What Incomplete Combustion Produces and Why It Matters

Why this matters

Incomplete combustion is not a state an appliance is either in or out of. It is a continuum, and its products arrive in a fixed order with carbon monoxide first, long before anything a person can see, smell or feel. That ordering is the entire reason a combustion analyser exists: it gives you the first product rather than the last one. It is also why the two habits that feel most reasonable in the field, taking one reading and comparing it to a limit, and trusting your senses when the reading looks fine, both fail in the same direction.

What to do before, and instead of, investigating

Carbon monoxide is odourless and colourless at every concentration that will hurt you, so put a monitor in the occupied space and have it running before the appliance fires. If ambient carbon monoxide climbs toward OSHA's general industry permissible exposure limit of 50 ppm as an 8-hour time-weighted average at 29 CFR 1910.1000 Table Z-1, or toward the 200 ppm NIOSH ceiling, or if anyone reports headache, nausea or confusion, 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 instead, that is the other instruction and it is not a diagnosis: 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.

Two hazards specific to this subject. If you smell a sharp, acrid odour at an operating appliance and your eyes sting, stop sampling: you are already breathing aldehydes, formaldehyde among them, which OSHA regulates as a carcinogen under its own standard at 29 CFR 1910.1048. Shut the appliance off at its gas shutoff and ventilate the space from outside before you do anything else. And if you find soot and intend to remove it, do not dry-brush or blow it out. Soot is a respirable particulate and the route is inhalation, so it needs a HEPA-filtered vacuum and respiratory protection selected and fit-tested under 29 CFR 1910.134, not a glove and a rag. Let the appliance cool before you reach into it, or handle it with gloves rated for contact heat.

Never check a draft hood for spillage by putting your face or hand in the relief opening.

The reaction stops in stages, and the order does not vary

The chain from hydrocarbon to carbon dioxide and water runs through intermediates. Interrupt it and it stops wherever it got to, and it stops at the same places in the same order every time.

Carbon dioxide is not on the fault list. It is a normal, required product, and an analyser reports it mainly as a measure of how diluted the sample is. Confusing carbon dioxide with carbon monoxide on a printout is a mistake that gets made often enough to be worth saying once.

Oxides of nitrogen belong on a different axis entirely. Thermal formation of nitrogen oxides rises with peak flame temperature and available oxygen, which means it does not track carbon monoxide and frequently moves in the opposite direction as an appliance is adjusted. How far it moves is specific to the burner and its combustion chamber, so treat it as a separate measurement with a separate limit rather than as another symptom of the same fault.

What a carbon monoxide reading is not

The useful discipline here is subtraction, because most misreadings come from asking the number a question it does not answer.

• It is not a gas leak detector. Carbon monoxide in the flue is about how well the fuel burned. Unburned fuel escaping a fitting is a different fault, found with a different instrument, and one does not screen for the other.
• It is not proportional to what you can see. Serious carbon monoxide is produced at levels far below anything that makes visible soot or a detectable odour. By the time an appliance is depositing carbon, its carbon monoxide left the interesting range a long way back. The visible symptom is a late indicator, not a confirming one.
• It is not a spot value. The same appliance can read tens of ppm two minutes into a run and hundreds fifteen minutes in. A single sample records a moment, and which moment you chose decides the answer.
• It is not a room reading. Ambient carbon monoxide in the occupied space and air-free carbon monoxide in the flue are two different measurements, with different thresholds and different actions. A clean room reading tells you the vent is working today. It says nothing about how the appliance is burning, and a sound vent will carry a badly burning appliance's products outdoors for years, right up until the day it does not.
• It is not comparable between appliances until it is converted. A raw ppm figure depends on how diluted the sample was. Putting it on an air-free basis is what makes two readings comparable, and that conversion, along with the sampling discipline it depends on, is covered in the flue gas reading article.
• It is not something a person can sense. There is no odour, no colour and no taste. The early symptoms are indistinguishable from a headache. Anyone who tells you they can smell carbon monoxide is smelling aldehydes, which means they are already deep into the table above.

The rule worth holding, stated so it can be applied

For one appliance, on one complete cold-start run, sample air-free carbon monoxide at least three times: about 2 minutes in, about halfway through a typical run, and at steady state. Treat the appliance as a fault if either of these is true:

• Absolute: air-free carbon monoxide at steady state exceeds your acceptance threshold. The ANSI Z21 and CSA listing standards cap products of combustion at 400 ppm air-free for listed gas appliances, which is the outer limit and not a target; a common shop acceptance is 100 ppm air-free.
• Trend: the steady-state air-free value is more than 2.0 times the 2-minute value, regardless of whether any single reading crossed the threshold.

The two arms fire independently, which is the point of the word "either." An appliance reading 30 ppm at 2 minutes and 85 ppm at steady state passes a 100 ppm acceptance at every single reading, and its ratio of 2.8 makes it a fault anyway.

Two runs, one instrument, two different faults

The first appliance. Air-free carbon monoxide, sampled across an 18-minute run from a cold start:

The value rose at every sample through 12 minutes, then flattened, gaining only 15 ppm across the final six minutes. End to end that is 355 divided by 22, or roughly a 16-fold rise. Against the 400 ppm listing ceiling, this appliance never crossed the line at any point in the run, and a tech checking only against 400 signs it off. Against a 100 ppm acceptance it fails on the absolute arm, and against the trend arm it fails by 355 divided by 41, or 8.7 times the 2-minute value.

What the shape says is more specific than either number. Something is opening up as the appliance heats: the fault is absent cold, grows with temperature, and stabilises when temperatures stabilise. That is the signature of thermal distortion, and it points at the heat exchanger, at a burner rack shifting in its supports, or at a vent whose behaviour changes as it warms. A tech who sampled once at 2 minutes wrote down 41 ppm and left.

The second appliance. 120 ppm at 30 seconds, 119 ppm at 2 minutes, 118 ppm at 5 minutes, 121 ppm at steady state. Across the whole run it moved 3 ppm, which is about 2.5% of its own value. It fails the absolute arm against a 100 ppm acceptance and passes the trend arm at 121 divided by 119, or 1.02.

Flat and elevated is a fixed geometry problem: it is the same cold as it is hot, so temperature is not the variable. Look at a blocked or partially obstructed burner port, a wrong or damaged orifice, a mixture that is set wrong, a primary air shutter that has moved. Nothing here is going to be found by cycling the appliance for twenty minutes, and nothing about the first appliance's fault would have been found by sampling this one's way.

Same instrument, same conversion, opposite investigations, and the only thing that separated them was taking more than one reading.

What changes the reading

Cold start behaviour is normal, to a point. Every sound appliance shows elevated carbon monoxide during ignition and the first seconds afterwards, while surfaces are cold and quenching the flame. That is why the trend arm is anchored at 2 minutes rather than at the first breath. Anchoring it at 30 seconds would flag healthy equipment.

Modulating and staged appliances. The trend arm assumes one firing rate across the run. On staged or modulating equipment, run and sample each stage separately, because a unit that is clean at high fire can be poor at low fire where flame velocity and mixing are weakest.

Recent service. An appliance sampled immediately after a burner cleaning can read high from residue burning off. Give it a full run before you record a number, and note on the ticket that it was sampled post-service.

Ambient contamination in the sample. If the analyser is drawing air from a space that already contains carbon monoxide from a vehicle, a space heater or an adjacent appliance, the reading includes it. That is why the zeroing step happens in clean outdoor air.

How to verify you got this right

Repeat the whole run rather than the reading. Let the appliance cool completely, start it cold again, and take the same three or more samples at the same points in the cycle. A genuine thermal fault reproduces its curve, and a sampling artefact does not. If the two runs disagree in shape rather than in magnitude, the disagreement is in your sampling, and the sample point and probe seal are the first things to check.

Then record the whole series, not the worst number. A ticket that says "355 ppm air-free" and a ticket that says "22 ppm at 30 seconds rising to 355 ppm at 18 minutes, flattening after 12" describe the same appliance and hand the next tech completely different amounts of information. The second one has already narrowed the search.

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
• 29 CFR 1910.1048, the OSHA formaldehyde standard
• 29 CFR 1910.134, respiratory protection program requirements including selection and fit testing
• See related: How to Read a Flue Gas Result; How to Read What a Flame Is Telling You