How to Choose an EMC Test Lab and Prepare to Avoid Retests
How to choose an accredited EMC test lab and prepare for it: scope checks, test plan, performance criteria, samples, operating modes and ways to avoid retests.
An EMC test session is billed by the hour, booked weeks ahead and often sits on the critical path to a product launch. Most of the cost of a failed session is not the lab fee but the delay: finding the fix, rebuilding samples and waiting for the next free slot. Many of those failures have nothing to do with the design. The EUT crashed for an unrelated reason, the wrong cables were shipped, nobody had defined what counts as a pass in the immunity tests, or the report was missing a test the certification body needed. Here is how to choose a lab and prepare so that the first session produces a usable result.
Choosing a lab: what to check first
Start with accreditation. A lab accredited to ISO/IEC 17025 has been assessed for competence, but only for the methods listed in its scope of accreditation. Ask for the scope document and check that it lists every standard you need, in the edition you plan to declare against, for example CISPR 32 for emission and CISPR 35 or the IEC 61000-4 series for immunity. Accreditation bodies that have signed the ILAC mutual recognition arrangement recognize each other's accreditations.
Some markets also require the lab itself to be recognized. FCC certification in the United States requires testing by an FCC-recognized accredited laboratory, VCCI in Japan accepts data from measurement facilities registered with VCCI, and KC certification in Korea relies on designated test laboratories. In the European Union, the EMC Directive (2014/30/EU) lets the manufacturer declare conformity on its own, so the choice of lab is commercial, although an accredited lab makes the technical file easier to defend.
| Criterion | What to ask | Why it matters |
|---|---|---|
| Scope of accreditation | Are all required standards and editions listed? | A test outside the scope may not be accepted |
| Radiated emission site | 3 m or 10 m chamber? Free-space conditions above 1 GHz? | Some standards or customers require a 10 m measurement; floor absorbers are needed above 1 GHz |
| EUT size and power | Maximum size and weight on the turntable; voltage, frequency, phases and current available | Large floor-standing or three-phase products fit only some chambers |
| Immunity capability | Field strength and frequency range for radiated immunity; surge and EFT generator ratings | Industrial and automotive levels exceed what some labs can generate |
| Engineering support | Can you debug and modify on site? Is a bench with a soldering station available? | Fixing a failure the same day avoids a second booking |
| Rebooking | How soon is the next slot after a failure? Are partial retests possible? | Lead time after a failure often decides the launch date |
| Report | Format and content accepted by your certification body; turnaround time | A report that must be rewritten delays certification |
Write a test plan before you book
A test plan is a short document, often 2 to 5 pages, that the lab reviews before the session. Writing it yourself forces decisions that would otherwise be made in the chamber at hourly rates. It should cover:
- EUT identification: model, hardware revision, firmware version, and which variants the report must cover.
- Standards, class and test levels, for example CISPR 32 Class B and the immunity levels of CISPR 35.
- Port list: every power, signal and telecom port with cable type, maximum length and whether the cable is shielded. Many immunity standards apply EFT, surge or conducted RF tests to signal ports only if the cable can exceed a stated length, often 3 m, so the maximum length decides which tests apply.
- Power supply: rated voltage, frequency and phases, or DC input range.
- Highest internal frequency. In CISPR 32 and FCC Part 15, it decides whether and how far above 1 GHz the radiated emission scan must go.
- Configuration: support equipment, cable layout and how the EUT is exercised.
- Performance criteria and monitoring method for each immunity test (see below).
Define performance criteria the lab can judge
Immunity results are judged against performance criteria. The generic standards and CISPR 35 use three levels: criterion A, normal operation within specification during the test; criterion B, temporary degradation during the test with automatic recovery afterwards; and criterion C, temporary loss of function that recovers by itself or can be restored by operator action, such as a reset. Only the manufacturer can say what normal operation means for a specific product. If the plan says only criterion A, the lab engineer has to decide on the spot whether a flicker on the display or one dropped packet is a failure.
Translate each criterion into something measurable and observable from outside the chamber. Examples: an Ethernet link that stays up with no more than an agreed packet error count; an analog reading that stays within a stated tolerance of its pre-test value; an audio output whose demodulated level stays below a stated value; a display with no visible distortion as seen on a camera. Decide also how the lab will see it: a video camera, a fiber-optic link, or a monitoring cable through a filtered bulkhead. A monitoring cable that is not filtered can carry the test signal into the EUT and cause a failure that does not happen in the field.
Prepare the samples and the operating modes
For emission, the EUT should run the mode that produces the most emission: highest clock rates, all interfaces active, maximum data throughput, and displays or motors running. For immunity, it should run the mode most likely to be disturbed, which is not always the same one. If unsure, scan every mode with the peak detector during pre-compliance and run the full test only in the worst one. Each extra mode in the formal test adds a full set of scans.
- Bring at least two working samples. Surge and ESD can damage a sample, and a second unit lets the session continue.
- Use production-representative hardware: final enclosure, gaskets, screws, cables and power supply. A prototype without its cover is not the product you will sell.
- Let the firmware boot straight into the test mode and run unattended, with a visible status indicator. Restarting a test mode by hand inside the chamber costs time on every scan.
- Pack a fix kit: clip-on ferrites, copper tape with conductive adhesive, spare capacitors and resistors, gasket material and spare fuses.
- Bring schematics, layout files and the bill of materials, so the person debugging can find the source of a frequency quickly.
Common causes of retests and how to avoid them
| Cause | What happens | Prevention |
|---|---|---|
| EUT crashes for a non-EMC reason | Immunity results are invalid; time is lost restarting | Run the test mode for several hours before shipping |
| Configuration differs from the plan | Results do not match pre-compliance; the report describes the wrong setup | Photograph the pre-compliance setup and send a cable diagram |
| No margin at the limit | A small difference between sites turns a pass into a fail | Pre-compliance with several dB of margin at every frequency |
| Vague performance criteria | Disputed results or a conservative fail | Measurable criteria agreed with the lab before the session |
| Fixes not recorded | Production units do not match the tested configuration | Log every modification; the report must describe it |
| Wrong standard edition or missing test | The certification body rejects the report | Confirm the required standards and tests before booking |
On the test day
Arrive early enough to set up and check the EUT before the clock starts. Stay for the tests, or at least for the first scans, when questions about cable routing and operating modes come up. Ask for the peak scans with the margin at each frequency, not only a pass or fail, so that you know how close the product is. If you fix a failure on site, make sure the change is written into the report as a modification, and repeat the tests it could affect: a ferrite added to cure radiated emission can change conducted emission too. Before you leave, review the list of completed tests against the plan.
Common mistakes
- Booking the formal test as the first EMC measurement on the product.
- Shipping a new firmware build the day before the test. An untested build can crash, and its emission can differ from what you measured.
- Assuming one report covers every market. Check each target market's lab and report requirements first.
Calculators used in this post
More posts
Results and summaries are for reference. For certification and test reports use the latest official standard text and calibrated instrument data.