A combustion analyser gives you a handful of numbers in thirty seconds. Knowing what they mean turns a service visit into a tuning exercise, and the difference between a well-set burner and a badly-set one is several percent of the fuel bill, permanently.
Everything hangs on the air ratio
Combustion needs a specific amount of air — 17.2 kg per kg of natural gas, 15.1 for petrol. The ratio of what you actually supply to that ideal is λ (lambda), and it is the one number the whole exercise is about. Below 1 you make carbon monoxide and soot; well above it you heat the atmosphere. The flue gas tells you which side you are on.
Reading O₂ and CO₂
They move in opposite directions as excess air rises: extra air dilutes the CO₂ and leaves oxygen behind. So either identifies λ, and analysers usually show both. O₂ is the more reliable in practice because it barely depends on fuel composition, whereas maximum CO₂ varies by fuel — 11.7% for natural gas, around 15.5% for oil. That is why the instrument has to be told which fuel it is looking at. A field rule worth memorising: 3% O₂ is about 15% excess air on gas.
Why the readings are "dry"
Burning a hydrocarbon makes a lot of water — for natural gas, nearly 19% of the flue gas by volume. The analyser chills the sample and knocks that water out before it reaches the cell, so every reading is on a dry basis. Comparing a dry analyser reading against a wet-basis calculation is a quiet and common error.
Combustion air is free. A boiler is running at 3% flue O₂; someone opens the air damper to 8% "for safety". What happens to efficiency?
The air is free; heating it is not. Every extra kilogram of air drawn in is heated from ambient to flue temperature and then thrown away up the stack, and air is about 79% nitrogen which does nothing at all except carry heat out of the building. Going from 3% to 8% O₂ roughly doubles the excess air and costs a few points of efficiency. The floor is set by the opposite failure: too little air gives incomplete combustion, CO and soot, which is dangerous rather than merely wasteful — so the target is the lowest O₂ that stays safely clear of CO, typically 2–4% on gas.
Try it — the efficiency curve and its two cliffs
Where the losses actually are
The dominant one is dry flue gas loss: the energy carried up the stack by hot combustion products. It rises with stack temperature and with excess air, which is why both appear in the Siegert formula every analyser implements. Then radiation and convection from the boiler shell — usually taken as 1–3% — and, for solid and heavy liquid fuels, unburnt carbon.
Net and gross: the same boiler, two numbers
This confuses more people than anything else in the subject. Net (LHV) efficiency ignores the latent heat of the water formed by combustion; gross (HHV) counts it. Europe quotes net, the US quotes gross, and for natural gas they differ by about 10%. The same appliance is honestly "92% efficient" in one market and "83%" in the other. It also explains how a condensing boiler can exceed 100% — it recovers heat the net basis never counted in the first place.
What good looks like
For a well-set gas boiler: 2–4% O₂, stack temperature not much above 180 °C on a non-condensing unit or under 60 °C on a condensing one, CO in single-digit ppm, and net efficiency somewhere in the high 80s to low 90s. If the stack is hotter than it should be with the burner correctly set, the heat exchanger is fouled — clean it before touching the air. Work the numbers through on the Combustion & Flue Gas tool.