Skip to content
AkerminMeasure. Monitor. Improve. Compare Products
Learning Center

NDIR against metal oxide, and why two carbon dioxide monitors disagree

Two monitors on one desk can differ by hundreds of parts per million. Part of that gap is the sensing principle, part is the baseline each device assumes, and part is simply where each one sits. This article separates the three so a reading can be judged instead of averaged.

As an Amazon Associate, Akermin earns from qualifying purchases. Product links on this page are affiliate links.

An infrared monitor measures carbon dioxide by how much of a specific infrared band the gas absorbs, which is a direct physical count of the molecules in the light path. A metal oxide monitor measures the electrical resistance of a heated film that reacts with other gases entirely, and converts that resistance into a carbon dioxide looking number. Put the two side by side and they are not two attempts at the same measurement, so a disagreement of hundreds of parts per million is the normal outcome. Where both devices are infrared, the remaining gap comes from baseline assumptions, placement and averaging, in roughly that order of size.

Infrared absorption gives a direct count, a heated oxide film does not

Carbon dioxide absorbs infrared strongly in a narrow band near four and a quarter micrometres, and almost nothing else in ordinary room air absorbs there. A non dispersive infrared bench exploits that directly. A broadband source shines down a chamber, a filter selects that band, and a detector reports how much light survived the trip. More molecules in the path means less light arrives, and the relationship follows an exponential law in the product of concentration and path length. The word non dispersive simply means the wavelength selection is done with a fixed filter rather than a prism or a grating.

A metal oxide film works on an unrelated principle. A tin oxide layer is heated to a few hundred degrees, and oxygen from the air adsorbs on its surface and traps electrons, raising its resistance. When a gas that can be oxidised reaches the hot surface it reacts with that adsorbed oxygen and returns the electrons, and the resistance drops. The size of the drop depends on how much reducing gas is present. Carbon dioxide is already fully oxidised and has essentially nothing to give, so the film barely notices it.

That is the whole of the first disagreement. One device counts the target molecule. The other watches a proxy and applies a conversion. The conversion can be plausible in a room where people are the only source of anything, and it falls apart the moment a solvent, a candle or a kitchen enters the picture. The pillar explainer on infrared and metal oxide sensing compares the two technologies in more depth, and how carbon dioxide monitors work walks through the optical bench itself.

Single beam against dual beam, and what the reference channel fixes

An infrared bench has one structural weakness. The reading is derived from how much light reaches the detector, so anything that reduces the light looks exactly like more gas. The source dims as it ages. Dust and condensate settle on the windows. The filter degrades. Each of these pushes the reported concentration up over months, in a direction that never reverses, and nothing in the signal itself distinguishes a dirty window from a crowded room.

A second detector solves most of it. The reference channel watches a wavelength that carbon dioxide does not absorb, so its signal responds to source ageing and window fouling and not to the gas. Dividing the measurement channel by the reference cancels the common part and leaves the absorption. That is what a dual beam arrangement buys, and listings that care about long term stability tend to say so. The DANOPLUS carbon dioxide controller states a dual beam low drift infrared sensor with an accuracy figure of plus or minus fifty parts per million, which is the kind of pairing that lets a buyer reason about drift rather than guess.

The reference trick is not a complete fix. It cancels changes that affect both bands equally and leaves anything that affects only one, which includes degradation of the measurement filter itself and a film on the window that absorbs selectively. That residue is slow, one sided and invisible, and it is the reason even a dual beam bench benefits from an occasional check against outdoor air.

Baseline assumptions that quietly move the whole scale

Almost every consumer monitor sets its own zero. The common routine keeps a rolling record of the lowest concentration seen over a period of days and asserts that this minimum must have been outdoor air, then adjusts the calibration so that the minimum reads the outdoor figure the firmware was told to assume. In an office that empties at night, this is clever and nearly free. In a bedroom used every night with the door closed, or a room with a gas injection system running continuously, the minimum is not outdoor air and the routine drags the whole scale down by the difference.

Manual routines expose the same assumption more honestly. The INKBIRDPLUS infrared air quality monitor documents a field calibration performed by holding three buttons together in an open environment and leaving the unit undisturbed for two hundred seconds, which is a user triggered version of the same idea. The quality of the result depends entirely on the air it was standing in, and an open environment beside a busy road or a flue is a different reference from an open environment in a garden.

Displayed thresholds add a second layer of apparent disagreement that is not a measurement difference at all. The same listing defines three bands, a normal stage from four hundred to seven hundred parts per million, a warning stage from seven hundred to fifteen hundred, and a dangerous stage above that up to five thousand. Another monitor showing an identical number can colour it differently, and to anyone reading the colour rather than the figure the two devices appear to disagree while actually agreeing exactly.

Placement, breath and averaging explain the rest of the gap

Carbon dioxide is not evenly mixed in a room on the timescale a monitor samples. A seated person emits a plume that rises and spreads, so a monitor at desk level a short distance away sees a quite different concentration from one on a shelf. A unit near a door or a trickle vent sits in the incoming air and reads low. A unit in a corner with poor circulation reads high. Moving one of two disagreeing monitors is frequently a more effective diagnostic than calibrating either.

Sampling behaviour then decides how the same air becomes a number. A device updating every second and displaying the instantaneous value shows every breath, while one averaging over a minute looks calmer and lags the room. Neither is wrong, and a side by side comparison of instantaneous against averaged output will show a persistent gap whenever concentration is changing. The Inkbird smart carbon dioxide monitor states real time monitoring on a backlit display with a wireless link to an application at up to a hundred metres unobstructed, which means the figure on the screen and the figure in the history can be inspected separately.

What the listings here state about sensing principle and tolerance

Listing Sensing principle named Accuracy stated Baseline routine stated
DANOPLUS carbon dioxide controller Dual beam low drift infrared Plus or minus fifty parts per million Not stated in the listing bullets, though a day and night detection mode is
INKBIRDPLUS infrared air quality monitor Infrared, named as the sensor type Not stated in the listing bullets Three button field calibration over two hundred seconds in open air
Inkbird smart carbon dioxide monitor Infrared, named as the sensor type Not stated in the listing bullets Not stated in the listing bullets

Two of those three name the technology and stop. That is still worth something, because naming an infrared bench rules out the possibility that the carbon dioxide figure is a conversion from a volatile organic compound sensor, which is the single largest source of disagreement between two monitors. What it does not supply is a tolerance, and without a tolerance there is no way to judge whether a gap between two units is within what both devices admit to. A buyer comparing monitors on this point is really comparing how much the manufacturer was willing to be held to.

Deciding which of the two readings to believe

Work through it in order. Establish the technology of each device first, because if one derives its figure from a volatile organic compound sensor the comparison is over and the infrared reading is the only carbon dioxide measurement in the room. If both are infrared, put them together outdoors in moving air, away from traffic, flues and vegetation, shaded from direct sun, and let both settle properly. Outdoor air is the one reference available without a cylinder, and a device that cannot land near it has a baseline problem rather than a sensing problem.

Then bring them back and place them side by side at the same height for a full day before concluding anything, since a difference that disappears when they share a location was never about the sensors. If a gap survives all of that and exceeds the tolerances both devices publish, the one that cannot be brought back by its own baseline routine is the one to doubt. For the specific case of two units in normal use, see why two monitors show different readings, and for the companion measurement that a heated film device is genuinely good at, see our overview of volatile organic compound monitors.

Frequently Asked Questions

If one monitor shows nine hundred and the other shows fourteen hundred, which is right?

Start by establishing which is measuring carbon dioxide at all. If one is an infrared device and the other derives its figure from a volatile organic compound sensor, there is no sense in which the second is close or far, because it is reporting a different quantity on a scale that happens to use the same units. If both are infrared, the gap is a calibration question, and the way to settle it is to put both outdoors in moving air away from traffic and see which one lands nearer the outdoor baseline.

Does a metal oxide sensor ever measure carbon dioxide directly?

No. The mechanism depends on a gas reacting with oxygen adsorbed on a heated oxide surface and releasing trapped electrons, and carbon dioxide is already fully oxidised, so it has almost nothing to donate. The film responds strongly to alcohols, aldehydes, hydrogen and many solvents, and barely at all to the gas the display names. Any carbon dioxide figure from such a device is computed from the response to those other gases.

Why is accuracy quoted as a parts per million figure plus a percentage of reading?

Because two different errors dominate at the two ends of the scale. Near the bottom the limit is the noise floor and the zero offset, which are fixed quantities independent of concentration, so they are expressed in parts per million. Higher up, the absorption curve flattens and any error in the gain multiplies with the reading, so that part is expressed as a percentage. A single number would overstate the accuracy at one end and understate it at the other.

How long should a monitor warm up before its reading means anything?

Longer than most people allow. An infrared source has to reach a stable output and the whole optical block has to reach a stable temperature, since the detector responsivity moves with it. A heated film device has a worse problem, because the film has to burn off whatever adsorbed while it was switched off, and a device that has been in a box can read wildly for a considerable period. Treat the first readings after a cold start as unreliable on any technology.

Do two identical monitors agree with each other?

Closer than two different ones, but not exactly, and the residual gap is informative. Two units of the same model share a sensing principle, a filter band and a firmware baseline routine, so most systematic differences cancel. What remains is unit to unit variation in the optical path and whatever difference their baseline histories have accumulated, which is why a pair that has lived in different rooms can disagree more than a pair taken from the same box.

Does humidity or temperature change a carbon dioxide reading?

On an infrared device the effect is mostly indirect and modest. Water vapour absorbs infrared, but not in the narrow band the filter passes, so a good design rejects it. Temperature matters more, because gas density changes with it and the detector itself is temperature sensitive, which is why benches carry a temperature sensor and apply a correction. On a heated film device humidity is a first order effect, since water competes directly for adsorption sites on the surface.