Why cleanrooms fail certification

What actually gets tested

A cleanroom does not fail certification. A parameter does. The certificate describes one room measured against its own specification, on one day, in one occupancy state, and what comes back is a set of separate results rather than a single verdict. So when a report arrives with a failure on it, the useful question is never why the room failed. It is which parameter went out, and what moved it.

The battery run on a conventional room is short, and it has not changed much in years:

• airflow velocity and uniformity, on unidirectional systems
• airflow volume, converted to air changes per hour, on non-unidirectional systems
• the installed filter leak test
• room particle counts
• enclosure pressurization
• temperature and humidity

The methods are the ones described in ISO 14644-3:2019 and in IEST-RP-CC006.4. The acceptance criteria are not: those come from the room's own specification, which is the document that says what each number has to be for your process. That split is worth holding onto when a result comes back marginal. The standards tell a tester how to measure. Your specification decides whether the measurement passed.

Where each test actually breaks

The middle column below is what the test is really looking at, which is not always what its name suggests. The right-hand column is industry consensus across the trade rather than anything a standard publishes, and it is offered as that: the causes a certifier sees over and over, not a list anyone promulgated.

Read down the right-hand column and one shape repeats in four of the six rows: something in the air path changed, and nobody re-tested afterwards. That is the single largest category of failure in this trade, and almost none of it is discovered by the people who caused it.

A leak test is not an efficiency test

This is the most common misunderstanding on a certification report, and it matters because the two tests answer different questions about different objects.

FDA's 2004 aseptic processing guidance, which is guidance and not regulation, gives the purpose of a leak test as detecting leaks from the filter media, the filter frame, or the seal. An efficiency test, in the same document's words, is a general test used to determine the rating of the filter. The rating belongs to the filter as a product and was established long before it reached your ceiling. The leak test belongs to the installed system: media, frame, gasket, and the joint between the filter and the grid holding it.

So a filter can carry a perfectly good efficiency rating, on paper and in fact, and still leak. The same guidance puts an intact HEPA filter at retaining at least 99.97 percent of particulates greater than 0.3 µm in diameter, and that figure describes the media rather than the installation. A gasket that lost its compression when a trade opened the ceiling is a leak past a filter that is doing exactly what its rating claims.

Two further points from the same guidance are worth knowing before anyone reads a leak-test result:

• A single probe reading equivalent to 0.01 percent of the upstream challenge is treated as indicative of a significant leak, and calls for replacement of the filter or, where appropriate, repair in a limited area, with a confirmatory retest in the area of any repair.
• Leak testing without introducing a sufficient upstream challenge of particles of known size upstream of the filter is, in that guidance's own words, ineffective for detecting leaks.

The second one is the part a buyer cannot audit from the report. A leak test run without an adequate challenge produces a clean result and tells you nothing, and the page it prints on looks the same either way. That is the honest practical case for credentialed independent supervision rather than a self-issued result. An independent party carries a reputational stake in its own method, and an inspector reading the report is going to ask who signed it and against what.

The failures that are not the room's fault

Pressure cascade is a relationship between rooms, not a property of one room. That single fact accounts for a large share of the failures that arrive with no plausible cause inside the four walls being tested.

The same FDA guidance, again guidance rather than regulation, recommends a positive pressure differential of at least 10 to 15 Pa between adjacent rooms of differing classification with the doors closed, and at least 12.5 Pa where an unclassified room adjoins an aseptic processing room. Nothing in those numbers is something your room owns by itself. Rebalance the suite next door, prop a door, block a return path, and your differential moves without one thing changing inside your walls.

The same reading applies to the two other figures that guidance carries. It puts unidirectional airflow velocity at 0.45 metres per second, or 90 feet per minute, with a range of plus or minus 20 percent around the setpoint. It says airflow sufficient to achieve at least 20 air changes per hour is typically acceptable for ISO 8 supporting rooms, and that significantly higher rates are normally needed for cleaner ones. Both describe air delivery, and air delivery is a building system that your room shares with its neighbours.

Three outside causes account for most of this category:

• Adjacent construction. That FDA guidance names facility renovations which might be the cause of disturbances to ceiling or wall structures as a reason for additional filter leak testing. Work in a space sharing a wall, a ceiling plenum, or an air handler with your room is work inside your room's pressure boundary, whatever the floor plan says. The controls that keep it out of a running room are in keeping a cleanroom clean while you build.
• A neighbouring room rebalanced. Someone corrects a complaint two doors down, the air moves to answer it, and your cascade absorbs the correction. Nobody involved knew your room was downstream of the adjustment, because nothing in a balancing report says so.
• The building itself. Stack effect, outside air conditions, and a seasonal swing that only shows up for a few weeks a year will move a marginal differential across its limit and then move it back before anyone gets there to look.

None of the three is a defect in the room. All three fail the room.

What happens after a failed result

Two things, in order, and the second is the one that gets skipped.

First, remediate and re-test the failed parameter. A failed result is not advisory. The room sits outside its class until a measurement says otherwise, so the corrective work and the confirming test together are what closes it out, and neither one alone does.

Second, investigate whether the other parameters share the root cause. Air delivery sits underneath most of the battery, so a cause large enough to fail one parameter is usually large enough to have moved two others that happened to stay inside their limits on the day. A velocity failure traced to loaded filters is saying something about the counts and the differential that passed beside it, and a report that re-tests only the failed line leaves that unread.

Then there is the part that decides what the whole thing costs. A failure caught on schedule is a maintenance item. One caught by an auditor is a finding. The remediation work is identical in both cases, and the difference between them is set months earlier by the certification cadence a room is actually on, which how often a cleanroom should be certified covers in full.