Why oxygen is the control variable
Burn fuel with exactly the theoretical amount of air and, in the real world, some of it fails to find oxygen — you get carbon monoxide, soot and wasted fuel. So every combustion system runs with excess air. The question is how much, and flue gas oxygen answers it directly:
The 20.9 is the oxygen content of ambient air. If none were consumed the flue gas would read 20.9%; the shortfall tells you how much air was actually used, and the remainder is what passed through unused.
The cost of too much air
Every kilogram of excess air enters at ambient temperature, gets heated to stack temperature and leaves. That energy is simply gone. The Siegert formula estimates the loss:
with A₁ and B depending on fuel — 0.65 and 0.009 for natural gas. Notice the O₂ term sits in the denominator, so loss climbs steeply as oxygen rises. As a rule of thumb, each additional 1% of flue gas oxygen costs roughly 0.5% of efficiency.
Worked example
Natural gas boiler, 3% O₂, stack at 180 °C, air at 20 °C:
- Excess air = 3 ÷ (20.9 − 3) × 100 = 16.8%
- λ = 20.9 ÷ 17.9 = 1.168
- Stack loss = 160 × (0.65 ÷ 18 + 0.009) = 7.22%
- Combustion efficiency ≈ 92.8% — well tuned
Let the same boiler drift to 8% O₂ and excess air jumps to 62%, stack loss rises to about 9.4%, and roughly two points of efficiency disappear — on a large boiler, a substantial fuel bill for nothing.
Target oxygen by fuel
| Fuel | Typical O₂ target | Excess air |
|---|---|---|
| Natural gas | 2 – 4% | 10 – 24% |
| LPG / propane | 2 – 4% | 10 – 24% |
| Fuel oil | 3 – 5% | 17 – 31% |
| Coal (pulverised) | 3 – 6% | 17 – 40% |
Field notes
- The optimum sits just above the CO knee. Reduce oxygen and efficiency improves — until carbon monoxide starts climbing sharply. That knee is the real target, which is why good combustion control trims on O₂ with a CO limit rather than O₂ alone.
- Air leaks look like excess air. A leaking boiler casing or a failed damper seal raises measured O₂ without the burner receiving more air. Rising O₂ with no control change usually means tramp air, not a tuning problem.
- Zirconia probes read wet or dry — know which. In-situ zirconia cells measure wet flue gas; extractive analysers usually dry the sample first. The two give different numbers on the same stack, and mixing them up shifts every calculation here.
- Stack temperature is the other lever. A rising stack temperature at constant oxygen means fouled heat transfer surfaces — worth trending alongside O₂.
- For safety limits on the gas side, see the gas concentration converter — combustion tuning and flammability monitoring are separate duties.
Frequently asked questions
How do I calculate excess air from oxygen?
Excess air percent = O₂ ÷ (20.9 − O₂) × 100, using dry flue gas oxygen by volume. A reading of 3 percent O₂ corresponds to about 16.8 percent excess air.
What is lambda in combustion?
The ratio of actual air supplied to the stoichiometric requirement. Lambda = 20.9 ÷ (20.9 − O₂). Lambda 1.0 is exact stoichiometric combustion; 1.17 means 17 percent more air than theoretically needed.
What is the optimum oxygen level in flue gas?
Usually 2 to 4 percent for gas firing and 3 to 5 percent for oil. Too little risks incomplete combustion and carbon monoxide; too much carries heat up the stack. The economic optimum sits just above the point where CO starts to rise.
Why does excess air reduce efficiency?
Every extra kilogram of air is heated from ambient to stack temperature and then discharged, carrying that energy away. Roughly speaking, each 1 percent of additional flue gas oxygen costs about half a percent of boiler efficiency.
Simplified Siegert estimation for reference and education; it excludes radiation, blowdown and unburnt losses, so it is not a full boiler efficiency test. Burner adjustment is safety-critical work. See our disclaimer.