The short answer: a first-round failure is normal, so do not start by changing the structure
Failing the first EMC screening round is a common state for active medical devices, not an accident. What actually separates a cheap project from an expensive one is the three steps that follow: locating the cause, choosing the fix, and defining the retest scope. Do those three in the right order and most problems converge at the level of filtering, grounding and cable routing. Do them in the wrong order and you tend to get pushed all the way down to re-laying the PCB and re-cutting tooling.
One convention has to be settled before any of that. IEC 60601-1-2 and YY 9706.102 sit within the same framework of requirements, and the applicable documents have to be fixed against the target market at protocol stage. If the device itself contains a radio-frequency generating part — radiofrequency therapy, radio-frequency identification, or induction heating, for example — the disturbance requirements of GB 4824 for industrial, scientific and medical radio-frequency equipment also have to be considered. Start screening before the applicable documents are settled and the resulting data may be unusable as a set.
Step one: sort the symptom into emission or immunity
The diagnostic logic for these two families is entirely different, and investigating them together burns a great deal of time. What you do with a set of failing data is assign every exceedance or failure record to one row of the table below — not open a discussion about "whether to add a shield can".
| Symptom | Family | Most likely coupling path | Where to look first |
|---|---|---|---|
| Conducted disturbance exceeds the limit in the low-frequency range | Emission | Switching supply common-mode current returning through the mains cable | Mains inlet filtering, return path of the earthing capacitors |
| Conducted disturbance exceeds the limit in the high-frequency range | Emission | High-speed digital signals coupling back as common-mode current on cables | Cable shield termination method, clock routing |
| Narrowband peaks in radiated disturbance | Emission | Clock and its harmonics | Crystal placement, clock drive strength, spread-spectrum settings |
| Broadband rise across radiated disturbance | Emission | Cables acting as an effective antenna | Cable length and routing, shielding, enclosure seams |
| Lock-up or reset after electrostatic discharge | Immunity | Discharge current entering the ground plane through the enclosure | Enclosure bonding, interface protection devices, reset circuit |
| Display anomaly under burst testing | Immunity | Common-mode injection on power and signal lines | Common-mode chokes, placement of isolation devices |
| Reading drift under a radio-frequency field | Immunity | Pickup by the analogue front end and sensor cabling | Front-end filtering, sensor cable shielding and grounding |
| No automatic recovery after a voltage dip | Immunity | Supply hold-up capability and software reset logic | Brown-out detection, power-up self-recovery sequence |
The value of the classification is that it narrows the conversation. The same exceedance record, filed as "emission, returning through the mains cable" versus "immunity, entering the ground plane through the enclosure", implies completely different corrective actions; discussing both at once only means the two interfere with each other.
Step two: three fixed moves for locating the cause
Move one is reproduction. Get the symptom to reproduce stably under the same setup first. EMC data are extremely sensitive to setup — change the cable routing, equipment placement or grounding arrangement a little and the conclusion changes a little. A problem you cannot reproduce reliably is a problem whose fix you cannot verify.
Move two is decomposition. Disconnect or substitute functional blocks one at a time: remove external cables one by one and see whether the exceedance disappears, power a suspect module from a separate external supply, separate the display unit from the main controller. The aim is to pin the problem down to "which port, over which path", not to stop at "this unit has a problem".
Move three is hypothesis testing. Verify the direction with temporary means: clip on a ferrite, tape down a piece of copper foil, tack a capacitor in parallel. If a temporary measure works, the hypothesis is pointing the right way. The value of this step is that it screens out wrong directions cheaply, instead of committing straight to a product-level change.
The order of the three cannot be shuffled. Skip reproduction and go straight to modifying, and when the retest still fails you cannot tell whether the measure was ineffective or the setup shifted. Skip decomposition and go straight to overall shielding, and the cost goes up while the problem may still be there.
Step three: work the fixes in priority order, cheap before expensive
| Measure level | Symptoms it suits | Cost | Risks to watch |
|---|---|---|---|
| Cable routing and grounding adjustments | Broadband radiated rise, common-mode injection immunity failures | Low, affects assembly work instructions only | Must be frozen into assembly documentation or production will drift |
| Adding ferrites and filter components | Conducted exceedances, burst test failures | Low to medium | Effect on signal integrity and safety isolation needs assessment |
| Enclosure bonding and seam treatment | Electrostatic discharge, high-frequency radiation | Medium | Touches cosmetic parts and assembly tolerances; drawings must be updated in step |
| Software fault tolerance and self-recovery logic | No recovery after dips or transients | Medium | A software change, which may trigger updates to software-related documentation |
| PCB layout and ground plane rework | Clock harmonic peaks, sensitive analogue front end | High | Safety and function have to be re-verified; the schedule stretches noticeably |
| Mechanical and tooling modification | Shield cavity unworkable, seams cannot be closed | High | New tooling required, plus repeat of the related safety verification |
The point of the priority order is that each level down raises the cost of backing out by a full step. Reserving a filter footprint, a bonding point and space to mount a ferrite at design stage is close to free; adding them at prototype stage means changing assembly; adding them after the design is frozen means changing tooling plus re-verification.
One class of failure should skip straight to the lower levels. If several unrelated symptoms appear together — radiated peaks, plus electrostatic sensitivity, plus drift under a radio-frequency field — that usually points to a problem in the grounding system itself, and patching symptom by symptom just moves the bulge around. Going back to the layout early is the cheaper path in that case.
The special problem with immunity failures: defining essential performance
Immunity verdicts depend on how you yourself define essential performance and acceptable degradation. That definition has to be written down before testing, and it has to hold together.
Defining it too broadly is a common problem. Treating "does not smoke or catch fire" as the entire floor, and passing a run in which readings jump around, does not stand up on the review side, because output accuracy in a diagnostic or therapeutic device is itself part of essential performance.
Defining it too narrowly is the opposite extreme. Declaring every display flicker a failure means acceptable momentary behaviour gets a fail verdict, which adds corrective work and schedule for nothing.
More awkward still is a definition that contradicts the instructions for use. The instructions promise a given function in a given use environment, while the immunity verdict excludes that function from essential performance. Once that contradiction is noticed, the technical documentation has to be revised and part of the testing repeated.
The right approach is to write the essential performance list, the assessment method for each item, and the permitted range of degradation into a document at protocol stage, considered together with the safety-related verdict conventions.
Retest scope: can you get away with the affected items only?
A full repeat is not always required after corrective work, but the reasoning has to hold.
| Change made | Retest scope |
|---|---|
| Cable routing and earth points adjusted only, hardware unchanged | Retest the affected items, retain the original data for the rest with a written rationale |
| Filter components or ferrites added | Retest all emission items plus the affected immunity items |
| Enclosure bonding method changed | Retest radiated items and electrostatic discharge items |
| Software logic change affecting reset or self-recovery | Retest immunity items; emission items may be retained depending on the nature of the change |
| PCB revision | Treat as a full retest |
| Power supply module replaced | Treat as a full retest, and reassess the safety-related items in step |
The heart of a retest plan is the technical explanation of why the items not retested are unaffected — not merely a list of what was done. If you cannot write that explanation, retest everything honestly. That is far cheaper than being asked for supplementary testing afterwards, and far quicker than having the report sent back.
Three frequent misconceptions
The first is treating screening data as a pre-test conclusion. Screening is usually run in a simplified setup, with different arrangement, cabling and auxiliary equipment from the formal test. Passing screening does not mean passing the formal test, and equally, failing screening does not mean the formal test will fail. State the two conclusions separately.
The second is verifying only the item that failed. A filter component is added, the conducted exceedance is resolved, and nobody goes back to check whether radiated emission has been pushed up. EMC measures often trade one effect against another, so the verification scope has to cover items that may interact — which is exactly why "all emission items" appears in the retest table above.
The third is treating EMC as the last operation in the sequence. Scheduling EMC only after the mechanical design, cabling and PCB are all frozen shuts off every low-cost option in advance and leaves only the expensive ones. The sensible approach is a screening round at prototype stage, exposing directional problems while they can still be fixed.
Pre-delivery self-check
- Are the applicable documents matched to the target market, and have devices containing a radio-frequency generating part been considered against the requirements for industrial, scientific and medical radio-frequency equipment
- Are the essential performance list and permitted degradation documented, and consistent with the wording in the instructions for use
- Is the screening setup configured with cabling and auxiliary equipment as the formal test requires
- Has every exceedance or failure record been assigned to a specific port and coupling path
- Have the corrective measures been frozen into drawings and assembly work instructions, rather than stopping at the prototype
- Can the exemption rationale for items not retested be supported with a technical basis
For project arrangements and screening options see EMC testing, for how this interacts with safety work see electrical safety testing, and for test packages covering other active devices see testing for other devices. For report issue and retest milestones, see our service process.
Turning a failure list into a concrete corrective path
Once a failing report is in hand, the time sink is usually not the fixing but the decision about which level to fix at. Send over the list of failed items, the test setup description and the power and cabling configuration, and we can classify the symptoms, propose an investigation order and give a first view on retest scope before discussing the corrective and retest schedule. SUNGO Lab (Shanghai Shage Medical Technology Co., Ltd.) is a medical device testing laboratory accredited by CNAS, CMA and IAS (USA), with laboratories in Shanghai and Hefei. Please note that accreditation marks only demonstrate that the laboratory holds the relevant technical competence within its accredited scope; they do not constitute a commitment regarding market access in the target market, and the final conclusion still rests on the test data and the review requirements of that market. To discuss a project, call +86 132 4819 8029 or submit your documentation through request a quote.