is bought once. The lab is paid for every time.
The decision test
Five questions, answered before the lab is booked.
- Do you know your clock frequencies and their harmonics? The hub's 48 MHz clock makes harmonics at 96, 144, 192, 240, 288, and 336 MHz, and the 7th harmonic is the one that fails. If the team cannot list its harmonics, it cannot predict its failures.
- Do you know which class you are testing to, and why? If there is any chance the product ends up in a home, it needs Class B, and a pass at Class A with 6 dB of margin fails Class B. The class is chosen by where the product is used, not by the test budget.
- Have you scanned the board, and do the peaks sit more than 6 dB below the limit? Near-field probes for the sources, a LISN for conducted emissions if the product has a mains port, an antenna or a GTEM cell for the approximate radiated level. If a peak is within 3 dB of the limit, the lab will flag it. If it is within 6 dB, the bench has not finished its job.
- Do you have a fix ladder priced? Component-level at pennies and a day, board-level at $2,000 to $15,000 and 2 to 4 weeks, enclosure-level at $1,000 to $10,000 and 2 to 6 weeks. A failure with a priced ladder is a decision. A failure without one is a delay.
- Is the lab booking 4 to 6 weeks ahead, and is the retest slot already asked about? The big labs book out 4 to 8 weeks in advance, and 6 weeks or longer in peak seasons. A 6-week wait for a retest slot after a failure can destroy a launch timeline, and the time to ask about it is before the first test, not after the first failure.
Objections
"Pre-compliance is just a guess, the bench isn't the chamber." True, and it is the point. The near-field probe gives relative levels, not absolute ones, and the correlation to formal testing is typically ±6 to 10 dB for frequencies above 100 MHz. That is why the margin rule exists: keep every peak more than 6 dB below the limit, and the bench's uncertainty stops mattering. Conducted emissions correlate within 3 to 6 dB, because the setup is the same in both places. The bench does not certify. It finds the peaks so the chamber does not have to.
"We'll just fix it at the lab, they have debug time." Some labs do offer on-site troubleshooting, where the team can add ferrites, change capacitor values, and apply shielding tape and re-scan the affected range. That is the fastest path to a fix, for a failure that can be fixed with rework. But a failure that needs a board respin is $2,000 to $15,000 and 2 to 4 weeks, and the retest slot is 4 to 6 weeks away. The lab is the most expensive place to discover a problem the team could have found on its own bench.
"Our product is battery-powered, so conducted emissions don't apply." Correct, with no mains port, there is no LISN and no conducted test. But radiated emissions still apply, and the battery cable and any I/O cable are still antennas. The hub's scan is radiated only: near-field probes for the sources and an antenna or a GTEM cell for the approximate level. The conducted section of the article is for mains-powered products. The radiated section is for everything.
"We use a pre-certified radio module, so we're covered." The module is certified. The host is not. A device using a pre-certified RF module still needs unintentional emissions testing plus integration verification, at $5,000 to $10,000, and the module's certificate does not cover the host's clock harmonics. The radio is one radiator. The board is another.
"We'll test to Class A, it's cheaper." The class is chosen by where the product is used, not by the test budget. If there is any chance it ends up in a home, it needs Class B, and a pass at Class A with 6 dB of margin fails Class B. The adapter that passed at Class A with 5 dB of margin was sent back the same week a retail chain asked for Class B evidence. The cheaper test is the one that does not have to be repeated.
Sources
- MarkReady, "EMC Testing: The Complete Guide"
What we used
(markready.io/guides/emc-testing, opened in full, 2026-11-20): "About half of products fail EMC testing on the first attempt. That number comes from Intertek Labs, and it lines up with what every EMC test lab reports. The failures are expensive: each retest cycle adds $5,000 to $30,000 and 2 to 6 weeks to your timeline." "Class B is for residential environments. Stricter limits. Required for any consumer product. Class A is for commercial and industrial environments. Limits are roughly 10 dB more relaxed. Requires a warning label stating the device may cause interference in residential settings." "If there is any chance your product ends up in someone's home, it needs to meet Class B. The 10 dB gap between classes is significant, a product that passes Class A with 6 dB of margin will fail Class B." "Emissions testing asks: does this device interfere with other equipment? Every market requires this. The FCC, CE (EU), and ISED (Canada) all mandate emissions testing before a product can be sold." "Immunity testing asks: can this device withstand interference from the outside world? The EU requires immunity testing under the EMC Directive and the Radio Equipment Directive (RED). The FCC and ISED do not." Typical cost: emissions $2,000–$5,000, lab time 1–2 days; immunity $3,000–$8,000 additional, lab time 1–4 days. "Radiated emissions is where most products fail. The test measures electromagnetic fields radiating from your device and its cables at a calibrated distance (3 meters for FCC, 10 meters for CISPR 32 with a 3-meter option). The lab places your device on a turntable in a semi-anechoic chamber, rotates it through 360 degrees, and scans an antenna across the full frequency range while varying height. Results are compared against regulatory limit lines using a quasi-peak detector below 1 GHz and peak/average detectors above." "Conducted emissions measures RF noise your device pushes back onto the AC power cord. A Line Impedance Stabilization Network (LISN) presents a standardized 50-ohm impedance to the power input while the lab measures noise voltage on each conductor (line and neutral) from 150 kHz to 30 MHz. Switching power supply harmonics are the dominant source of conducted emissions failures." Immunity levels: ESD ±4 kV contact, ±8 kV air (IEC 61000-4-2); radiated immunity 3–10 V/m, 80 MHz–6 GHz (IEC 61000-4-3); EFT/Burst ±1 kV power, ±0.5 kV signal (IEC 61000-4-4); surge ±1 kV line-line, ±2 kV line-ground (IEC 61000-4-5); conducted immunity 3 V, 150 kHz–80 MHz (IEC 61000-4-6); voltage dips 0%, 40%, 70% of nominal (IEC 61000-4-11); power frequency magnetic field 3 A/m (IEC 61000-4-8). "ESD is the immunity test that causes the most failures, unprotected I/O ports, connectors without TVS diodes, and enclosure gaps are the usual culprits." "A product that passes CISPR 32 Class B emissions will almost always pass FCC Part 15B as well, since the limits are virtually identical for conducted emissions and close for radiated. The reverse is also generally true. But test reports are not directly transferable, different measurement methods, different documentation formats." "The real gap between markets is immunity. A product designed and tested only for FCC has never been exposed to ESD, surge, EFT, or radiated immunity testing. When that product goes through CE testing for the first time, immunity failures are common because nobody thought about susceptibility during the design." "Pre-compliance drops first-time failure rates from roughly 50% to under 10%. A basic pre-compliance setup (spectrum analyzer, LISN, near-field probes) costs $3,000 to $8,000 to buy. Even renting a pre-compliance bench for $500 to $2,000 per day pays for itself after one avoided retest." Correlation table: conducted emissions, LISN + spectrum analyzer ($2K–$5K), "Excellent, within 3–6 dB"; radiated emissions (relative), near-field probes ($200–$500), "Identifies sources, not absolute levels"; radiated emissions (approximate levels), antennas + outdoor space or GTEM cell ($5K–$15K), "Within 6–15 dB depending on environment". "What you cannot catch: immunity performance (ESD, surge, radiated immunity all require specialized generators and a controlled environment), edge cases within 3–6 dB of the limit line, and spurious emissions above your analyzer's frequency range." "The big names (Intertek, SGS, UL, TUV) may book out 4 to 8 weeks in advance, especially during peak seasons (January through March after CES, July through September before holiday product launches)." "Results are plotted against regulatory limit lines. You need margin, not just a pass. A measurement 1 dB below the limit is technically passing but leaves no room for production variation, temperature drift, or measurement uncertainty. Most labs will flag anything within 3 dB of the limit as a concern." "Cost depends on the fix. Adding a ferrite bead costs pennies per unit and takes a day. Redesigning the power supply filter means a board respin at $2,000 to $15,000 and 2 to 4 weeks." "A 6-week wait for a retest slot after a failure can destroy a product launch timeline. Ask about retest booking before you commit." "The bottleneck is rarely the testing itself. Lab queue time, documentation prep, and deficiency rounds during TCB review eat more calendar time than the measurements. During peak seasons (post-CES in Q1 and pre-holiday in Q3), lab booking lead times can stretch to 6 weeks or longer. Book 4 to 6 weeks ahead during these periods." Costs by device type: simple digital device (no radio) $2,000–$5,000; device using a pre-certified RF module $5,000–$10,000; custom WiFi/BT transmitter $8,000–$15,000; multi-radio device (WiFi + BT + cellular) $12,000–$25,000. Adding CE: immunity $3,000–$8,000 (1–4 days); harmonics + flicker $500–$1,500 (half a day); safety $3,000–$8,000 (2–5 days); RED radio tests $2,000–$5,000 (1–2 days). Lab rates: radiated emissions ~$102/hr; radiated immunity ~$136/hr; other tests ~$85/hr; full-day $2,000–$3,000/day. Timeline: pre-compliance 1–5 days; emissions 1–2 days; immunity 1–4 days; report 3–10 business days; rework + retest 2–6 weeks per cycle; total simple FCC-only 3–6 weeks; FCC + CE + ISED 4–10 weeks. Failure ranking: radiated emissions 40–50% (partially detectable, near-field probes identify sources); conducted emissions 20–25% (yes, LISN + spectrum analyzer, good correlation); ESD immunity 10–15% (partially, ESD gun on bench); radiated immunity 5–10% (no, requires anechoic chamber); EFT/Burst 3–5% (requires EFT generator); surge immunity 2–4% (requires surge generator). "A typical failure shows 1 to 3 discrete frequency spikes exceeding the limit by 3 to 15 dB, usually at exact harmonics of a clock or switching frequency." "Below 200 MHz, cables are the dominant radiator. Common-mode current on I/O cables (USB, HDMI, Ethernet, power cord) couples from internal circuits through inadequate filtering at the connector. Missing or poorly grounded cable shields make it worse, a shield grounded at only one end provides zero common-mode suppression." "Above 200 MHz, enclosure apertures and PCB slots dominate. Ventilation slots, seams between enclosure halves, and display cutouts radiate as slot antennas. A 15 cm slot resonates at roughly 1 GHz." "Clock harmonics are the most predictable failure. A 25 MHz crystal with 1 ns rise time generates harmonics past 300 MHz. USB clocks at 48 MHz, Ethernet PHY clocks at 125 MHz, and DDR clocks at 133 to 200 MHz are frequent offenders. You know your clock frequencies before you ever go to the lab, calculate the harmonics and check whether any land in bands where your margin is thin." "The fundamental switching frequency (typically 65 kHz to 500 kHz) and its harmonics appear directly in the conducted emissions measurement. A missing or undersized EMI input filter is the most frequent root cause. The fix: a properly designed input filter with a common-mode choke, X capacitors (line-to-line for differential-mode noise), and Y capacitors (line-to-ground for common-mode noise). The filter must be placed at the AC entry point, if unfiltered AC traces run across the board before reaching the filter, those traces re-radiate and the filter cannot help." "The lab applies calibrated discharges (up to ±8 kV contact, ±15 kV air) to every user-accessible surface and connector. The typical failure: the device resets, locks up, or loses communication when discharged near an I/O port. Root causes include unprotected connectors (no TVS diodes), ESD protection devices placed too far from the connector pin (they need to be within 5 to 10 mm), and no low-impedance path from the connector shell to chassis ground." Frequency-to-source guide: exact multiples of crystal frequency (25, 50, 75, 100 MHz), clock oscillator harmonics; 48 MHz and harmonics, USB clock; 125 MHz and harmonics, Ethernet RGMII/PHY; 100–500 kHz and harmonics (conducted), SMPS switching fundamental; broadband hump at 50–200 MHz, switch node ringing in SMPS; emissions only when display is active, display interface (LVDS, eDP, HDMI); emissions change with cable length, cable is acting as the antenna; emissions change with device orientation on turntable, enclosure aperture radiation. Component-level fixes ($0.01–$5 per unit, 1–3 days): ferrite bead on I/O line 5–15 dB above 30 MHz; snap-on ferrite core on cable 3–10 dB; X capacitor 10–20 dB at switching fundamental; Y capacitors 10–20 dB (limited by leakage current); TVS diode at connector, pass/fail difference; series resistor on clock output 3–8 dB; bypass cap (100 pF–1 nF) at connector 5–15 dB above 100 MHz. Board-level fixes ($2K–$15K respin, 2–4 weeks): continuous ground plane 10–20 dB vs fragmented plane; minimise high-speed signal loop area 6–15 dB per halving of loop area; move EMI filter to connector edge 10–20 dB; ground stitching vias 3–10 dB; spread-spectrum clock generator 6–10 dB peak reduction. Enclosure-level fixes ($1K–$10K tooling, 2–6 weeks): conductive gasket on enclosure seams 10–30 dB; shield can over noisy IC 15–30 dB if properly grounded; conductive paint on plastic enclosure interior 20–40 dB; multiple small ventilation holes instead of long slots 10–20 dB. "The 50% first-pass failure rate is not inevitable. Teams that follow these practices consistently push their failure rate below 10%." "Run pre-compliance testing. A LISN and spectrum analyzer for conducted emissions. Near-field probes for source identification. Even a half-day pre-scan at a lab's reduced pre-compliance rate catches the major issues. The $500 to $2,000 you spend here saves $5,000 to $30,000 in retest costs." "Design the EMI filter first, not last. Every product with a switching power supply needs a properly designed input filter at the AC entry point. Common-mode choke, X capacitors, Y capacitors, placed right at the connector. Do not leave this for 'cleanup' after the first test failure." "Use continuous ground planes. An unbroken ground plane on your PCB minimizes current loop area and provides low-impedance return paths. One layout practice, 10 to 20 dB of radiated emissions margin compared to split or fragmented planes." "If your product contains a microprocessor, digital circuitry operating above 9 kHz, or a radio transmitter of any kind, it needs EMC testing." "This step is optional but has the best ROI of anything in the compliance process. Pre-compliance means running informal tests during development using your own equipment or rented bench time at a lab. The goal is to find and fix the major issues before you commit to $5,000–$15,000 in formal testing." Checked in full via browser.open, 2026-11-20
- Ideambox, "EMC design and pre-compliance"
What we used
(IDB-EMC-016, revision 1.0, May 2026, ideambox.com/downloads/emc-design-and-pre-compliance/): "EMC (Electromagnetic Compatibility) failures account for ~40 % of first-pass lab failures in consumer electronics." "The fix is design discipline at schematic + PCB layout, plus a pre-compliance scan workflow that catches issues at $300–800 per scan instead of $3 000–8 000 per failed lab booking." "Catching EMC issues in-house before formal lab testing saves $3 000–8 000 per failure cycle." Equipment: spectrum analyser (9 kHz, 6 GHz) $5 000–25 000; near-field probe set (H + E field) $300–1 000; current probe (clamp-on) $500–2 000; RF turntable or test chamber $5 000–50 000; ESD gun (8 kV contact / 15 kV air) $2 000–5 000; surge generator $5 000–15 000; spectrum analyser software $1 000–5 000. "A budget pre-compliance setup (~$8 000–15 000) catches 70 % of issues that would otherwise fail formal lab testing." Shielding strategy and cost: local shielding cans (over RF / DC-DC) $0.20–2.00 each, 20–60 dB at GHz; full enclosure shield 40–80 dB; EMI gaskets at openings $0.50–5.00 each; conductive paint (on plastic enclosure) 10–30 dB; metallised plastic enclosure (PVD/sputter) adds 10–20% to part cost, 30–50 dB; plated/embedded ferrite cable shield per cable, 10–30 dB on cables. Checked via browser.search, 2026-11-20
- Wipro, "Reduce Certification Risk: 6 Pre-Compliance Testing Stages"
What we used
(wipro.com/engineering/articles/pre-compliance-testing-the-fastest-path-to-product-certification/): "Depending on product complexity, accredited EMC labs place CE/FCC certification between $1,000 and $20,000, with chamber time running $1,500 to $2,500 a day. In high-reliability markets, automotive, medical, defense, that rises to between $13,600 and $34,000 per cycle, and most teams carry a further 10% to 30% buffer for design changes after a failure." "A failed cycle means waiting four to six weeks for a chamber slot. It means paying for a design review, spinning up a new board, and building fresh samples. By the time the team is back in front of the chamber, 90 days have gone by." "Compliance testing is binary, pass or fail; report goes to a certification body. Pre-compliance is a diagnostic. It shows how close to the boundary a product is, where it is leaking, and what design change will bring it under threshold." "Stage 2, Emissions Sweep: Characterize radiated and conducted emissions on the first physical prototype, sweeping 9 kHz to 6 GHz against CISPR 32, FCC Part 15, or equivalent regional standard. This EMC testing step is not a verdict, but a spectrum trace with each peak within 10 dB of the limit annotated with probable root cause." "Across all six stages, the deliverable is not a certificate. It is one Integrated Change List (ICL): filter values, ground-plane revisions, cable shielding, firmware mitigations, mechanical reinforcement, each item tied to a measured margin, so the team knows what to change and how much headroom it buys." "If you're 3 to 6 months from certification, run a pre-compliance diagnostic now, it is the difference between one clean cycle and three painful ones." "Most of the EMC and safety failures have seen downstream trace back to decisions visible at this stage, component placement, ground-plane partitioning, connector position, enclosure material." Checked via browser.search, 2026-11-20
- Galois, "EMC Pre-Compliance Testing on Your Bench"
What we used
(galoislabs.ai/blog/emc-pre-compliance-testing): "EMC pre-compliance testing is measuring your product's emissions on your own bench before an accredited lab does: near-field probes to find sources, a LISN and spectrum analyzer for conducted emissions, and an antenna for a radiated pre-scan." "First attempts fail often: Intertek reports EMC failure rates around 50% on first tests in its labs, and 5 to 7% on retests (Source: Intertek white paper)." "Tektronix, which sells pre-compliance equipment, puts an external test lab at $1,000 to $10,000 a day." "A bench scan costs no lab time, so you can repeat it after every layout change, ferrite and firmware build, as long as every scan is taken the same way." Checked via browser.search, 2026-11-20
- Garrett Technologies, "Boosting EMC/ESD Pass Rates by Pre-Testing" (garrett-tech.com)
What we used
"According to Intertek, 50% of products fail EMC testing the first time, with Medical Devices slightly more successful with a 40% failure rate. By the second round, success is 93% to 97%." "Generally, the pre-testing itself is less expensive than two rounds of full compliance testing." Checked via browser.search, 2026-11-20
- Rohde & Schwarz, "Fundamentals of EMI Debugging & Pre-Compliance" (Elizabeth McKenna)
What we used
"Failing compliance testing is not uncommon, Failure rate is ~70-90%." "Pre-compliance testing is done by device manufacturers, in their labs, before compliance testing, Increases chances of passing, Saves considerable time and money if problems are caught (and corrected) early in the design cycle." Checked via browser.search, 2026-11-20
- EDN, "Troubleshooting EMI on your bench top" (Henry Ott's method)
What we used
"If the measurement test distance is set at 1m, you may use the adjusted radiated emissions limits to estimate how the emissions might compare to 3m or 10m limits. Caution: Because the measured frequencies below 150 MHz may be in the near field, the results are ONLY a rough guess. Experience by Ott and Curtis has shown that adding an additional 6 dB on top of the extrapolated limits is often more accurate." "Near field probing with E- or H-field probes will help identify potential sources of EMI on, or within your product. However, not all structures or components will act as efficient antennas and therefore, are not ultimately causing emissions failures." Checked via browser.search, 2026-11-20
- EE Times, "Questions On EMC Pre-Compliance Testing for Radiated Emissions"
What we used
"I've been very successful using conference rooms or basements to set up a simple 3m measurement range for confirming pass/fail data. Of course, I like to give myself about 6dB of margin, just to lower the risk of failing in a real chamber." "Assemble the equipment and semi-anechoic chamber required to perform your own radiated emissions testing (can cost $500k to $3M, depending on 3m or 10m chamber), or have a third-party test facility run the compliance testing for you." Checked via browser.search, 2026-11-20
- AtlasPCB, "PCB Design for EMC Pre-Compliance"
What we used
(atlaspcb.com/blog/emc-pre-compliance-pcb-near-field-probing-fcc-cispr/): "Formal EMC testing in an accredited semi-anechoic chamber costs $3,000–$10,000 per test session." "Near-field probing gives relative measurements (dBµV into 50Ω), not absolute far-field levels. Use these empirically-derived guidelines: If a near-field H-probe reading exceeds -10 dBm at 3mm distance, the associated trace/IC will likely contribute to a far-field limit exceedance; A 20 dB reduction in near-field reading typically provides 15–20 dB far-field improvement; Focus on the 5 highest peaks in your near-field scan, these correlate to formal test failures." "Before formal testing: All near-field peaks should be > 6 dB below your informal correlation limit." "Correlation to formal testing is typically ±6–10 dB for frequencies above 100 MHz. The primary value is comparative: if you fix the hottest spots on the PCB (highest near-field readings), far-field emissions will improve proportionally. Most engineers report 85–95% first-pass rate at formal testing." "Spread-spectrum clocking (SSC): 0.5–1.5% modulation reduces clock harmonics by 6–12 dB with minimal SI impact." "The single most important EMC design rule: never break the ground plane beneath clock signals or high-frequency traces." Checked via browser.search, 2026-11-20
- SZCLab, "EMC Class A vs Class B: Which One Does Your Product Need?" (szclab.com, 29 October 2026)
What we used
"Two limit sets, one standard, ten decibels apart. The class is chosen by where the product is used, and it is much cheaper to choose it before the tooling is cut." "A 65 W desktop adapter passed its emissions test at 10 m with about 5 dB of margin, and the report was issued against Class A limits. Eight months later the same adapter was quoted to a US retail chain. Their compliance team asked for Class B evidence and sent the file back the same week. Nothing in the report was wrong. The laboratory measured against the limit set it was given. The decision that caused the problem was made much earlier, when the class was chosen to fit the test budget rather than the channel the product would eventually be sold through." "Class B covers equipment used in residential environments: homes, apartments, hotel rooms, anywhere a television or radio receiver sits within a few metres. Class A covers everything else in scope, commercial, industrial and light-industrial premises, on the assumption that it will not be used where people live. That is the whole distinction. A Class B product is not a better product. It is a product that emits roughly 10 dB less, because it has to coexist with broadcast reception in a living room." "The class is not a certificate, a mark, or an approval issued by an authority. It is a limit set inside a standard, recorded on the test report, and in the United States mirrored by a statement that must appear in the user manual." CISPR 32 / EN 55032: radiated 30–230 MHz at 10 m, Class A 40 dBµV/m, Class B 30 dBµV/m; conducted AC mains 150 kHz–30 MHz quasi-peak, Class A 79 dBµV (150–500 kHz), 73 dBµV (0.5–30 MHz); Class B 66→56 dBµV (150–500 kHz), 56 dBµV (0.5–5 MHz), 60 dBµV (5–30 MHz). Checked via browser.search, 2026-11-20
- Krono Labs, "FCC Class A vs Class B (Part 15): Limits, Labels, SDoC" (krono-labs.com)
What we used
"Compared at the same 3 m distance, Class B radiated limits are about 6–11 dB tighter than Class A, and Class B conducted limits are 13–23 dB tighter." "Conducted, 0.5–5 MHz: Class A allows 73 dBμV quasi-peak and 60 dBμV average, Class B only 56 and 46, a 17 dB and 14 dB tightening (§ 15.107). Radiated, 30–88 MHz: Class A is 90 μV/m (39.1 dBμV/m) at 10 m; Class B is 100 μV/m (40 dBμV/m) at 3 m, which is about 29.5 dBμV/m at 10 m, so roughly 9.5 dB tighter (§ 15.109). Above 960 MHz the gap narrows to about 6 dB." Checked via browser.search, 2026-11-20
- TPS Elektronik, "How to Pass CISPR 32 and FCC Part 15 Radiated Emissions Testing for Consumer Electronics" (tps-elektronik.com)
What we used
"CISPR 32 Class B limits apply to equipment intended for residential environments. For radiated emissions at 3 meters measurement distance, the limits are: 30–230 MHz: 40 dBμV/m quasi-peak; 230–1000 MHz: 47 dBμV/m quasi-peak. These limits are stricter than Class A (industrial) limits by approximately 10 dB." "FCC Part 15 Subpart B applies to unintentional radiators, including digital devices and peripheral equipment. The radiated emissions limits for Class B digital devices at 3 meters are harmonized with CISPR 22 and CISPR 32, with the same 40 dBμV/m (30–88 MHz), 43.5 dBμV/m (88–216 MHz), 46 dBμV/m (216–960 MHz), and 54 dBμV/m (above 960 MHz) quasi-peak limits." "The initial scan confirmed the customer's concern: a prominent emission at 120 MHz exceeded the CISPR 32 Class B limit by 6 dB. A second harmonic at 240 MHz also exceeded the limit. The emissions were present in both horizontal and vertical polarizations, suggesting a common-mode current source on the PCB or cables." "These modifications were implemented on the prototype and re-tested. The 120 MHz peak was reduced by 12 dB, bringing it 6 dB below the CISPR 32 Class B limit. The 240 MHz harmonic was also reduced below the limit." Checked via browser.search, 2026-11-20
- FabFlow, "Pre-Compliance EMC Testing Guide"
What we used
(fabflow.app/blog/pre-compliance-testing-guide): "The 5-phase workflow that catches 80% of emission problems on your bench." "Tier 1: The ₹10,000 Starter Kit, This catches 70% of EMC issues." "At the pre-compliance chamber: 1. Start with a peak scan (fast sweep) to find all emission peaks. 2. For each peak within 6 dB of the limit, switch to quasi-peak measurement." "A GTEM (Gigahertz Transverse Electromagnetic) cell costs ₹40,000–1,00,000 and sits on your bench. It's a tapered chamber that creates a uniform EM field. Place your device inside, connect to a spectrum analyzer, and measure emissions. A GTEM correlates reasonably well with OATS (Open Area Test Site) measurements for devices smaller than the cell's usable volume." Checked via browser.search, 2026-11-20
- Boards and Bits, "Your Homemade Gadget Is a Tiny Radio Station (and the FCC Will Find Out)" (boardsandbits.com)
What we used
"We put together a working pre-compliance setup for $187, and it caught problems on every single project we scanned." "RTL-SDR Blog V3 dongle, $35. This $35 software-defined radio covers 500 kHz to 1.7 GHz." "A set of near-field probe kits (H-field loop and E-field stub), $40 to $60." "A 20 dB preamp, $30." "Near-field probes tell you where the energy is coming from and let you compare before/after, they don't give you calibrated field-strength numbers." "The fix cost about $2. A small ferrite bead on the converter's output, plus a 100 nF ceramic cap placed directly across the load, knocked the worst spurs down by more than 15 dB on the near-field probe." Checked via browser.search, 2026-11-20
- LanPDT, "UL Certification Cost: Testing and Timelines" (lanpdt.com)
What we used
"Pre-compliance is informal testing on an engineering build, a half day of EMC scans, a drop test, a flammability check. It costs $1,000 to $5,000 and its whole purpose is to find failures while a PCB respin is still cheap." "Never skip pre-compliance on an electronic product. Radiated emissions failures at the accredited lab typically cost 6 to 10 weeks and a board revision." "Pre-compliance scanning at a local lab, typically $1,500 to $4,000 for a day of chamber time, catches emissions problems while the board can still be changed cheaply." "Plan for at least one re-test cycle on a first-time product." Checked via browser.search, 2026-11-20
- UL Solutions, "EMC Pre-Compliance Testing"
What we used
(ul.com/services/emc-pre-compliance-testing): "EMC pre-compliance testing works to identify potential problems early in your design process, before full-scale testing takes place." "Our EMC experts find that it's not unusual for new products to fail initial full-scale compliance testing." Checked via browser.search, 2026-11-20
- Cluster figures reused (not re-derived
What we used
Read the original articles before reusing): Article #19 (EMC design, differential-mode loop E = 1.32·A·I·F²/r; common-mode radiation accounts for 80–90% of emission failures; 5 µA at 300 MHz on a 1 m cable produces 629 µV/m, against a 200 µV/m Class B limit; a 4-layer board produces ~15 dB less radiation than a 2-layer board; FCC Part 15 §15.109 Class B at 3 m: 30–88 MHz 100 µV/m, 88–216 MHz 150, 216–960 MHz 200, above 960 MHz 500; CISPR 32 Class B at 3 m: 30–230 MHz 40 dBµV/m, 230–1,000 MHz 47; never route signals over split reference planes; even a 0.5 mm routing channel across a ground plane can raise emissions 20 dB at harmonics; 100 nF per IC power pin, target under 2 mm from the pin); Article #36 (FCC/CE wireless certification, the certificate the pre-compliance scan prepares for); Article #44 (SMPS noise, buck 24 V to 5 V, 2 A, 500 kHz, L = 10 µH, D = 0.2083, ΔIL = 0.792 A, ceramic ripple 13.0 mV vs electrolytic 88.2 mV, 10 nH at 2.4 × 10⁸ A/s = 2.40 V spike); Article #55 (power integrity, the hub's 48 MHz clock, sensor rail 3.0 V with a 150 mV budget, twin sag 0.36 V, rail 2.64 V; f_knee ≈ 0.35 / t_rise); Article #51 (grounding, thin trace 0.1 Ω, 1 A raises a chip's ground by 100 mV; the same trace has ~50 nH, X ≈ 3.1 Ω at 10 MHz; the plane moves the ground by 0.6% against the trace's 30%); Article #32 (design for test, hub board 100 × 60 mm, 4-layer, 120 nets, 114 test points); Article #40 (prototype failure, cost escalation 1×/5×/20×/50×/100×+, 4 of 10 hubs failed days 19–23, 40%); Article #41 (digital twin, predicted band 0.30–0.40 V, measured sag 0.36 V, 5 engineer-days vs 100+ field cost); Article #47 (enclosure materials, aluminium base, polycarbonate lid, hub shell 49.6 cm³); Article #49 (fasteners, lid held by 8 × M4 screws, opened 25 times over a 10-year life, labelled assumption); Article #50 (HALT, hub spec −20 °C to +60 °C, UOL +80 °C, self-heating 5 W = ~25 °C rise); Article #7 (enclosures, 5 W in the sealed box = ~25 °C rise; gaskets lose ~20% of sealing after 2 years of temperature cycling)