The IP Rating Red Flag Hiding in Your Supplier's Datasheet

A datasheet IP number is usable in a qualification file only when a test report sits behind it, and only when that report names the standard edition, the exact enclosure revision that went into the tank, the connector and gland state during the test, the build status and age of the sample, and the party that ran it. Remove any one of those and the claim stops being evidence. It becomes a sentence in a PDF, and a sentence will not carry a PPAP submission or a quality agreement clause.

I keep the reports that taught me the most in a folder of their own. Not the fakes — fakes announce themselves, usually with a badge-shaped graphic and no test data anywhere near it. The ones worth studying are honest tests on honest hardware that still cannot be used, and the one I hand to new buyers has twenty-eight pages, a clean pass, and no way to tell what part was actually underwater.

You'll recognize it. It's the PDF sitting on your second monitor while the supplier call is still live, the one you opened because the datasheet says IP67 and the program schedule says the qualification file closes on Friday.

Page 1 is a title page. Page 2 is a scope statement that reads like it came off a template. Page 3 is the summary table, and the summary table is where the trouble starts. Four rows:

• Sample description: lithium-ion battery pack
• Standard: IEC 60529
• Test: IPX8
• Result: Pass

That is the entire story on that page. No part number. No drawing revision. No note about whether the connector was capped, mated, or open to the air. No depth, no duration. And it is the page everyone reads, because it is the page that fits in a slide, and the page your program manager forwards with the subject line "IPX8 confirmed."

The rest of the document is better than that. Somewhere around page 9 there are photographs of a tank and a pressure trace, and on the last page a technician's initials and a date. It is a real test. It is just not obviously a test of your part.

There is no certificate, and that is the first thing to unlearn

IEC 60529 is a test method, not a registry. Nobody issues an "IP68 certification." What exists is a report from a laboratory, or from the supplier's own rig, describing what went into the water and what came out of it. So when the datasheet carries a badge and the document behind the badge is a one-page declaration, you have not received evidence. You have received a claim with better typography.

This matters more for battery packs than for, say, a junction box, because of seam count. A pack can have a welded or gasketed housing, a vent, a service disconnect, a connector interface, a pressure-equalization membrane, and a label, and every one of those is a place where the seal either is or is not part of what was tested. A single IP number flattens all of it into two digits.

What the summary page left out

The configuration

I want the drawing number and revision on the report, matching the drawing in your file. Not "similar to." Not "representative unit." When the tested revision and the revision you are buying are different, you are holding a report about a different part — and the difference can be as small as a gasket durometer change, or a relocated screw boss that changes how evenly the cover compresses the seal.

The connector state

This is the one that costs people real money. A pack tested with the connector port plugged by a shipping cap is a different product from a pack tested with the mating half installed and the harness strain-relieved. Both can produce a report that says IP67. Only one of them describes what your pack looks like once it is on the vehicle.

So I ask it plainly: during the test, what was in the connector? Was the mating connector installed, was there a cap, or was it open? If the answer is a cap, the next question is whether that cap ships with the unit and stays on in service. If it doesn't, the report describes a condition that will never exist on your build.

The build and the age of the sample

Reports get run on prototypes, on hand-built units, on units pulled from a pilot run, and on units off the production line. Those are four different claims wearing the same cover page. A housing machined from a block and a housing molded from the production tool do not seal the same way, because the molded part has a weld line, a gate location, and a shrink profile the machined part never had. And a report on a unit that went into the tank the same week it was assembled says nothing about what happens after six months of thermal cycling and a few hundred connector mate cycles.

The practical move is to log the gap. The qualification file should state, in writing, that the IP claim rests on prototype-level evidence and that a production-unit re-test is required before final approval. That one line stops the claim from quietly upgrading itself from DV to PV while nobody is watching.

Depth and duration for IPX8

The standard deliberately leaves immersion conditions to whoever is making the claim. Two packs can both carry IPX8, one tested at a meter for thirty minutes and one at ten meters for a day. If the report does not state depth and duration, the number cannot be compared to anything, including your own use case. Same problem for IPX6 and IPX6K, where jet pressure and distance matter, and for IPX9K, where temperature and nozzle distance do. If the conditions are not written down, you are holding a label rather than a specification.

Who ran it

An accredited third-party laboratory and the supplier's own tank are both legitimate sources of evidence, but they carry different weight and they get reviewed differently. When the result came out of the supplier's own facility, the audit has to cover the rig itself. Where is the depth measured from? When were the gauges last calibrated? How was the sample handled between build and test? What does the logbook entry for that day say? Does the serial number in the photograph exist in the build records? A report that cannot be tied back to a unit in the paperwork is a printout.

Photographs carry more weight than they should

I lean on them more than I'd like to admit. A photograph of the unit in the fixture tells you the configuration that actually went in. A photograph of the same unit opened afterward tells you where the water reached and where it didn't, and which points the technician checked. A file that shows a fixture and then jumps straight to a signed conclusion — no teardown images, no indication of how many units went in and how many came back dry — has given you a result without a trail.

Sample size belongs in the same conversation. One unit that passed is a data point. Five units tested with one failure and a note about what was done with the failing unit is an actual test program, and that note is often the most informative page in the document.

The report does not cover aging, and this is where teams relax

Ingress testing happens on units in the condition the laboratory received them. Thermal cycling, vibration, mechanical shock, UV exposure, salt fog, and connector mating cycles generally sit outside that scope. A clean IP67 report on a fresh pack and a separate environmental qualification plan are not the same evidence, and passing the first does not excuse you from the second. Gaskets take compression set. Potting compounds and adhesives behave differently across temperature. A pack that sealed perfectly in week one can weep at a seam in year three, and the IP report was never asked that question.

If you want the ingress claim to hold for the life of the program, the environmental qualification plan has to say at which point in the sequence the pack gets re-checked. The honest answer is usually not "at the end of everything."

When the tank belongs to the supplier

Most of the packs I've worked on were ingress-tested at the supplier, which is fine, and which means the audit is where the claim gets validated. What I look at, roughly in order: the rig and how depth is established, the calibration status on the gauges and the date on the sticker, the sample storage and handling between build and test, the raw records behind the report — logbook, setup sheet, technician's notes — and the link between the serial number in the photographs and the unit in the build record.

None of that has to be adversarial. Suppliers who run good testing are usually proud of it and will walk you through the tank in ten minutes. The ones who get uncomfortable at the word "logbook" are telling you something useful.

Put the requirement in the file before you need it

The most effective change I've made to this process is boring. The IP evidence line goes into the quality agreement and onto the PPAP checklist, alongside the other qualification documents, so it arrives with the rest of the package instead of after someone notices it is missing in a design review. The line enumerates what "usable" means: standard edition, tested part number and revision, connector state, sample build and age, depth and duration for IPX8, quantity tested, photographs, and the testing party.

When a supplier sees that list at the start, the report comes back complete, because they know exactly what page 3 is going to be compared against. That is the whole trick. You are not asking for more paperwork. You are asking for the four or five facts that turn a number into evidence, and asking for them early enough that nobody has to go back to the laboratory twice.

A datasheet IP rating is a conclusion. The folder behind it is either evidence or it isn't, and you can usually tell which one you're holding within about four pages.