Start with the cost: three failure types, three very different repair windows
Failing electrical safety is rarely a "tune it a little and it will pass" situation. Creepage and clearance, protective earthing and leakage current sit in completely different repair windows, and the cost gap between them is an order of magnitude.
| Failure type | What it looks like | Where it can still be fixed | Cost of finding it late |
|---|---|---|---|
| Creepage distance and clearance too small | PCB trace spacing, transformer winding separation or terminal spacing does not meet the requirement for the insulation grade at that point | Schematic and mechanical design stage | At prototype stage this effectively means a new board spin and new tooling, and the whole project gets rescheduled |
| Protective earth path broken or unreliable | The path between an accessible metal part and the earth terminal is interrupted, or the connection impedance is high | Mechanical definition and process documentation stage | Often rescuable with local conductive treatment, different screws and anti-loosening washers, but process documents and first-article approval have to be redone |
| Leakage current out of tolerance | Scrapes through in normal condition, fails in single fault condition | Power architecture and filter design stage | Touches filter capacitors, transformer and power module selection; one change pulls everything, and it drags EMC down with it |
The one-line version: creepage is a design problem, protective earthing is a process problem, leakage current is an architecture problem. Testing only exposes these three; it does not solve any of them. The genuinely economical move is to draw the insulation diagram during design review, against the general requirements of GB 9706.1 and IEC 60601-1, instead of waiting for the prototype to land on the bench. Below is the decision basis for each of the three.
Creepage and clearance: draw the insulation diagram first, argue about spacing second
Plenty of teams reach for the callipers first. That order is backwards. Four questions have to be answered before any measurement means anything, and getting one of them wrong makes the numbers worthless:
- How is the working voltage across this barrier determined: the mains-side voltage, or the voltage actually appearing across that barrier, and does anything have to be added on top?
- Which kind of insulation is this: basic, supplementary, reinforced, or merely functional?
- Is this barrier protecting the operator or the patient, and does it carry one means of protection or two?
- What is the pollution degree at that location, and which comparative tracking performance group does the insulating material belong to?
Only once those four inputs are settled does the spacing requirement have a definite value. Laying out a board from habit, the way you would for ordinary information technology equipment, is the usual root cause of failures here, because patient-side protection is held to a tighter line than operator-side protection to begin with.
| Barrier location | What has to be confirmed first | Common misjudgement |
|---|---|---|
| Between transformer primary and secondary | How many means of protection this barrier carries, and how the windings are physically separated | Counting only the enamel layer on the wire, and using functional insulation as if it were reinforced insulation |
| PCB traces in the mains part | Pollution degree, whether a coating is applied, and whether that coating process is under control | Assuming that conformal coating means spacing no longer has to be checked |
| Between the applied part and the rest of the circuit | Applied part type and the patient protection required | Laying it out to operator-protection rules and leaving the patient side short |
| Connectors and terminal blocks | Where terminals sit relative to each other when being mated, unmated or worked loose | Measuring only the fully mated state, ignoring where a single detached conductor could reach |
| Between internal live parts and accessible metal | How accessibility is determined | Assuming that if a finger cannot reach it once assembled, it is not accessible |
Coating is the item most often skipped. Whether conformal coating lets you relax a spacing requirement depends on whether the coating forms a continuous, repeatable insulating layer, and on whether you hold process validation and incoming-material control evidence for it. Without that evidence, the coating counts as absent during evaluation. The cost of getting this wrong is very concrete: boards are already prototyped, perhaps already in pilot production, but the evaluation is carried out on the uncoated basis, and you are back to re-routing.
Creepage and clearance carry a hidden cost as well: they set the size of the product. Discovering a shortfall after the mechanical design is frozen is usually not a matter of moving one trace, it is a matter of the enclosure no longer being big enough. Which is why this item belongs in design review with the mechanical engineer in the room, not at prototype stage.
Protective earthing: the failure is almost never in the green-and-yellow wire
Earthing problems rarely start in the conductor itself. They sit at the two ends and at the connection points along the path.
- Surface finishing. Once the enclosure has been anodised, powder-coated or painted, your own process has cut the conductive path. None of this is visible on a drawing; the only things that catch it are first-article measurement and a local masking requirement written into the process documentation.
- Fastening. An earth screw shared with signal or shield wiring, or a joint with no serrated anti-loosening washer and no locking feature, will drift in connection impedance after the equipment has been shipped, taken apart and reassembled.
- Detachable power supply cords. For equipment with a detachable cord, the earth path has to be looked at all the way to the plug. Measuring only inside the equipment misses the contact resistance between plug and inlet.
- Choice of test points. On the same unit, picking different accessible metal parts as test points can produce opposite conclusions. A pre-submission self-check means listing every accessible conductive part and confirming each one, not measuring whichever spot happens to be convenient.
The self-check needs no instruments. When the prototype arrives, do three things: list the accessible conductive parts, mark the earth path for each one, and confirm the fastening and anti-loosening arrangement at every connection point along those paths. That alone stops a meaningful share of rejections. If an accessible conductive part is neither earthed nor backed by an explainable insulation arrangement separating it from live parts, it is already a nonconformity waiting to be written up — you do not need a test result to know that.
Leakage current: passing in normal condition does not count as passing
The item that most often stretches the schedule is leakage current, because it is the textbook case of "scrapes through in normal condition, fails in single fault condition". The rule is one sentence: leave your design margin for the single fault condition, not for the normal condition.
Single fault conditions include an interrupted protective earth conductor, an interrupted neutral and reversed supply polarity. How the equipment behaves in those states can be a long way from its normal-condition behaviour, and it is those states that decide the verdict.
| Leakage current type | Main source | Common remedies | Side effect |
|---|---|---|---|
| Earth leakage current | Common-mode filter capacitors at the supply input, coupling capacitance between switching transformer primary and secondary | Reduce the filter capacitance, move to a transformer with a lower-coupling construction | Common-mode noise rises and EMC risk increases |
| Touch (enclosure) leakage current | Coupling between the enclosure and live parts, redistribution of current paths when the earth is interrupted | Improve the insulation arrangement, revise how the enclosure is earthed | Mechanical changes, possibly tooling changes |
| Patient-related leakage current | How the applied part is isolated from signal ground, and the choice of isolation components | Raise the isolation grade, change to interface components with better isolation performance | Cost goes up and the signal chain has to be re-verified |
There is a conflict here worth knowing about in advance: enlarging the common-mode filter capacitance to suppress common-mode noise is a routine EMC remedy, but it pushes earth leakage current straight up. Cutting that same capacitor to fix leakage current can put conducted emissions at risk. So these two subjects should not be handed to two teams in two separate phases; look at them together in one exploratory round and you avoid changing the design back and forth. The EMC-side investigation approach is covered under EMC testing services and is not expanded here.
Applied part type also drives difficulty directly. For equipment with type B, type BF and type CF applied parts, the patient-related requirements tighten step by step, and a type CF isolation scheme essentially has to be settled at circuit architecture stage; changing it at prototype stage is a rebuild. If this is not nailed down at design review, it is hard to retrofit later.
Pre-submission self-check
| Self-check item | How to do it | What happens if you skip it |
|---|---|---|
| Insulation diagram | From the mains input through to the applied part, mark the insulation type and number of means of protection for every barrier | Manufacturer and test house read the same barrier differently, and you only find out after testing that the acceptance basis differed — a wasted round |
| Accessible parts list | List every accessible conductive part with its earth path or its insulation justification | Miss one accessible part and the report conclusion no longer covers the whole device |
| Critical component list | Power modules, transformers, optocouplers, filter capacitors, fuses and similar, with part numbers and safety approval status | Missing component documentation stops testing mid-stream while you chase paperwork, and your slot gets taken by another project |
| Single fault dry run | Under safe, controlled conditions, simulate an interrupted protective earth and reversed polarity yourself and watch what the equipment does | Discovering single fault failures at the laboratory means queueing again after the fix |
| Markings and accompanying documents | Confirm that markings and the safety information in the accompanying documents match the actual configuration | These nonconformities are cheap to fix but still cost you another retest round |
The value of the self-check is that it separates design problems from documentation problems. Documentation problems can be fixed on the spot; design problems have to go back to the design stage. Fixing it at design stage costs nothing. Fixing it once the prototype has arrived means rescheduling.
Three changes that keep getting treated as minor
- Swapping the power module. Even with identical input and output ratings, a different safety construction and different primary-to-secondary coupling capacitance will change leakage current. A change like this is equivalent to a fresh evaluation.
- Changing the enclosure supplier or the surface finishing process. Earth continuity can be lost outright, and nothing about the appearance gives it away.
- Adding one external interface. If the new interface has any electrical relationship to the patient side, it introduces a new isolation path, which means walking through the insulation diagram again.
Internally, these three often bypass the evaluation process entirely and surface only at the next surveillance sample or system audit. The rule can be kept simple: any change that touches the insulation path, the earth path or the power architecture is handled as a fresh evaluation, regardless of how cheap the part is.
Report acceptance and how to pace the work
Whether a report is accepted depends on whether the item tested matches the product you are filing, whether the sample state is traceable, and whether the acceptance basis in the test plan is written down clearly. It should be noted that an accreditation mark only demonstrates that the laboratory holds the relevant technical competence within its accredited scope; it is not a commitment regarding market access in the target market, and the applicable acceptance position remains subject to review by the receiving authority. Where a standard is cited, the currently effective version of the standard text governs.
The pacing we suggest has three stages: a desk-based prediction before the design is frozen, an exploratory round on the first prototype, and the formal project after the design is finalised. Skipping the first two and going straight to formal testing looks like saving a round of fees; in practice it puts the entire remediation risk onto your schedule.
If you want someone to walk this line with you
If you have equipment stuck on an electrical safety item, or the design is not frozen yet and you want the risk points cleared first, send over the mechanical drawings, the power architecture and the critical component list. We will run a desk-based prediction on the basis described above, and then decide together whether an exploratory test round is worth scheduling. Scope of work and sample submission requirements are on electrical safety testing services and sample submission requirements; the overall workflow is on our testing process. Call +86 132 4819 8029, or simply request a quote.