NDIR gas analyzer

 How an NDIR Gas Analyzer Actually Works

Walk past a CEMS cabinet or a CO/CO₂ analyzer skid in a cement plant and you'll usually find an NDIR analyzer doing the heavy lifting for gases like CO, CO₂, SO₂, or NOx. Unlike the zirconia cell, which measures oxygen through an electrochemical reaction, NDIR works on something completely different — light absorption. It's a beautifully simple idea once you see how the pieces fit together.

The Core Idea: Gases Absorb Specific Colors of Infrared Light

NDIR stands for Non-Dispersive Infrared. Every gas molecule made of two or more different atoms — CO, CO₂, SO₂, NO, CH₄, and so on — absorbs infrared light at very specific wavelengths, determined by how its molecular bonds stretch and bend. CO₂, for instance, strongly absorbs infrared light around 4.26 micrometers. CO absorbs strongly around 4.6 micrometers. Each gas has its own fingerprint.

Symmetric diatomic gases like O₂ and N₂ don't have this property — their bonds don't create a shifting dipole moment when they vibrate, so they're essentially invisible to infrared absorption. This is exactly why NDIR is used for CO, CO₂, and SO₂, but never for oxygen — that measurement is left to zirconia or paramagnetic methods.

Building the Measurement Path

A basic NDIR analyzer has four main components sitting in a line:

An infrared source — typically a small heated element that glows and radiates a broad spectrum of infrared light, not unlike a miniature light bulb tuned for IR output.

A sample cell — a tube through which the process gas flows (or, in some modern designs, a folded optical path to shrink the physical size while keeping a long light path).

An optical filter — a narrow bandpass filter that only lets through the specific wavelength band the target gas absorbs. This is the "non-dispersive" part — rather than splitting light into a full spectrum like a spectrometer would, the analyzer simply filters for the one band of interest.

A detector — usually a pyroelectric or photoacoustic sensor that converts the remaining IR intensity into an electrical signal.

The Actual Measurement: Beer-Lambert Law

As infrared light passes through the sample cell, gas molecules of the target species absorb a portion of the light at their characteristic wavelength. The more of that gas present, the more light gets absorbed, and the less reaches the detector. This relationship follows the Beer-Lambert Law:

I = I₀ × e^(-α·c·L)

Where:

I₀ = the original light intensity entering the cell

I = the light intensity reaching the detector

α = the absorption coefficient specific to that gas and wavelength

c = gas concentration

L = the path length of the sample cell

In plain terms: more gas, less light gets through. The analyzer measures how much the light intensity dropped compared to a clean reference, and works backward through this equation to calculate concentration.

Why a Reference Channel Matters

A single detector measuring only the absorption wavelength has a problem — dust on the optics, source aging, or window fogging would all look identical to "more gas present," causing false readings. To solve this, most NDIR analyzers use a dual-beam or dual-wavelength design: one detector channel is filtered to the gas absorption band, and a second reference channel is filtered to a nearby wavelength the gas doesn't absorb at all. Since both channels see the same source, dust, and window condition, dividing the measurement signal by the reference signal cancels out those common errors and isolates the true gas absorption. This ratio-based approach is a big part of why NDIR analyzers stay stable over long runs in dirty industrial environments.

Some designs instead use a rotating chopper wheel with alternating filters, comparing measurement and reference readings from the same detector in rapid succession — same underlying logic, different mechanical execution.

Cross-Interference: The Practical Headache

Because absorption bands can be close together, gases sometimes partially interfere with each other's measurement — CO₂ and CO bands sit near enough that a high CO₂ background can nudge a CO reading. Good analyzers correct for this in firmware using cross-interference compensation tables, but it's worth knowing about when you're troubleshooting a reading that seems to track a different gas's concentration more than it should.

Where This Fits in Your Plant

For kiln and cooler exhaust monitoring, NDIR analyzers are typically used for CO and CO₂ — CO especially matters for combustion efficiency and as an early warning for reducing conditions or incomplete combustion in the kiln. Because these are usually extractive systems (sample drawn through a probe, filter, and conditioning system to the analyzer cabinet), the sample conditioning quality has a huge effect on NDIR performance:

Moisture in the sample can cause cross-interference and even damage optics if it condenses in the cell, so sample coolers and desiccant dryers upstream are essential.

Dust carryover fouls the sample cell windows over time, gradually reducing the reference-corrected signal quality even though the correction helps mask it for a while.

Sample flow rate needs to stay consistent — too slow and response time lags behind real process changes; too fast and residence time in the cell drops, affecting reading stability on some designs.

Where Things Usually Go Wrong in the Field

From a maintenance standpoint, most NDIR issues trace back to:

Sample conditioning failure — a saturated desiccant cartridge or a failing sample cooler lets moisture through, and readings start drifting or bouncing.

Optical window fouling — dust or condensate builds up on the cell windows faster than the reference channel can fully compensate for, especially past its calibration interval.

IR source aging — the source's output intensity slowly degrades over years of operation; most units flag this as a "low signal" or "source fault" diagnostic before it affects accuracy meaningfully.

Zero and span drift — like most analyzers, periodic zero gas (usually N₂) and span gas calibration is what keeps the Beer-Lambert calculation anchored to reality; skipping calibration intervals is the single most common cause of "the analyzer is lying to us" complaints.

Thinking about it as light being selectively absorbed by a specific gas fingerprint — rather than a mysterious black-box reading — makes it much easier to reason through a bad reading: is it really more gas, or is something in the optical path (dust, moisture, a tired source) making it look that way?