Air quality

Most air-quality sensors output a single number: TVOC, the sum of all volatile organic compounds. That number rises when someone installs furniture that outgasses formaldehyde — and it rises just the same when someone disinfects with alcohol. Anyone who wants to base a decision on it needs the breakdown, not the sum.

What changed on 6 August 2026

Since 6 August 2026 a tightened restriction on formaldehyde in articles applies across the European Union. Regulation (EU) 2023/1464 adds entry 77 to Annex XVII of the REACH Regulation and limits how much formaldehyde an article may release into indoor air.

Conversion at 25 °C and 1013 hPa, molar mass 30.03 g/mol. Source of the limits: Regulation (EU) 2023/1464 amending Annex XVII of the REACH Regulation, entry 77.

For manufacturers of furniture, floor coverings, insulation and electronics this is a hard requirement with a testing obligation. And it shifts the yardstick: formaldehyde is now regulated in an order of magnitude where a sum reading says nothing at all.

Why a sum reading is not enough

A classic VOC sensor measures a change in the resistance of its sensing layer. Reducing gases lower that resistance — practically all of them, to differing degrees, but indistinguishably. The result is a single number reported as "TVOC".

That number has three weaknesses, and indoors all three occur at once:

• It is not substance-specific. 200 ppb of ethanol and 200 ppb of formaldehyde produce different deflections, but both are only a deflection. The reading cannot be worked backwards to say which substance caused it.
• It weights the wrong things. The substances that matter for health are not the ones producing the largest deflection. A cleaning routine routinely drowns out a persistent formaldehyde source that would be orders of magnitude more important.
• It admits no threshold. Any threshold low enough for early detection is triggered by everyday activity. Raise it for the sake of quiet and you end up detecting only gross events.

The substances the datasheet warrants limits for

NZGS 2 carries four sensing elements on a die of 1.15 × 1.15 mm, with layers of tin dioxide and tungsten trioxide. Because each element has a different layer, a different position relative to the heater and therefore a different operating temperature, one gas event produces four different signal traces. The substance is determined from the ratio between them, not from the height of a single deflection.

Values warranted in datasheet rev. 3.3.0, determined under laboratory conditions: 24 h stabilisation, 25 °C, 45 % relative humidity. These are typical characteristics that vary from chip to chip. Ethylene (400 ppb) is the fifth warranted value and belongs to the food application.

Detected beyond these, but without a warranted limit of detection in the datasheet: BTEX — benzene, toluene, ethylbenzene and xylene — nitrogen dioxide, ozone, ammonia, hydrogen sulphide, hydrogen, methane and nitrous oxide. Measured values and test conditions for those are available on a project basis.

Formaldehyde in relation to the limit

The warranted value of 30 ppb corresponds to roughly 0.037 mg/m³ at 25 °C and 1013 hPa. The stricter of the two new limits is 0.062 mg/m³, that is about 50 ppb.

That is deliberately phrased as a ratio and not as a compliance claim. A limit of detection states the concentration from which a substance is reliably detected at all — not the accuracy with which its concentration can be reported. Reporting a concentration requires calibration for the specific use case.

Where the interferents come from

For a gas sensor, an occupied indoor space is a restless environment. The substances that most reliably mislead a TVOC sensor are precisely the harmless ones:

• Ethanol from disinfectants and cleaning agents. Ubiquitous indoors since 2020, at concentrations that mask any formaldehyde source.
• Acetone and solvents from cosmetics, adhesives and paints.
• Cooking, which releases a whole mixture — aldehydes, alcohols, fatty acids.
• People themselves. Exhaled air contains acetone, ethanol and hundreds of other volatile compounds in measurable quantities.

This is where a measurement proves useful or useless. A sensor that can tell ethanol from formaldehyde reports nothing while someone disinfects and something when a new cabinet arrives. A sum reading does the opposite — it reports a lot during disinfection and little for the cabinet, because one event is brief and strong and the other persistent and weak.

Carbon monoxide is a different case

CO is not a VOC and is detected on a different layer: tungsten trioxide rather than tin dioxide. The warranted value is 900 ppb, that is roughly 0.9 ppm.

For context: that is an order of magnitude at which CO serves as an indicator of incomplete combustion — a gas appliance not burning cleanly, a fireplace with poor draught, a vehicle in an adjoining garage.

Outdoor air: nitrogen dioxide and ozone

Different substances govern outdoor air, above all nitrogen dioxide from traffic and combustion, and ground-level ozone. Both are detected via the tungsten trioxide layer. The datasheet warrants no limits of detection for them — values and test conditions can be determined on a project basis.

It becomes interesting when both directions come together: a device that captures VOCs and CO indoors and NO₂ and O₃ outdoors can answer whether ventilating right now improves the air or makes it worse. For ventilation control that is the decisive piece of information, and it cannot be derived from a sum reading.

What the sensor does not do in this application

• It is not a reference instrument. Legally binding statements require standardised procedures with calibrated equipment. A MOx sensor provides continuous observation, not a defensible measurement.
• It does not demonstrate REACH compliance. The entry 77 limit is verified on articles in a test chamber under EN 16516, not on room air.
• It has to be trained for the use case. Which substances separate cleanly in feature space depends on the mixture and on the interferents at the installation point. Without a measurement campaign there is no dependable classification.
• It ages and it is vulnerable. Baseline drift through ageing is compensable if temperature and humidity are measured alongside. Silicone compounds, by contrast, poison the tin dioxide layer irreversibly — that belongs in the choice of housing, seals and adhesives.

How a project begins

At the outset stands the question of whether the target substances separate cleanly in the actual environment. That cannot be derived from specifications, only measured. The evaluation kit with read-out electronics and software exists for this: it records the four signal traces, extracts the features and shows whether the target substances form separate regions in feature space.

What gets recorded is reference air for the baseline, the target substances at known concentrations, and the interferents that realistically occur at the installation point. The last is routinely underestimated and decides the later false-alarm rate. Added to that are the boundary conditions temperature and humidity across the expected range, plus several sensor specimens, because characteristics vary.

For the component itself: QFN-8 at 2.8 × 2.8 × 1.6 mm, around 70 mW power consumption, supplied on tape and reel. All specifications are in the datasheet extract, the sensing principle on the technology page.

Sources: Commission Regulation (EU) 2023/1464 of 14 July 2023 amending Annex XVII to Regulation (EC) No 1907/2006 (REACH) as regards formaldehyde and formaldehyde releasers; NANOZ NZGS 2 datasheet rev. 3.3.0.