What an Air-Free Correction Is Doing

Why this matters

Two appliances in one building, two carbon monoxide readings, one more than twice the other. The shop condemns the worse one. Corrected for dilution, the two are the same appliance and the difference was excess air.

The air-free correction is the arithmetic that makes carbon monoxide readings comparable across appliances and across visits, and it is genuinely necessary - a raw ppm figure taken at an unknown dilution is not comparable to anything. It also does something less advertised. It is a multiplication, so it multiplies whatever error was in the reading by the same factor it multiplies the reading. A corrected value computed off a heavily diluted sample arrives on the screen with two clean digits and a band that can span a factor of several, and nothing on the display distinguishes it from a corrected value you could hang a decision on.

Before you act on a carbon monoxide number

A personal carbon monoxide monitor in your breathing zone, alarmed and switched on before the appliance fires, is not the analyser. The analyser is sampling the flue.

If your monitor alarms, or you find spillage, or you smell gas, none of this arithmetic applies. Everyone leaves, no switches or lights are touched, no phone is used inside, the call goes from outside, and the appliance gets shut down and tagged. A hazard finding is acted on immediately; it is not corrected, compared or averaged first.

This card is about whether a corrected number can carry a comparison, not about where the action line sits. The threshold at which a reading requires shutdown or correction comes from the fuel gas code your authority having jurisdiction has adopted, which binds you through that adoption, and from the appliance manufacturer's instructions. Where those two differ, the more restrictive one governs the work you do that day.

What the correction is actually doing

Dilution air adds oxygen and inert nitrogen and thins everything else in the same proportion. Dry air is close enough to 20.9 percent oxygen by volume that the instrument can work backward: measure how much oxygen is in the sample, infer how much air was mixed in, and scale the carbon monoxide back up to what it would have been with no excess air present at all.

The arithmetic is one line. Air-free carbon monoxide equals measured carbon monoxide multiplied by 20.9, divided by the difference between 20.9 and the measured oxygen percentage. Three conditions come with it, in the same breath:

• The diluent has to be ambient air. The correction assumes the extra oxygen arrived attached to nitrogen in normal proportion.
• The sample has to be dry. Field analysers report on a dry basis, after the water trap, and the 20.9 figure is a dry-air value.
• All the measured oxygen has to be air that did not participate. The instrument cannot tell whether that air was excess combustion air at the burner, dilution air at a draft hood, room air through an unsealed test hole, or a leak in your sample line. It corrects all four identically.

That last one is the important limit and it is not a flaw in the arithmetic. It means the correction makes carbon monoxide comparable while leaving every other reading on the screen describing whatever mixture you sampled.

What it does not do

• It does not fix the sample point. A reading taken downstream of a draft hood has a valid air-free carbon monoxide value and an invalid oxygen, carbon dioxide, excess air, stack temperature and efficiency. Correcting one number does not launder the set.
• It does not turn a spot reading into a trend. Comparability is not history.
• It does not decide anything. It puts two numbers on one basis. What the number then means belongs to the code and the manufacturer.

The multiplier grows fast

The multiplier is not linear in oxygen, and that is the whole story. Going from 3 to 6 percent oxygen adds 0.23 to the multiplier. Going from 15 to 17 adds 1.82.

What the multiplication does to your uncertainty

Three effects, and they behave differently. Keep them apart.

The carbon monoxide cell's relative error passes through unchanged. Multiplying a reading and its error by the same factor leaves the percentage alone. If the cell is good to 5 percent of reading at that level, the corrected value is good to 5 percent of the corrected value on that account.

The oxygen cell adds a relative error that grows with the multiplier. The sensitivity works out to the oxygen error divided by the difference between 20.9 and the measured oxygen. Take an illustrative oxygen accuracy of 0.2 percentage points, which you should replace with your own instrument's stated figure. At 6.0 percent oxygen that is 0.2 over 14.9, about 1.3 percent. At 15.0 percent oxygen the same 0.2 points is 0.2 over 5.9, about 3.4 percent. Same cell, same error, nearly triple the effect, purely because the denominator shrank.

In absolute parts per million, the band is multiplied by the same factor the reading is. A reading of 20 ppm with a band of plus or minus 3 ppm becomes, at a multiplier of 3.5, a corrected 70 ppm with a band of plus or minus about 11 ppm. The display will show a whole number and hide all of it.

Which of the two terms dominates depends on where you are sitting on the carbon monoxide cell's accuracy curve, and that curve is in your instrument's documentation, not in a rule of thumb. Most field cells are specified as a percentage of reading or a fixed floor, whichever is greater, so at low readings the fixed floor dominates and the relative error is large before any correction is applied. For the examples below, treat that floor as plus or minus 3 ppm and the relative accuracy above it as 5 percent, both illustrative and both replaceable from your own spec sheet.

The gate before you compare two corrected values

Use an air-free value for a comparison only when both of these hold, per sample, at the sample point: the raw reading is at least five times the cell's stated floor, AND the resulting absolute band is narrower than the difference you are trying to detect. Both, not either. The first test protects you from correcting noise; the second protects you from correcting a real reading into a band too wide to decide with.

If the first test fails, the fix is not a bigger multiplier. It is a less diluted sample: move upstream of the draft hood, seal the test hole, get the raw reading up where the cell can resolve it.

Two appliances, one gate, two answers

Case one: the comparison is licensed.

Two furnaces in the same building, both sampled between the heat exchanger and the draft hood, both at steady state, both holes sealed. Furnace 1 reads 42 ppm carbon monoxide at 8.4 percent oxygen. Furnace 2 reads 19 ppm at 14.6 percent oxygen.

Raw, Furnace 1 looks 121 percent worse than Furnace 2. Nobody would leave that alone.

Gate test one: 42 ppm is 14 times the 3 ppm floor; 19 ppm is about 6 times it. Both clear the five-times requirement.

Correct. Furnace 1's multiplier is 20.9 over 12.5, or 1.67, giving about 70 ppm air-free. Furnace 2's multiplier is 20.9 over 6.3, or 3.32, giving about 63 ppm air-free. The corrected difference is about 11 percent, down from 121 percent.

Gate test two, the bands. Furnace 1 carries 5 percent from the carbon monoxide cell plus 0.2 over 12.5, or 1.6 percent, from the oxygen cell: about 6.6 percent, which is roughly plus or minus 5 ppm on 70, giving 65 to 75 ppm. Furnace 2 carries 5 percent plus 0.2 over 6.3, or 3.2 percent: about 8.2 percent, roughly plus or minus 5 ppm on 63, giving 58 to 68 ppm. The two intervals overlap between 65 and 68 ppm.

The answer. On this evidence the two appliances are producing the same carbon monoxide. The 121 percent raw difference was a difference in excess air, not in combustion quality, and Furnace 2's higher oxygen is now the interesting finding - a separate question about setup and venting, not a heat exchanger question. Note also which appliance carries the wider band: Furnace 2, the one that needed the bigger multiplier.

Case two: the comparison is not licensed.

A third appliance in the same building has no accessible port upstream of its draft hood, so the sample is taken above the hood. It reads 4 ppm carbon monoxide at 16.8 percent oxygen.

The instrument displays an air-free value: 4 times 20.9 over 4.1, or about 20 ppm. Two digits, looks like a measurement, sits neatly alongside the 70 and the 63 from case one.

Gate test one fails. 4 ppm is about 1.3 times the 3 ppm floor, nowhere near five. Push the raw band through the same multiplier: a raw 1 to 7 ppm becomes a corrected 5 to 36 ppm. The corrected value spans a factor of seven.

The answer. That 20 ppm cannot be compared to anything, including itself next season. The correction did not create the problem, it exposed it: at 16.8 percent oxygen you are sampling mostly air, and the carbon monoxide the appliance actually produced is buried under the cell's resolution. The move is not to record 20 ppm with an asterisk. It is to find or make a sampling point upstream of the hood, get the oxygen down where the multiplier is under 2, and take a reading the cell can resolve. If the appliance genuinely offers no such point, the honest note on the ticket is that a comparable carbon monoxide value could not be obtained, which is a real finding a future tech can act on.

How to verify you got this right

• Check the oxygen cell against outdoor air. It should settle near 20.9 percent. That single check validates the denominator behind every corrected value you will produce that day, and it costs seconds.
• Take a second sample at a less diluted point and correct both. If the two corrected values agree inside their combined bands, the correction is doing its job. If they do not, the dilution was not all air, or one sample had a defect the correction cannot see.
• Compute the multiplier before you trust the corrected number. If it is above about 2.5, ask why the oxygen is that high before you ask what the carbon monoxide means.
• Record raw carbon monoxide and oxygen, not just the corrected value. Both survive a firmware change, an instrument change and a disagreement with another shop. A lone corrected number cannot be taken apart later by anyone.

References

• The fuel gas code adopted by your authority having jurisdiction, in the edition adopted, together with the appliance manufacturer's installation and service instructions, for action thresholds
• Analyser manufacturer documentation for carbon monoxide and oxygen cell accuracy, resolution and stated floor
• See related: What a Combustion Analyser Is Actually Computing; How to Take a Combustion Sample That Means Something; What Incomplete Combustion Produces and Why It Matters