Every electronic product sold in the United States answers to 47 CFR Part 15. The limits are numeric, and the pass or fail decision is made in an accredited chamber by a technician with no interest in your schedule. What varies between companies is when they find out.

The Cost Curve Is Not Linear

An emissions problem is cheap in the schematic. It costs a resistor value, a ferrite footprint, a decision to route a clock on an inner layer. The same problem found during a certification test costs a chamber rental, a lab day, a board respin, and a slot in the queue three weeks out. Found after injection mold tooling has been cut, it costs the tool.

That last case is worth planning against. Radiated emissions above 30 MHz are a function of the enclosure as much as the board. Seam length, gasket compression, apertures for connectors and displays, and the impedance of the joint between a shield can and the chassis all move the number. Once the mechanical design is frozen in steel, the remaining degrees of freedom are firmware and passives.

What Part 15 Actually Says

Subpart B covers unintentional radiators: anything that generates radio frequency energy for internal use and does not intend to emit it. That is a DAC, a streamer, a motor controller, a battery charger. Section 15.109 sets the radiated limits. For a Class B digital device measured at 3 meters, the limits are 100 microvolts per meter from 30 to 88 MHz, 150 from 88 to 216 MHz, 200 from 216 to 960 MHz, and 500 above 960 MHz. In logarithmic terms that is 40.0, 43.5, 46.0 and 54.0 dBuV/m. The tighter limit applies at a band edge.

Section 15.107 sets the conducted limits on the AC mains from 150 kHz to 30 MHz, measured through a 50 microhenry, 50 ohm line impedance stabilization network. For Class B the quasi-peak limit falls from 66 to 56 dBuV across 0.15 to 0.5 MHz with the logarithm of frequency, sits at 56 dBuV from 0.5 to 5 MHz, and rises to 60 dBuV from 5 to 30 MHz. The average limits sit 10 dB lower in each band.

Subpart C covers intentional radiators. Section 15.209 gives the general field strength limits, including 30 microvolts per meter from 1.705 to 30 MHz at 30 meters and the same 3 meter numbers as Subpart B above 30 MHz.

Section 15.31 names ANSI C63.4-2014 as the measurement procedure, with several clauses excluded. That standard specifies the site, the antenna height scan from 1 to 4 meters, the 360 degree turntable rotation, and the cable arrangement. Section 15.35 requires an average detector with a minimum 1 MHz resolution bandwidth above 1000 MHz, and caps peak emissions at 20 dB above the average limit. Section 15.31 also gives the extrapolation factors: 20 dB per decade of distance above 30 MHz, and 40 dB per decade below.

Class A and Class B Are Not the Same Product Decision

Section 15.3 defines a Class B digital device as one marketed for a residential environment, and a Class A device as one marketed for commercial, industrial or business use and excluded from home use. The Class A limits are measured at 10 meters. Extrapolate the 30 to 88 MHz Class A limit of 90 microvolts per meter back to 3 meters at 20 dB per decade and it lands near 49.6 dBuV/m, against 40.0 dBuV/m for Class B. That is roughly 9.6 dB of additional headroom.

Nine decibels is a lot of engineering. It is also a marketing decision that cannot be quietly reversed later, because Section 15.105 requires a Class A device to carry a statement telling the user that operation in a residential area is likely to cause harmful interference and that correcting it is the user's expense. Pick the class during requirements definition. Design to Class B if there is any chance the product reaches a consumer.

What a Pre-Scan Tells You

A useful bench setup is modest. A spectrum analyzer, near field probes, a LISN, and a current probe will find most problems. Run the conducted scan first. It is the most repeatable measurement available outside a chamber, because the LISN defines the source impedance and removes the ambiguity of the room. If the conducted profile sits 10 dB under the Section 15.107 limit with the real supply and the real load, one failure mode is closed.

For radiated, a pre-scan finds signatures rather than absolute values. Harmonic combs at exact multiples of a crystal or a PLL output point at a clock. Broadband humps that move when you change load current point at a switching regulator. A near field H-field probe walked along the board will localize the aggressor to a specific loop in a few minutes. A current probe clamped around a cable bundle will tell you whether the cable is the antenna, and a ferrite slipped over the same bundle will confirm it.

What a Pre-Scan Does Not Tell You

It does not give you a number that will survive a chamber. Ambient RF at any office site swamps the 40 dBuV/m region across the FM and broadcast bands. Ground plane reflections, antenna height scan, turntable azimuth and cable dress can each move a reading by several decibels, and C63.4 exists to constrain those variables. Treat a pre-scan margin of 6 dB as informational and 15 dB as comfortable. Neither is a grant.

A pre-scan also cannot substitute for the final configuration. Emissions are a property of the assembled unit with production firmware, cables and enclosure. A board on a bench with a lab supply and a short USB lead is a different device.

Where the Energy Usually Comes From

Four sources account for most failures.

Clock harmonics are the first. A 25 MHz oscillator with fast edges produces measurable content past 1 GHz. Series termination, a slower slew rate setting, and spread spectrum modulation where the protocol tolerates it all reduce the peak.

Switching regulators are the second. The loop formed by the input capacitor, the switch node and the ground return is a small transmitting loop. Shortening it on the layout is free at schematic review and impossible after fabrication.

Common mode current on cables is the third, and it is the most common cause of a failure between 30 and 300 MHz. A one meter cable is an efficient antenna in that range. The current is driven by a small ground voltage difference between the board and the connector shell.

Enclosure seams are the fourth. A slot behaves as a slot antenna when its longest dimension approaches a half wavelength. Above 1 GHz that dimension is under 150 mm.

Modules Versus Certifying Your Own Radio

If the product needs a radio, a certified module is usually the correct engineering answer. Section 15.212 lists eight conditions for single modular approval, including that the radio elements have their own shielding, buffered modulation and data inputs, its own power supply regulation, a compliant antenna arrangement, standalone testing, FCC identification labeling, integration instructions, and RF exposure documentation. A module that meets all eight carries its own grant. The host still has to demonstrate Subpart B compliance for the digital circuitry, and the integration conditions in the module manual are binding.

Certifying a discrete radio design means a full Certification through a Telecommunication Certification Body, with testing at an FCC-recognized accredited laboratory. It is the right choice when volume, cost or form factor demands it. It is also a program line item, and treating it as a late task is where schedules break.

Section 15.101 permits either Supplier's Declaration of Conformity or Certification for most unintentional radiators. Either way, the technical file has to exist and the measurements have to be real.

References