“The efficiency is 99.997%, so how can the filter have a leak?” Because an average can hide a local defect. The answer becomes straightforward once three terms are separated: integral efficiency is a whole-filter result, local efficiency concerns a small sampled region, and a leak scan is a method used to look for unacceptable local penetration.
They are related, but they are not three interchangeable certificates of quality. Strictly speaking, they are not even three equivalent kinds of test: two describe performance quantities and one describes a way of inspecting the filter.
That distinction matters when approving a delivery, writing a specification or comparing a factory certificate with an installed test.
Integral efficiency: how much gets through overall?
For a particle-counting measurement under defined conditions, penetration is the downstream concentration divided by the upstream concentration, after the applicable corrections. Efficiency is one minus penetration. Expressed as percentages:
Efficiency (%) = 100 − penetration (%).
An integral result represents the filter as a whole. It asks how much of the challenge passes through all the flow paths combined. For high-efficiency classification, the test particle size matters: the relevant standards use the most penetrating particle size, or MPPS, rather than assuming that every filter's weakest point occurs at exactly 0.3 µm. ISO 29463-5:2022.
Imagine an illustrative measurement giving 99.997% integral efficiency. The corresponding penetration is 0.003%, or 30 particles per million upstream particles under the measurement's stated conditions. That conversion does not predict the particle count in a cleanroom. The room has other sources, air paths and removal mechanisms.
It also says nothing by itself about where those particles passed through. Uniform small penetration across the pack and a concentrated defect can contribute to an overall result in very different ways.
Local efficiency: what is happening in this region?
A local measurement samples a limited part of the downstream area. Its result is tied to the specified probe, flow, sampling time and evaluation method. It is not the efficiency of an isolated square cut out of the filter.
The distinction is deliberate. A highly efficient pack surrounding a small defective region can dilute that region's contribution to the whole-filter result. Local assessment prevents an acceptable average from concealing excessive penetration at a particular location.
EN 1822's H14 limits illustrate the separation: minimum integral efficiency is 99.995%, while minimum local efficiency is 99.975%. These correspond to maximum penetration values of 0.005% and 0.025%, respectively. They are classification values at MPPS, not a universal installed photometer leak limit. EN 1822-1:2019, Table 1.
Do not reverse the labels. Reporting “99.975% H14 efficiency” without saying local makes the filter appear to miss the H14 integral requirement. Reporting a maximum local penetration of 0.025% as the filter's overall penetration makes a different error.
Leak scan: where does the result exceed the criterion?
A scan moves the sampling probe across the area prescribed by the method. The objective is to find and evaluate local indications, including relevant pack-to-frame regions, rather than take one convenient downstream reading.
ISO 29463-4 describes a reference scanning method for determining leakage of filter elements. The existence of alternative methods for some classes does not mean all tests called “leak tests” have identical sensitivity or report the same quantity. ISO 29463-4:2011.
A scan report may show a map, the maximum local penetration, a threshold result or other method-specific records. “Scanned” alone does not tell the buyer which aerosol, instrument, test flow or acceptance criterion was used. It also does not tell the buyer whether the scan was carried out on the filter at manufacture or on the completed installation.
Specify the method when it matters. An order requiring an individual particle-counter scan should say so; the word H14 should not be used as shorthand for every detail of the agreed testing route.
A numerical example that exposes the difference
Consider this deliberately simplified teaching example. Suppose 99.9% of the airflow passes through a region with 0.001% penetration, while 0.1% of the airflow passes through a defective region with 1% penetration. Assume the same upstream concentration and ignore sampling losses.
The combined penetration is:
0.999 × 0.001% + 0.001 × 1% = 0.001999%.
The corresponding integral efficiency is 99.998001%. As a numerical average, that exceeds the H14 minimum integral efficiency. Yet the defective region's assumed penetration is far above the H14 local limit.
This is not a model of an actual scan probe or a prediction of what an instrument would display over a small hole. It demonstrates why a whole-filter average cannot replace local assessment. Whether the real test detects a particular defect depends on the approved method and its sampling conditions.
The practical point is simple: do not approve a filter solely because one impressive percentage appears on the certificate.
A photometer percentage adds another distinction
A photometer detects the combined light scattered by the sampled aerosol. It does not provide a particle-by-particle, size-resolved efficiency curve. A particle counter uses a different measurement basis. Consequently, a photometer percentage should not be read as though it were automatically the MPPS particle-counting penetration printed on a factory certificate. NIOSH: aerosol photometer measurement principles.
Neither instrument is “wrong” because the percentages differ. The comparison is wrong if the detection basis, aerosol distribution and test boundary are ignored. Installed leak testing has a defined purpose within cleanroom verification; factory element testing has its own classification purpose. IEST explicitly addresses factory, preinstallation and installed leak testing as distinguishable contexts. IEST-RP-CC034.
Ask for the missing noun after every percentage
When reviewing a result, complete these descriptions:
| Result shown | Information needed to interpret it |
|---|---|
| 99.997% | Integral or local efficiency; particle size; test flow; method |
| 0.010% | Penetration or leakage indication; detection basis; location; applicable limit |
| Pass | Which test, which standard and which agreed acceptance criterion? |
| MPPS | Measured or established how, and for which media and velocity conditions? |
| Scan complete | Which surfaces and interfaces were included? |
Scroll the table sideways to see all columns →
Then check whether the reported result belongs to the individual element being purchased. A media development curve, a product-family datasheet and a serial-numbered filter record answer different questions.
If a whole-filter result passes and a valid local result fails, the average does not overrule the defect. That is precisely why the distinction exists.



