EMC, Noise, and Certification Basics
Your board radiates. The only questions are how much, at which frequencies, and whether a lab will let you sell it.
🎯 By the end you'll understand
The two mechanisms that put your board's noise into the air — small current loops radiating as differential-mode sources, and cables carrying common-mode current as accidental antennas — and why the second one wins by a wide margin. What a limit line actually is: the FCC Part 15 Subpart B Class B numbers at 3 m, what CISPR 32 and the EU's radio rules require instead, and why a battery-powered product still has to care. How to do pre-compliance with a near-field probe and a modest analyzer without lying to yourself about what those measurements mean. What a pre-certified module's grant does and does not cover, and what Indonesia's SDPPI type approval adds. And why Beacon passed on the bench and failed the moment someone plugged in a USB cable.
KEY INSIGHTThe cable is the antenna
A 35 × 25 mm board is a terrible antenna at 80 MHz — its dimensions are about a hundredth of a wavelength. A one-metre USB cable is an excellent one. Almost every radiated-emissions failure in a small product is the board driving the cable, not the board radiating on its own.
Beacon's pre-compliance session took an afternoon and produced two numbers that were seventeen decibels apart. Sitting on battery, on a foam block, running its normal one-report-per-minute cycle, the worst peak in the whole 30 MHz to 1 GHz sweep sat about thirteen decibels under the Class B limit — comfortable. Then someone plugged in a USB cable to charge it, because that is what you do with a device that has a USB-C port, and the peak at 78 MHz climbed above the limit line. Same board, same firmware, same room. The difference was a metre of copper hanging off the connector, and a few microamps of common-mode current riding on it.
That asymmetry is the organising fact of practical EMC. A small loop on a PCB radiates as a magnetic dipole, and its field strength scales with the loop's area and with the square of frequency — which means a well-laid-out board with small loops is quiet by construction, and the layout rules from Lesson 5.6 have already done most of the work. A cable, by contrast, is a resonant structure a metre long, and a metre is a quarter wavelength at about 75 MHz. It does not need much drive to reach the limit. A widely cited rule of thumb in the EMC literature is that a few microamps of common-mode current on a one-metre cable is enough to hit the FCC Class B radiated limit at three metres. Microamps. Your board has milliamps of high-frequency ground noise sloshing around; the only thing standing between that and a failure is how little of it becomes common-mode current on the cable.
Two questions answer most EMC problems. What loop is the current flowing in, and how big is it? — that governs what the board radiates directly, and it is fixed with layout, not with parts. What is driving common-mode current onto a cable? — that governs what the system radiates, and it is fixed with filtering right at the connector plus a ground plane that gives return currents nowhere else to go. Everything else, including most of the folklore about ferrite beads, is downstream of those two.
HOW IT WORKSLoops, edges, and cables: where the energy actually comes from
Differential mode: the loop is the radiator
Any current that flows out along one conductor and back along another forms a loop, and a loop carrying alternating current radiates. For a loop small compared with a wavelength, the far-field electric field strength is proportional to f2 × I × A, divided by distance — frequency squared, current, and loop area. Two consequences follow immediately and they are the reason Stage 5 spent so long on return paths. Halving the loop area halves the field: a 6 dB improvement for free, purely from where you put a trace. And doubling the frequency quadruples it: 12 dB worse, which is why the fifteenth harmonic of a switching regulator matters even though it carries a tiny fraction of the energy.
On Beacon the loops that matter are ranked, and the ranking is not intuitive. First is the buck converter's hot loop: input capacitor, high-side switch, low-side switch, back to the capacitor's ground. That loop carries a current that switches on and off in nanoseconds, so its spectrum extends to hundreds of megahertz even though the switching frequency is a couple of megahertz. Second is the module's supply loop, where the radio's 40 mA transmit bursts pull current with sharp envelopes. Third — and only third — are the signal loops: I2C at 400 kHz is slow enough that its harmonics fall off before they matter, and the WS2812B's data line is a 800 kHz stream with fast edges but almost no current.
Common mode: the cable becomes the radiator
Common-mode current is current that flows out along all the conductors of a cable together and returns through the surrounding environment — through capacitance to the bench, the room, and you. It has no matching return conductor, so there is no small loop; the loop is the size of the room. That makes an ordinarily unremarkable microamp into a serious radiator.
Where does it come from? From a voltage difference between your board's ground and the far end of the cable. Every millivolt of high-frequency noise across your ground plane — the inevitable result of switching currents flowing through a plane with finite inductance — appears as a drive voltage between the cable's shield or ground wire and the rest of the world. The cable, being roughly resonant, converts that drive into current very efficiently near its quarter-wave frequency and its odd multiples. A one-metre cable peaks near 75 MHz and again near 225 MHz, which is exactly where small-product emissions failures cluster.
A ferrite bead is a lossy inductor: it presents a specified impedance at a specified frequency (a "600 Ω at 100 MHz" bead really means it turns high-frequency energy into a few millidegrees of heat) and it does so to whatever current flows through it, differential or common. Put one in series with a USB data line and you attenuate the noise and the signal. A common-mode choke winds both conductors on the same core in opposite senses, so differential current produces cancelling flux and sees almost no inductance, while common-mode current produces adding flux and sees a large one. That is why a CM choke is the correct part on a data pair and a bead is not — and why a bead is perfectly correct on a power rail, where there is no differential signal to protect. Beacon has 22 Ω series resistors on D± (canon), which damp reflections at USB full speed and do exactly nothing for common mode.
DEFINITIONThe limit lines: what a regulator actually forbids
A limit line is a curve of maximum permitted field strength against frequency, measured with a specified detector at a specified distance in a specified test setup. Every part of that sentence carries weight — a measurement that changes any of those is not comparable to the limit.
In the United States, unintentional radiators — the digital circuitry, the switching regulator, everything that is not the radio — are governed by FCC Part 15, Subpart B. The Class B limits (residential use, which is what a device that ends up in an office or a workshop needs) for radiated emissions measured at 3 metres are:
| Frequency | Field strength (µV/m at 3 m) | Same in dBµV/m | What tends to live there |
|---|---|---|---|
| 30 – 88 MHz | 100 | 40.0 | Cable resonance for a 1 m lead — the strictest limit, on the frequency you are most likely to violate |
| 88 – 216 MHz | 150 | 43.5 | Cable harmonics, mid-order clock harmonics |
| 216 – 960 MHz | 200 | 46.0 | Clock harmonics, fast edges, switching-regulator ringing |
| above 960 MHz | 500 | 54.0 | High-order harmonics; also where the radio's own band sits under different rules |
Two conversions worth internalising. Field strength in dBµV/m is just 20 log10 of the value in µV/m, so 100 µV/m is 40 dBµV/m and every 6 dB is a factor of two. And because emissions scale roughly with 1/distance in the far field, a measurement at 1 m is about 9.5 dB higher than the same source at 3 m — a correction that pre-compliance setups lean on heavily and that is unreliable at the low end, as the next section explains.
In the European Union the picture is shaped differently, and the difference catches people out. A product that contains a radio does not fall under the EMC Directive at all; it falls under the Radio Equipment Directive (2014/53/EU), which folds EMC, spectrum efficiency, and safety into one conformity assessment. The EMC requirement is met by applying the EN 301 489 series (EMC for radio equipment), which for unintentional emissions leans on the same underlying standard a non-radio product would use, CISPR 32 / EN 55032. The radio's own behaviour in the 2.4 GHz band — transmit power, occupied bandwidth, spectral masks, adaptivity — is covered by ETSI EN 300 328, the harmonised standard for wideband data transmission equipment in that band. Immunity, which the FCC does not regulate at all for this class of product, is a genuine requirement in the EU: electrostatic discharge and radiated RF immunity per the IEC 61000-4 series are part of the assessment, and they are the tests a battery-powered plastic gadget most often fails.
The same emission measured three ways gives three different numbers. A peak detector reports the highest instantaneous level and is fast, which is what pre-compliance sweeps use. A quasi-peak detector weights by repetition rate — an interference that happens rarely reads lower, because the standard's concern is perceived interference — and it is slow, sometimes a second per point. An average detector reports the mean. Limits are specified against particular detectors, and a peak measurement is always the most pessimistic. That is genuinely useful: if the peak sweep passes, you are done, because quasi-peak can only be lower. If the peak sweep fails by 2 dB, you may still pass on quasi-peak, and you have just discovered why chamber time costs what it costs.
SEE ITSweep the spectrum — and see which fix costs you a battery
The lab below models Beacon's radiated emissions from 30 MHz to 1 GHz against the FCC Class B limit line at 3 m. Every control is a real design choice you make in Stage 4 or Stage 5. Watch which fixes move which peaks, and watch the battery-life readout while you do it — because the single most effective way to clean up a switching regulator's spectrum is also the one that turns a three-year product into a two-week one.
Start where Beacon started: power-save mode, a bare 1 m cable, a solid plane. The peak near 78 MHz sits above the 40 dBµV/m limit — a failure, and one you would never have seen on battery. Now try the instinct that feels like engineering: switch to forced PWM. The broadband burst noise collapses into a few tidy, filterable lines and the margin improves by about three decibels — and it still fails, because the cable is still the antenna. Meanwhile the sleep current jumps from about ten microamps to a few milliamps, taking the projected battery life from years to a fortnight. Requirement R1 is dead, in exchange for three decibels that did not fix the problem. Then undo that and fit a common-mode choke instead: the same peak drops by about ten decibels, the battery is untouched, and the whole fix costs a component and a footprint. Finally, split the ground plane and watch every clock harmonic rise together — the one change that makes the board itself, rather than the cable, the problem.
IN PRACTICEPre-compliance you can actually afford
A full radiated-emissions test needs a calibrated antenna, a semi-anechoic chamber or an open-area test site, a turntable, a mast, and a receiver with the right detectors. In Jakarta that is a booking and an invoice. What you can do beforehand, on a bench, is pre-compliance: measurements that are not comparable to the limit line but are extremely comparable to each other. Used that way they are worth a great deal.
Near-field probe: find the source, never the level
A small H-field loop probe on a spectrum analyser tells you which part of the board is emitting at a given frequency. Sweep the probe over the buck's hot loop, the module, the USB connector, and the battery leads, and note where each peak rises. It is a locating instrument. Its absolute reading has no defined relationship to a field strength at 3 m, because you are in the near field where E and H are not related by the free-space impedance, and because probe factors vary enormously. Anyone who quotes a near-field probe number in dBµV/m against a limit is confusing themselves.
A/B comparison: the measurement that is actually valid
Set up any repeatable geometry — a broadband antenna at 1 m, the board on a foam block, the analyser on peak detection, max-hold over a sweep — and then change exactly one thing. The difference between two sweeps is trustworthy even when neither absolute number is. "Fitting the choke dropped the 77 MHz peak by 10 dB" is a real result from a garage. "The 78 MHz peak is 38 dBµV/m" is not.
Know your ambient before you blame the board
Sweep with the board switched off. In any city you will see FM broadcast at 88–108 MHz, digital TV, and cellular bands sitting well above the limit line. Those are the environment, not you. If your ambient covers the band where you have a suspect peak, you cannot measure it there without a chamber; what you can still do is the A/B, because the ambient is common to both sweeps.
The 1 m to 3 m correction, and where it stops working
In the far field, field strength falls as 1/distance, so a 1 m measurement is about 20 log10(3) ≈ 9.5 dB higher than the 3 m one. The trap is that the far field does not start until roughly λ/2π from the source. At 30 MHz, λ is 10 m and λ/2π is about 1.6 m, so a 1 m measurement is in the near field and the correction is simply wrong. It works acceptably above a few hundred megahertz and misleads you exactly where the limit is tightest.
Aim for 6 dB of margin, not zero
Chamber-to-chamber variation of several decibels is normal, and so is the difference between your cable dressing and the lab's. A design that pre-compliance says passes by 1 dB is a design that fails on the day. Six decibels is the number practitioners use, and it costs less to buy at layout than at a re-test.
The unit under test must be in its enclosure, with the cables it will ship with, running the firmware it will ship with, exercising its worst-case mode. Beacon's worst case is not the one-report-per-minute cycle everyone thinks of; it is a Wi-Fi OTA session with the USB cable plugged in, which is the only time the radio, the charger, and a metre of cable are all active at once. A lab will ask you what the worst case is and will test what you tell them. Choosing the quiet mode because it passes is not clever; it is a certificate that does not describe your product.
COMPARISONCertification: what the module's grant buys you, and what it does not
The single best decision on Beacon's bill of materials, from a regulatory standpoint, was choosing a pre-certified module rather than laying the ESP32-C3 chip down with your own RF front end and antenna. That decision, made back in Lesson 1.6 for reasons of pin count and price, quietly removed the most expensive item from the certification budget. It is worth being precise about what it removed and what it left behind, because the folklore here is unusually wrong.
Modular approval is an FCC concept. A module that has been granted certification as a single-modular transmitter carries its own FCC ID, and a product that integrates it does not need to re-test the transmitter — provided the integrator satisfies the conditions in the module's grant and its integration guide. Those conditions are specific and they are engineering constraints, not paperwork: the antenna must be the one covered by the grant (for a module with an integrated PCB antenna, that means not modifying it and respecting the keep-out); the module must not be modified; the host layout must follow the manufacturer's guidance about ground and antenna clearance; and the finished product must be labelled to show it contains the module's FCC ID.
What the grant does not cover is the rest of your board. Your switching regulator, your microcontroller's clock, and your USB port are unintentional radiators, and they are your responsibility under Part 15 Subpart B for the finished product. That is exactly the sweep this lesson has been about. In other words, the module removed the hard, expensive, radio-specific testing and left you with the test you were always going to have to do — the one Beacon failed with a cable attached.
| Jurisdiction | What governs the radio | What governs the rest of the board | Does the module's certificate carry over? |
|---|---|---|---|
| United States (FCC) | Part 15 Subpart C (§15.247 for 2.4 GHz digital modulation) — covered by the module's grant | Part 15 Subpart B, Class B, verified on the finished product | Yes for the transmitter, under the grant's conditions; host labelled "Contains FCC ID: …" |
| European Union | Radio Equipment Directive 2014/53/EU; ETSI EN 300 328 for the 2.4 GHz band | EN 301 489 series, leaning on CISPR 32 / EN 55032; plus immunity | Partly — the module's test reports can be reused as evidence, but the CE declaration of conformity is yours for the finished product, including immunity |
| Indonesia (SDPPI) | Type approval (sertifikat perangkat telekomunikasi) for telecommunication and radio equipment, issued by the Directorate General of SDPPI | Assessed as part of the same certificate; testing at a designated laboratory | No — the certificate attaches to the finished product model and must be held by an Indonesian legal entity; the module's own certificate does not transfer |
Everything in that table is the shape of the landscape, not advice for a specific product, and national requirements, fees, and accepted test reports change. In Indonesia the practical route for a small hardware company is to work through a local test house and a certification consultant who files on behalf of an Indonesian entity; the honest budget line for that, in Lesson 6.8, is not small compared with a 25-unit BOM. What you can do as an engineer is make certification cheap when it happens: use a pre-certified module, keep the antenna keep-out sacred, do not modify the module, keep the pre-compliance evidence, and design the filtering in rather than bodging it on. The unit that goes to the lab should be the unit you sell.
One more distinction that matters for a pilot. Building 25 units and running them inside your own organisation for evaluation is a different activity from selling them, and most jurisdictions treat it differently. The moment a unit is sold, imported for sale, or handed to a customer, the certificate question becomes live and is not something to resolve after the fact. Beacon's pilot ran in a workshop belonging to the project. Lesson 6.9's list of what a real product would still need begins with this line.
BEACONBeacon's sweep: thirteen decibels of margin, until the cable
The pre-compliance session used a borrowed spectrum analyser, a broadband antenna at 1 m in a quiet room outside the city, peak detection with max-hold, and the A/B discipline above. Absolute numbers were corrected to a nominal 3 m for readability and are reported here as indicative rather than certifiable. The unit under test was a pilot board in its printed enclosure, running the shipping firmware.
| Configuration | Worst peak | Frequency | Limit | Margin |
|---|---|---|---|---|
| Battery only, normal 1 report/min cycle | 30.0 dBµV/m | 200 MHz | 43.5 | +13.5 dB |
| Battery only, Wi-Fi OTA session running | 33.4 dBµV/m | 200 MHz | 43.5 | +10.1 dB |
| 1 m USB cable, charging at 500 mA | 43.7 dBµV/m | 77 MHz | 40.0 | −3.7 dB — fails |
| Same, with a clamp-on ferrite at the connector | 38.9 dBµV/m | 80 MHz | 40.0 | +1.1 dB — passes, no margin |
| Same, cable dressed differently on the bench | 41.1 dBµV/m | 81 MHz | 40.0 | −1.1 dB — and the only change was the cable's shape |
The last row is the most instructive and the most annoying. Moving the cable — coiling it differently on the bench — changed the result by more than two decibels and flipped a pass into a fail. That is not measurement error; it is the physics. The cable's resonance depends on its geometry and on its capacitive coupling to everything around it, so a common-mode problem is inherently a variable problem. Any fix that leaves you within a couple of decibels of the line has not fixed anything; it has bought a coin flip. That is why the six-decibel rule exists and why the clamp-on ferrite, which passed, was not accepted as the answer.
Two probe sessions located the source. With the near-field loop over the buck's input capacitor, the 77 MHz peak was strong; with the probe over the module, it was not. Disabling the LED rail changed nothing. Putting the board into a state where the buck was in power-save mode with essentially no load — the deep-sleep condition — moved the peak's shape but not its frequency, which is the signature of a cable resonance being driven by a broadband source rather than of a narrowband emitter. The mechanism was therefore: switching noise on the ground plane near the USB connector, coupled onto the cable as common-mode current, radiating from a structure that resonates near 77 MHz. The board was not the antenna. The board was the transmitter and the cable was the antenna.
Two changes were written up with this evidence attached. First, a common-mode choke footprint on D± at the connector, in place of one of the 22 Ω series resistors' positions, so the part can be fitted or bridged. Second, a row of stitching vias along the USB edge tying the top and bottom ground pours together right where the connector's shell lands, shortening the return path that currently lets ground noise develop a voltage across the connector. Together, modelled, they should put the peak near 33.6 dBµV/m — about six and a half decibels of margin, which is a number worth taking to a lab. What did not go on the list: forcing the regulator into PWM mode. It cleans the spectrum and it costs requirement R1, which is the requirement the entire product exists to satisfy. Trading a six-month battery for three decibels is not a trade; it is a different product.
Neither change was made for Rev A, and the pilot ran anyway — inside the project's own workshop, on units that were never sold, with the honest note in the log that the product as built would not pass a Part 15 Subpart B test with a cable attached. Writing that down is the whole point. An engineering team that knows it has an open compliance item behaves differently from one that assumes it is fine.
WEAKNESSWhere this bites
This is the most expensive misunderstanding in small hardware. A modular grant covers the transmitter, under conditions about the antenna and the host layout that you must actually meet. It says nothing about your regulator, your clock, or your cables, all of which are unintentional radiators that the finished product is responsible for. It also does not travel between jurisdictions: an FCC modular approval is not a CE declaration and is not an Indonesian type approval, and only one of those three even has a modular concept. The module removes the hardest test from your budget. It does not remove the budget.
When a board fails, the reflex is to buy components: beads on everything, capacitors everywhere, a shield can over the noisy corner. Sometimes that works; more often it moves the peak somewhere else, costs unit price forever, and hides the real defect. Emissions are a function of loop area and return-path continuity, and both are decided in the layout for free. A split plane that adds nine decibels across every clock harmonic cannot be filtered away, because the noise is being radiated by the board's own geometry. The parts that are worth fitting are the ones at the boundary — at the connector where a cable leaves — because that is where filtering intercepts a mechanism rather than a symptom.
Both instruments are enormously useful and neither produces a number that can be compared with a limit. The near-field probe is in the near field by definition, where the ratio of electric to magnetic field is not the free-space impedance and where a probe factor is a fiction. A garage sweep sits in an ambient full of broadcast signals, on a ground plane that is a workbench, with reflections from the walls. Use them for differences and for locating sources, and treat any absolute number as a hypothesis to be tested at a real site. The failure mode here is not being wrong; it is being confidently wrong and cancelling the chamber booking.
KEY INSIGHTHow this connects
- Decoupling, Bypassing, and the Return Path (2.5) — the loop that a decoupling capacitor closes is the loop that would otherwise radiate; everything in this lesson's differential-mode half is that idea at higher frequency.
- Grounding, Planes, and Return Currents (5.6) — the split-plane control in the widget is that lesson's central warning, priced in decibels.
- Test Jigs, Pogo Pins, and Production Test Firmware (6.3) — a jig measures unit-to-unit variation; an EMC limit measures the design. A jig can never tell you whether a design passes, because every unit shares the design.
- Enclosures, Thermal, and Mechanical Design (6.5) — next: the enclosure changes the antenna too, and the same keep-out that protects the module's grant protects your link budget.
- Cost, BOM, and Scaling From 25 to 25,000 (6.8) — certification is the largest non-recurring cost in the whole project, and at 25 units it dwarfs the bill of materials by a factor you will not believe until you see it.
EMC is where a board stops being a private object. Everything before this was between you and the design. From here on, there is a limit line drawn by someone else, and a measurement that decides whether you may sell.
GLOSSARYKey terms introduced
- Differential-mode emission
- Radiation from a current flowing out and back through a defined loop on the board. Field strength scales with loop area and with the square of frequency, so it is controlled by layout rather than by components.
- Common-mode emission
- Radiation from current flowing in the same direction along every conductor of a cable and returning through stray capacitance to the environment. A metre-long cable resonates near 75 MHz, and a few microamps there can reach the Class B limit.
- FCC Part 15 Subpart B
- The United States rules for unintentional radiators — digital circuitry, regulators, everything that is not the licensed or licence-exempt transmitter. Class B radiated limits at 3 m are 40, 43.5, 46, and 54 dBµV/m in ascending frequency bands.
- CISPR 32 / EN 55032
- The international and European emissions standard for multimedia equipment, referenced by the EU's radio-equipment EMC standards. Sets Class A (industrial) and Class B (residential) limits analogous to the FCC's.
- ETSI EN 300 328
- The harmonised European standard for wideband data transmission equipment in the 2.4 GHz band, covering transmit power, occupied bandwidth, and spectrum-sharing behaviour. It governs Beacon's BLE radio in the EU; the module's certification addresses it.
- Modular approval
- An FCC grant that certifies a radio module in its own right, so an integrator need not re-test the transmitter — provided the module is unmodified, the antenna is as granted, the host layout follows the manufacturer's conditions, and the product is labelled with the module's FCC ID. It does not cover the host's unintentional emissions.
- SDPPI type approval
- Indonesia's certification for telecommunication and radio equipment, issued by the Directorate General of SDPPI. It attaches to the finished product model, is filed by an Indonesian legal entity, and requires testing at a designated laboratory; a certified module inside the product does not transfer its certificate.
- Common-mode choke
- Two conductors wound on one core so that differential current sees almost no inductance and common-mode current sees a large one. The correct filter for a data pair, where a plain ferrite bead would attenuate the signal along with the noise.
ACTIVE RECALLCheck your understanding
Answer in your own words first, then reveal.
1Beacon passed with 14.5 dB of margin on battery and failed by 3.7 dB with a USB cable attached. Nothing on the board changed. Explain the mechanism, and say why moving the cable changed the number again.›
On battery, the only radiating structures are the board's own current loops, and a 35 × 25 mm board is a hopeless antenna at 78 MHz — the dimensions are about a hundredth of a wavelength. Plugging in a metre of cable attaches a structure that is resonant near 75 MHz. The drive is the high-frequency noise voltage that switching currents develop across the ground plane near the connector; that voltage appears between the cable and the environment and pushes common-mode current along it. Because the return path is stray capacitance to the whole room, the effective loop is enormous and the radiation efficiency is high, so a few microamps is enough. The cable's shape matters because resonance depends on its length, its coupling to the bench, and its coupling to you. That is the deeper point: a common-mode failure is inherently variable, so passing by one decibel is not passing.
2Forcing the TPS62840 into PWM mode measurably improves Beacon's spectrum. Why was it rejected, and what does that tell you about EMC fixes in general?›
Because power-save mode is what makes the regulator draw 60 nA of quiescent current, and forced PWM makes it switch continuously regardless of load. At no load the inductor still carries the full ripple current, so the input current rises from tens of nanoamps to a few milliamps. Beacon's whole sleep budget is about 8 µA; a few milliamps is three orders of magnitude larger, and the projected life falls from years to roughly a fortnight. Requirement R1 — six months on a 1000 mAh cell — is the product's reason for existing, so a fix that violates it is not a fix. Generally: EMC fixes are traded against other requirements, and the currency is usually power, cost, or size. The good fixes are the ones with a small denominator — a common-mode choke costs a component and a footprint and no battery at all, which is why it won.
3Your colleague reports "the near-field probe reads 52 dBµV over the regulator, which is over the 46 dBµV/m limit — we fail." What is wrong with that sentence?›
Almost everything. The units differ: the probe reads a voltage into the analyser, in dBµV, while the limit is a field strength at a distance, in dBµV/m; there is no general conversion. The probe sits in the near field, millimetres from the source, where the electric and magnetic fields are not related by the free-space impedance, so no distance correction recovers a far-field number. And the limit is defined for a specific measurement geometry at 3 m with a specific detector, which a handheld loop on a bench does not reproduce. What the reading is good for is location and comparison: it says the regulator is a strong source at that frequency, and if a design change drops it to 41 dBµV, that 11 dB improvement is real and will show up at a proper site. Use probes for differences, never for verdicts.
4Beacon uses a pre-certified ESP32-C3-MINI-1. List what you still have to do for the United States, for the European Union, and for Indonesia.›
United States: the module's grant covers the 2.4 GHz transmitter under Part 15 Subpart C, provided the module is unmodified, the antenna and keep-out are as granted, and the host follows the integration conditions. You still owe Part 15 Subpart B compliance for the finished product's unintentional emissions — the sweep this lesson describes — plus the "Contains FCC ID" labelling and the associated documentation. European Union: a product containing a radio falls under the Radio Equipment Directive, and the CE declaration of conformity is yours for the finished product, not the module's. You can reuse the module's test evidence for the radio requirements under EN 300 328, but you must address host EMC under the EN 301 489 series and — unlike in the US — immunity, where electrostatic discharge and radiated RF immunity are real tests a plastic battery-powered device often fails. Indonesia: SDPPI type approval attaches to your finished product model, is applied for by an Indonesian legal entity, and requires testing at a designated laboratory; the module's certificate does not carry over, and the certificate number must appear on the device. The common thread: the module deleted the hardest and most expensive test, and left you with all the ones that describe your product rather than its radio.