Effective indoor air quality monitoring means measuring three different things with three different technologies: carbon dioxide with an NDIR infrared cell accurate to about ±(30–50 ppm + 3% of reading), PM2.5 with a laser scattering counter accurate to roughly ±10 µg/m³, and volatile organic compounds with either a low-cost metal-oxide sensor that reports only relative trends or a photoionisation detector that gives absolute readings in ppb. Practical targets for occupied commercial space are CO&sub2; below 800–1,000 ppm, PM2.5 below 15 µg/m³ as a 24-hour mean, and TVOC below 300–500 µg/m³. Sensor placement and calibration drift decide whether the data is worth acting on.
What each sensor actually measures
An NDIR CO&sub2; sensor passes infrared light through a sample chamber and measures absorption at 4.26 µm. Single-channel designs rely on automatic baseline correction (ABC), which assumes the sensor sees fresh outdoor air near 420 ppm at least once every few days – a valid assumption in an office that empties overnight, but wrong in a 24-hour facility, where ABC slowly drags the reading downward. Dual-channel or dual-beam sensors avoid that dependence and hold drift to roughly 20–50 ppm per year.
Optical PM2.5 sensors count light pulses scattered by particles crossing a laser beam, then convert counts to mass using an assumed density and refractive index. Two limitations follow. They under-detect particles below about 0.3 µm, which is precisely where combustion aerosol concentrates, and they over-read when relative humidity exceeds 70% because hygroscopic particles swell – typical over-reading is 20–50% at 85% RH unless the firmware applies humidity compensation.
VOC measurement is the least standardised. Metal-oxide semiconductor sensors respond to a broad mix of reducing gases with no compound specificity, so a reading of "450 ppb TVOC" is a relative index, not a laboratory result. A photoionisation detector with a 10.6 eV lamp measures a wide range of solvents in absolute ppb but does not detect methane, formaldehyde or carbon monoxide. Formaldehyde needs its own electrochemical or spectroscopic sensor, typically resolving 10–20 ppb.
Sensor comparison for building monitoring
| Parameter | Typical technology | Typical accuracy | Useful range | Calibration or replacement | Action threshold |
|---|---|---|---|---|---|
| CO&sub2; | NDIR, dual channel preferred | ±(30–50 ppm + 3%) | 0–5,000 ppm | Field check every 12 months | Boost outdoor air above 1,000 ppm |
| PM2.5 | Laser light scattering | ±10 µg/m³ or ±10% | 0–500 µg/m³ | Clean optics yearly, replace in 3–5 years | Increase filtration above 25 µg/m³ |
| TVOC | Metal-oxide semiconductor | Relative index only | 0–5,000 ppb equivalent | Replace every 3–5 years | Investigate a sustained rising trend |
| VOC, absolute | Photoionisation detector | ±10–20% of reading | 1 ppb–20 ppm | Lamp cleaning every 3–6 months | Local exhaust above 500 ppb |
| Formaldehyde | Electrochemical | ±20 ppb | 0–2 ppm | Replace cell every 2–3 years | Ventilate above 80 ppb |
| Carbon monoxide | Electrochemical | ±2 ppm | 0–500 ppm | Bump test every 6–12 months | Alarm and purge above 25 ppm |
Where to put the sensors
- Mount wall sensors at breathing height, 1.1–1.7 m above the floor, which represents occupant exposure far better than a ceiling position.
- Keep at least 1 m clear of doors, operable windows, supply diffusers and exhaust grilles, so the sensor reads room air rather than a jet.
- Avoid direct sunlight and external walls; solar gain can bias the internal temperature reference and shift NDIR readings.
- Provide roughly one sensor per 400–500 m² of open plan, plus one per enclosed high-occupancy room such as a meeting room or classroom.
- Place duct-mounted sensors in the return, not the mixed air plenum, and never within three duct diameters downstream of a branch.
- Add one outdoor reference sensor per building; without it, PM2.5 data cannot distinguish an indoor source from an outdoor episode.
When to monitor what
- Monitor CO&sub2; when occupancy varies by more than about two to one, since demand-controlled ventilation typically cuts conditioning energy 10–25% in meeting rooms, lecture halls and auditoria.
- Monitor CO&sub2; when a green-building credit or wellness certification requires continuous ventilation-rate evidence.
- Monitor PM2.5 when the building sits in an urban corridor, near construction, or in a region with seasonal smoke, and when recirculating filtration can be modulated in response.
- Monitor formaldehyde and specific VOCs when the space is newly fitted out, or where processes use solvents, adhesives or printing chemicals.
- Use only a relative MOS VOC index when the goal is simply to trigger extra exhaust in washrooms, kitchens or copy rooms – and do not report its numbers as compliance data.
- Add carbon monoxide monitoring when the space adjoins a loading bay, car park or fuel-fired plant room.
- Skip continuous monitoring when nobody owns the response; a dashboard with no control sequence or maintenance trigger produces alarm fatigue and nothing else.
Turning readings into control action
Data only pays back when it is wired into a sequence. A common demand-controlled ventilation strategy resets outdoor air linearly between 600 ppm and 1,000 ppm of CO&sub2;, holding a minimum ventilation floor so that building materials and background emissions are still diluted when the space is empty. On the particulate side, when the outdoor reference exceeds 35 µg/m³, shifting the economiser toward recirculation through a MERV 13 filter usually lowers indoor PM2.5 faster than pulling in more outdoor air. For VOCs, a rising trend that persists after occupancy ends generally indicates a material or process source rather than a ventilation shortfall, and should trigger investigation instead of a fan-speed increase.
Data quality problems that mislead building operators
Three artefacts account for most false conclusions drawn from IAQ dashboards. The first is self-heating: a sensor board dissipating 0.3–1 W inside a sealed enclosure can read 1–2 °C above room temperature, which distorts the relative humidity calculation and therefore any humidity correction applied to particle mass. The second is averaging window mismatch, where a one-minute CO&sub2; sample is compared against a guideline written as an eight-hour mean, making transient peaks look like violations. The third is uncorrected sensor ageing, since an NDIR cell drifting 40 ppm per year will read 200 ppm low after five years and quietly suppress the ventilation the sequence should be calling for. Fix these by logging raw values alongside averaged ones, recording firmware and calibration dates, and reviewing one week of trend data before changing any control setpoint.
Keruite manufactures air handling units, energy recovery ventilators and fan coil units at an 80,000 m² facility in Dezhou, China, with an in-house AMCA-210-style test laboratory, and can factory-fit CO&sub2; and PM2.5 sensing into the return or mixing section so that demand-controlled sequences arrive pre-wired.
FAQ
What CO2 level indicates poor ventilation?
A sustained reading above 1,000 ppm generally signals that outdoor air per person is below typical design values, and above 1,400 ppm the shortfall is significant. Judge the difference from outdoor concentration, currently around 420 ppm, rather than the absolute number alone.
Are low-cost PM2.5 sensors accurate enough for building control?
Usually yes for trend-based control, within about ±10 µg/m³ when humidity compensation is applied. They are not a substitute for gravimetric reference instruments in regulatory reporting, and they over-read by 20–50% above 85% relative humidity if uncompensated.
Why does my TVOC sensor read high with no obvious source?
Metal-oxide sensors respond to alcohols, cleaning agents, hand sanitiser, perfume and cooking vapour without distinguishing them. Sudden spikes after cleaning are normal. Look at whether the baseline has shifted upward over weeks, which is the trend that actually matters.
How often should IAQ sensors be recalibrated?
Check NDIR CO&sub2; sensors annually against a known reference, clean PM optics annually and plan replacement in 3–5 years, bump-test electrochemical CO cells every 6–12 months, and clean PID lamps every 3–6 months. Log every intervention against the sensor identifier.