EMC Pre-Compliance Testing

Every electronic product that switches fast — a switching power supply, a microcontroller board, a motor drive, an LED driver — radiates and conducts some unintended radio-frequency energy. Electromagnetic compatibility (EMC) rules limit how much, so that products do not interfere with radio, television, mobile phones and each other. Formal EMC testing at an accredited laboratory is expensive, and failing it is more expensive still, because every redesign means another test. Pre-compliance testing on your own bench, with a spectrum analyzer and a few accessories, finds the problems early, while they are cheap to fix.

Conducted-emissions pre-compliance: the product is powered through a LISN and the spectrum analyzer scan is compared with the limit lines. Peaks above the limit show which harmonics need filtering.

View Larger

What EMC Covers

Electromagnetic compatibility has two halves. Emissions are the unwanted energy a product gives off, either conducted along its power and signal cables or radiated through the air. Immunity (or susceptibility) is a product's ability to keep working when exposed to interference: electrostatic discharge, radio fields, fast transients and surges on the mains. A product sold in most markets must meet both. This lesson concentrates on emissions, which are the most common reason for failing a test and the easiest to measure on the bench.

Most emissions come from the same few sources: the fast edges of switching converters and digital clocks, and the harmonics of those edges extending to hundreds of megahertz. They escape mainly through the cables attached to the product, which act as accidental antennas — which is why a product often passes with its cables removed and fails with them connected.

Standards and Limits

The limits depend on the type of product and where it will be used. The main families are:

Class B (residential) limits are about 10 dB stricter than Class A (commercial), because a radio or TV in a home is likely to be close to the product. As a guide, the CISPR 32 Class B conducted limits on the mains cable are 66–56 dBµV quasi-peak from 150 kHz to 500 kHz (falling with frequency), 56 dBµV from 0.5 to 5 MHz, and 60 dBµV from 5 to 30 MHz, with average-detector limits 10 dB lower. Radiated limits apply from 30 MHz upward, measured with an antenna at 3 or 10 m.

dBµV and Detectors

EMC limits are expressed in dBµV (decibels relative to one microvolt) for conducted emissions and dBµV/m for radiated field strength. A spectrum analyzer reads in dBm into 50 Ω. The conversion is:

The standards also specify the detector. A peak detector reads the highest value; a quasi-peak detector weights short bursts less, roughly matching how annoying a sound is to a listener; an average detector reads the mean. Peak is always the highest of the three. A useful pre-compliance strategy is therefore to measure with the peak detector first: anything that passes the quasi-peak and average limits with peak readings will certainly pass; only frequencies that fail on peak need to be checked more carefully.

Conducted Emissions and the LISN

Conducted emissions are measured on the mains cable from 150 kHz to 30 MHz using a line impedance stabilization network (LISN). The LISN sits between the mains supply and the product. It passes mains power through, presents the product with a standard 50 Ω impedance at radio frequencies (so that results do not depend on the building's wiring), blocks noise from the mains reaching the measurement, and provides a BNC output for the spectrum analyzer. A transient limiter between the LISN and the analyzer protects the analyzer's input.

For DC-powered products and modules, the same measurement is made on the DC supply leads with a DC LISN, and for vehicle products according to CISPR 25.

Radiated Emissions and Near-Field Probes

True radiated-emissions measurements need a calibrated antenna, a known distance (usually 3 or 10 m) and a quiet site such as an anechoic chamber, so they are hard to reproduce on a bench. Pre-compliance therefore uses two cheaper techniques:

• Near-field probes: small magnetic loops and electric-field tips, connected to the spectrum analyzer, that are moved over the circuit board and cables. They show which component, trace or cable is the source of each emission. Readings are relative, not calibrated to the limits, but they show clearly whether a change makes things better or worse.
• Current clamps on cables: since cables are the main radiators, a high-frequency current clamp around each cable measures the common-mode current that drives the radiation. A few microamps of common-mode current at 100 MHz on a 1 m cable is enough to exceed Class B radiated limits.

Many makers also do an informal radiated scan with a small antenna a meter or so from the product in an ordinary room, comparing results before and after each change. Because the room has reflections and other signals, the absolute numbers are only rough, but the comparisons are useful.

A Pre-Compliance Procedure

1. Measure the background first. With the product switched off, record the analyzer's display. Broadcast stations, mobile phones and other equipment will appear; anything that is still there with the product off is not your problem.
2. Conducted scan: 150 kHz to 30 MHz through the LISN, peak detector, 9 kHz RBW. Convert to dBµV and compare with the limit line. Note every peak within 10 dB of the limit.
3. Radiated survey: 30 MHz to 1 GHz (or higher) with a current clamp on each cable and a near-field probe over the board. Identify the frequencies and sources of the largest emissions; harmonics of a known clock (for example every 24 MHz) point directly at their source.
4. Fix the worst problem first, re-measure, and repeat. Keep a record of each change and its effect.
5. Test in the worst-case configuration: all cables connected, the product doing whatever makes the most noise (full load, all interfaces active).

Common Fixes

• Layout: a solid ground plane under all fast signals, short switching loops in power converters, and decoupling capacitors right at every IC are the cheapest and most effective fixes, but they must be done before the board is made.
• Filtering at the cable entry: common-mode chokes, X and Y capacitors on mains inputs; ferrite beads and capacitors on DC and signal lines, placed right where the cable leaves the board.
• Slowing the edges: a small series resistor on clock and data outputs, or a lower drive-strength setting, removes harmonics above the frequencies the signal needs.
• Spread-spectrum clocking: many clock generators and switching controllers can dither their frequency slightly, spreading each harmonic over a wider band and lowering its peak (and quasi-peak) reading by several decibels.
• Shielding and cable choice: metal enclosures with good seams, shielded cables with the shield bonded to the enclosure all the way round, and ferrite clamps on cables as a last resort.

dBm, dBµV and Margin Calculator