Combustion Tuning with Oxygen Analysers: Finding the CO Knee

8 min read · updated 2026-09-12

Every percent of unnecessary oxygen in your flue gas costs about half a percent of efficiency. Every percent you remove past a certain point starts producing carbon monoxide. The whole craft of combustion tuning is locating that point and sitting just above it.

Why oxygen is the measurement

In theory, fuel plus exactly the stoichiometric air gives complete combustion and no leftover oxygen. In practice, mixing is imperfect — some fuel never finds an oxygen molecule, and you get carbon monoxide, soot and unburnt fuel going up the stack. So every real burner runs with excess air, and the leftover oxygen in the flue gas tells you exactly how much:

Excess air % = O₂ / (20.9 − O₂) × 100 λ = 20.9 / (20.9 − O₂)

The 20.9 is ambient air's oxygen content. Three percent O₂ in the flue means 16.8% excess air, or lambda 1.168 — the excess air calculator works through it along with stack loss and efficiency.

The cost of carrying air

Excess air is not free. Every kilogram of it enters at ambient temperature, gets heated to stack temperature, and leaves carrying that energy with it. The Siegert relation estimates the loss:

Stack loss % = (T_stack − T_air) × ( A₁ / (21 − O₂) + B )

The oxygen term sits in the denominator, so the loss climbs steeply as O₂ rises. A natural gas boiler at 3% O₂ with a 180 °C stack loses about 7.2% — roughly 92.8% combustion efficiency. Let it drift to 8% O₂ and loss rises to about 9.4%, with excess air at 62%. On a large boiler those two points of efficiency are a substantial annual fuel bill for no benefit whatsoever.

The CO knee

So reduce the air. Efficiency improves steadily as oxygen falls — until it doesn't. Somewhere below 2–3% O₂ (the exact point depends on burner design, firing rate and fuel) carbon monoxide begins to climb, and it does not climb gently. It goes from tens of ppm to hundreds within a fraction of a percent of oxygen.

That sharp rise is the CO knee, and it is the real target. Sitting just above it captures nearly all the available efficiency while keeping combustion complete. Sitting below it wastes more fuel as unburnt CO than the reduced excess air saved, and produces a genuine safety hazard.

The knee moves. It shifts with firing rate, ambient temperature, fuel composition and burner condition, which is why good combustion control trims on oxygen with a CO limit rather than holding a fixed O₂ setpoint forever. A single tuning exercise at one firing rate is not a combustion control system.

Wet or dry — know which you're reading

This catches people constantly. In-situ zirconia probes measure wet flue gas, including the water vapour from combustion. Extractive analysers usually dry the sample first and report on a dry basis. The same stack gives different numbers depending on which you have.

The difference is not small — water vapour can be 10–18% of flue gas by volume when firing natural gas, so a wet reading is meaningfully lower than the dry equivalent. Every excess-air formula, including the one above, assumes dry basis. Mixing the two shifts every downstream number and produces tuning targets that are quietly wrong.

When rising oxygen isn't a burner problem

If O₂ climbs with no control change, check for air getting in where it was not invited before touching the burner:

Tramp air through a leaking casing, a failed damper seal, or an inspection door left ajar. It raises measured O₂ without the burner receiving more combustion air — so trimming the burner in response makes the actual combustion worse.
A drifting analyser. Zirconia cells age and their reference degrades. Verify against a portable before believing a trend.
Probe location. A sensor near a leak or in a stratified duct reads a local pocket rather than the bulk gas.

Meanwhile, a rising stack temperature at constant oxygen points somewhere else entirely — usually fouled heat transfer surfaces. The two readings together tell you more than either alone, which is why both belong on the same trend.

A safety note worth stating plainly

Burner adjustment is safety-critical work. Reducing combustion air on a live burner without CO monitoring risks incomplete combustion, and in the wrong circumstances, accumulation of unburnt fuel. This is permit work performed by competent people with the right instruments — the calculations here tell you what the numbers mean, not that you should go and change them.

Calculators used in this guide: excess air & efficiency · gas concentration converter · thermocouple conversion