How to Read a Flue Gas Result

Why this matters

A combustion analyser prints six or eight numbers on one slip of paper and most of them are calculated from the others. Read them in the wrong order and you will confidently interpret a result that was never valid, which is how an appliance gets signed off on a 40 ppm carbon monoxide reading that was really 155 ppm on the basis that mattered. The numbers are not a list. They are a cascade, and each one decides whether the next one means anything. This article is about that order.

The reading that comes before the appliance fires

Put a carbon monoxide monitor in the occupied space and let it run before you light anything. This is a life-safety reading rather than a diagnostic one, which is why it is not step one of the procedure below, it is the gate on doing the procedure at all. OSHA's general industry permissible exposure limit is 50 ppm as an 8-hour time-weighted average at 29 CFR 1910.1000 Table Z-1 and NIOSH publishes a 200 ppm ceiling. If ambient carbon monoxide in the occupied space is climbing toward those numbers, everyone leaves the building immediately, nobody operates a switch on the way out, the space is ventilated from outside, and nobody re-enters until an instrument reads it clear.

If you smell gas, that is not a measurement problem: everyone out immediately, no switches, no lights, no phone used inside, shut off at the exterior shutoff only if it is on your way out, call from outside.

If you have to make a sampling port, drill only metal vent, with eye protection against chips. Never drill a plastic vent on a condensing appliance; it is a listed sealed pressurised system and the manufacturer provides its own port. Let a probe cool before handling it, or hold it with gloves rated for contact heat, because a probe pulled from a flue is hot enough for a deep contact burn.

Step 1: Zero the analyser in clean air and set the fuel

Warm the analyser and zero it outdoors or in air you know is clean, away from vehicle exhaust and away from the appliance. Then set the fuel selection to the fuel actually in the pipe.

What breaks if you skip it. Zeroing inside a mechanical room sets the instrument's baseline to whatever carbon monoxide is already present, so it subtracts a real hazard from every subsequent reading. And the fuel setting drives every derived number on the slip. Natural gas and propane have different theoretical carbon dioxide ceilings, 11.7% and 13.7% respectively, so an analyser set to the wrong fuel prints a carbon dioxide figure, an excess air figure and an efficiency figure that are all wrong together, in a way that looks perfectly plausible.

Step 2: Put the probe where the appliance's own products are, undiluted

Sample upstream of anything that admits room air: upstream of a draft hood, upstream of a barometric damper, and away from any joint that might be drawing. On a sealed-combustion or condensing appliance, use the manufacturer's designated port.

What breaks if you skip it. Everything below this line. A sample diluted with room air is not a weaker version of the real result, it is a different gas, and no amount of arithmetic downstream reliably rescues it. This is the single most common way a flue gas result gets read wrongly.

Step 3: Let the appliance reach steady state, and watch the whole run

Give the appliance long enough to stabilise, and record readings at more than one point in the cycle: shortly after ignition, and again once temperatures have settled.

What breaks if you skip it. Carbon monoxide during the first minute or two of a cold start is routinely several times the steady-state value on a perfectly sound appliance, so a reading taken too early condemns good equipment. More importantly, the shape of the curve across the run carries information a spot reading does not, and that shape has its own article.

Step 4: Read oxygen first, and use it to judge whether the sample is real

Oxygen is the credibility check on the whole slip. Before you interpret anything, ask whether this oxygen reading is one the appliance in front of you could plausibly produce.

Convert it to the flue gas multiple, which is 20.9 divided by (20.9 minus the measured oxygen percentage). That number is how many times more dry flue gas you are looking at than a perfect burn would make. A well-set appliance typically lands somewhere between about 1.4 and 1.9. A multiple much above 2 on an appliance at steady state means the sample is diluted or the probe or hose is leaking, not that the burner is running at extraordinary excess air.

What breaks if you skip it. You interpret a carbon monoxide number without knowing what it is diluted in, and interpret an efficiency number computed from a temperature that is not the appliance's flue temperature.

Step 5: Convert carbon monoxide to an air-free basis before you judge it

A raw carbon monoxide reading in parts per million is a concentration in whatever mixture the probe happened to sample. To compare it to a limit or to last year's reading, put it on an air-free basis:

Air-free CO = measured CO times the flue gas multiple from step 4.

The listing standards in the ANSI Z21 and CSA series cap carbon monoxide in the products of combustion at 400 ppm air-free for listed gas appliances, which is the outer limit rather than a target. A common shop acceptance is 100 ppm air-free, and that is the number worth setting and holding.

What breaks if you skip it. Two appliances at 60 ppm raw are not comparable at all if one was sampled at 6% oxygen and the other at 12%. And a raw number compared against an air-free limit understates the result every time, always in the unsafe direction.

Step 6: Compute net stack temperature from the air the burner actually breathes

Net stack temperature is measured flue temperature minus the temperature of the air entering the burner. Not minus room temperature by default, and not the raw flue reading.

What breaks if you skip it. On a sealed-combustion or ducted-combustion-air appliance in winter, the burner may be breathing air far colder than the room. Using room temperature there understates the real temperature rise, and every efficiency figure derived from it is optimistic.

Step 7: Read efficiency last, and only if steps 4 through 6 held

The efficiency figure is the most derived number on the slip and the least diagnostic. It is a calculation from oxygen, stack temperature and fuel constants. If any of those was wrong, efficiency is wrong, and it will still print to a tenth of a percent.

What breaks if you skip the ordering. Techs read efficiency first because it is the number customers ask about, then reason backwards to justify it. What excess air actually costs in efficiency terms, and how to decide whether to change it, is a separate article and a separate decision.

The worked case: a printout that looked fine

Handed to you at a service call on a natural-draft gas appliance, sampled at the outlet above the draft hood:

Run step 4 first. Flue gas multiple is 20.9 divided by (20.9 minus 13.5), which is 20.9 divided by 7.4, or 2.824. That is well above 2, so the sample is diluted and the credibility gate has failed. The location explains it: above the draft hood, the appliance's products have already been mixed with room air by design.

Note the carbon dioxide reading is internally consistent with that, at 11.7 divided by 2.824, or 4.1%. Most analysers calculate carbon dioxide from measured oxygen rather than measuring it, so a carbon dioxide figure that disagrees with oxygen is a fuel-setting error, not a second opinion.

Move the probe into the connector between the appliance flue collar and the draft hood, let it settle, and take it again:

Multiple is now 20.9 divided by 12.4, or 1.686, which is inside the plausible band. Air-free carbon monoxide is 92 times 1.686, or 155 ppm. Net stack temperature is 402 minus 68, or 334 degrees F.

Now compare what the two sample points told you. The first slip's air-free conversion would have given 40 times 2.824, or 113 ppm. The correct sample point gives 155 ppm. Those two numbers should agree if the diluent were clean air and both readings were perfect, and they are 37% apart. That gap is the lesson: the air-free conversion multiplies the sensor's own error by exactly the same factor it multiplies the reading, so at 2.824 the raw reading would only have to be understated by 15 ppm to produce the whole discrepancy. Arithmetic does not rescue a diluted sample. Sampling correctly does.

The practical consequence is a different job. At 155 ppm air-free the appliance is under the 400 ppm listing ceiling but well over a 100 ppm shop acceptance, which makes it a fault to investigate rather than a unit to sign off. On the first slip it read 40 ppm and looked clean.

The efficiency figure on the first slip described nothing. The analyser saw a net stack temperature of 247 degrees F, which is 87 degrees F below what the appliance actually produced, and an oxygen reading 5.0 percentage points above what the burner actually made. An 81% figure computed from those two inputs is not an optimistic estimate of the appliance. It is a measurement of the draft hood.

How to confirm the result you are about to write down

Take the whole set twice, at two sample depths in the same port, and confirm oxygen agrees within a few tenths of a percentage point. A sample point sitting in a stratified or eddying part of the flue will move, and if oxygen moves, nothing derived from it is stable enough to record.

Then hold the probe out in clean air until the analyser returns to its zero before you disconnect. A sensor that will not come back down is telling you it is drifting or saturated, and it is worth knowing that before the number goes on a report rather than after. Record the sample location, the point in the cycle, the fuel setting and the building condition alongside every value. A flue gas result without its sample point is not a result, and the next tech has no way to know whether your number and theirs describe the same gas.

References

• ANSI Z21 and CSA appliance standards, which limit carbon monoxide in the products of combustion to 400 ppm air-free for listed gas appliances
• 29 CFR 1910.1000 Table Z-1, general industry permissible exposure limit for carbon monoxide
• Analyser manufacturer documentation for warm-up, zeroing, fuel selection and sensor drift
• Appliance manufacturer installation instructions for designated sampling port locations on sealed-combustion and condensing equipment
• See related: What Excess Air Does to Efficiency; What Incomplete Combustion Produces and Why It Matters