How to Take a Combustion Sample That Means Something

Why this matters

The number on the analyser is a property of the sample, and the sample is whatever made it from the flue, down the hose, through the trap and into the cell. The appliance only gets a vote at the first step. Everything after that can add air, drop carbon monoxide on a wet surface, slow the response so you read early, or hand you a perfect measurement of a mixture that never existed inside the heat exchanger.

Which means a bad combustion reading usually is not a wrong number. It is a right number describing something other than what you meant to describe, and it looks exactly as clean as a good one. Techs argue about analyser brands while the real spread between two sets of readings on one appliance comes from where the probe sat and how long the burner had been running.

This procedure walks the path the gas actually takes, from the flue to the display, then back out to the room. Each step owns one way the sample gets spoiled and the direction it pushes the answer.

Before you fire anything

Put a personal carbon monoxide monitor in your breathing zone and turn it on before the burner lights. You are about to stand next to a firing appliance for several minutes with a hole drilled in its vent, and the analyser is watching the flue, not you.

If you smell gas at any point, the job stops. Everyone leaves the building, no switches or lights are touched, no phone is used inside, and the call goes out from outside.

Spillage, a disconnected vent, a blocked flue or soot at the draft hood is a shutdown, not a data point. Shut the appliance off, tag it, and tell the customer plainly. You do not sample an appliance you have already found unsafe to run.

Before a drill bit touches the vent, identify what is behind it. Deburr the hole and close it afterward with a listed high-temperature plug, not tape.

The probe tip and the vent reach flue temperature. Handle the probe by its handle with heat-rated gloves, seal around it with a high-temperature material rather than a bare thumb, and set it down on something non-combustible.

Opening the appliance itself brings in different rules. Stored energy and mechanical isolation are 29 CFR 1910.147; work inside the control compartment is electrical work under 29 CFR 1910.333(b)(2), because 1910.147 excludes exposure to electrical hazards from work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C), with live-dead-live proving per NFPA 70E-2021, 120.5 in the edition your employer's electrical safety program adopts.

Step 1: The point in the flue decides what you are sampling at all

On an appliance with a draft hood or a barometric damper, the vent above that device carries combustion products mixed with room air in a ratio nobody knows and nothing controls. Sample there and every raw value is diluted: oxygen high, carbon dioxide low, carbon monoxide low, stack temperature low.

Sample between the heat exchanger outlet and the draft hood. On a fan-assisted or sealed-combustion appliance, use the port the manufacturer specifies, because those appliances have a defined sampling location and using the wrong one on them is the same error wearing different sheet metal.

Skip this and no correction rescues you. The air-free correction backs out dilution arithmetically, but it cannot separate the burner's own excess air from hood dilution, so the corrected value comes out plausible and the excess air, carbon dioxide, stack temperature and efficiency readings are all describing a mixture rather than the combustion.

Step 2: Depth and position, because flue gas is not uniform

Flue gas stratifies. At the wall it is cooler and generally leaner than in the core, so a probe left at the wall reads oxygen high and stack temperature low. Push the probe to the center and traverse slowly through the cross-section, watching the display. The position with the highest stack temperature and the lowest oxygen is the core; hold there and let it settle.

Skip the traverse and you get a reading that is repeatable and biased, the worst combination, because it agrees with itself every visit and never agrees with anyone else's.

Step 3: Seal the hole you just made

A vent under negative pressure will pull room air in around a loose probe, and that air joins your sample. The signature is the same as real dilution: oxygen up, stack temperature down, carbon monoxide down, computed excess air up, computed carbon dioxide down.

Seal around the probe with high-temperature material every time, including on a quick check. This is the single most common defect in field combustion readings and it is free to fix.

Step 4: The hose, the trap and the filter

Flue gas is wet and, on many appliances, acidic. Between the probe and the cells the instrument runs a water trap and a particulate filter, and both are consumables.

• Drain the trap before the job, not after. Liquid in the line absorbs carbon dioxide and can carry condensate into the cells.
• Check the filter. A loaded filter slows the sample, so the reading creeps toward the true value instead of settling, and a tech who reads at thirty seconds gets a number biased toward the fresh-air baseline: oxygen high, carbon monoxide low.
• Check the hose for cracks and loose fittings. A leak in the hose is room air in the sample, indistinguishable at the display from a leak at the test hole.

The tell for all three is a reading that never stops moving. A healthy sample path settles.

Step 5: The cells and the zero

Run the fresh air calibration in genuinely fresh air, outside or well away from any running equipment, with the probe out of the flue, and let it finish before the appliance fires. Zeroing in a mechanical room where an appliance has been running subtracts the room's real carbon monoxide from every reading you take for the rest of the visit, and it subtracts it silently.

Confirm the oxygen cell settles near 20.9 percent in that air. Electrochemical cells age, and an oxygen cell that will not reach 20.9 in outdoor air is telling you something about the instrument that no amount of care with the probe will fix.

Step 6: Steady state, defined so two techs get the same answer

"Warmed up" is not a specification. Use a settling criterion instead: record the sample when the stack temperature changes by less than 2 F over 60 seconds and the oxygen changes by less than 0.2 percentage points over the same 60 seconds, whichever settles last. Where the manufacturer's instructions state a warm-up time or a different criterion for that appliance, theirs governs and yours goes in the notes.

Two directions to know, because they conflict. Sampled early, the stack temperature is low because the heat exchanger has not come up. Sampled early, carbon monoxide is often high, because burners commonly produce a post-ignition peak that falls as the flame stabilizes. So an early sample does not bias everything the same way, and that is precisely why it cannot be corrected after the fact.

Step 7: The room and the building

The appliance breathes the room, and the room is connected to the building.

• A running exhaust fan, dryer or range hood can depressurize the space, which changes draft and, on a hood-vented appliance, can pull flue gas back into the room. If your personal monitor alarms, that is the finding and the sampling stops.
• A second appliance on a shared vent changes the draft yours sees, and so do closet doors and panels. Sample with the installation in its normal operating configuration and record which that was.

Step 8: Record the conditions with the number

Fuel selection, firing rate or stage, ambient temperature, the room's own carbon monoxide level, sample location relative to the draft hood, whether the hole was sealed, other appliances running, and the settling criterion you met. Without those, the reading answers a question only on the day it was taken.

Worked example: same appliance, two techs, forty minutes apart

Tech A, an early-morning check on a hood-vented natural gas appliance: oxygen 12.1 percent, carbon monoxide 24 ppm, stack temperature 268 F, ambient 66 F. His notes show three departures from the procedure above: the sample was taken about 6 in above the draft hood, the hole was not sealed around the probe, and the reading was recorded 90 seconds after ignition.

Tech B, returning the same morning for an unrelated reason: oxygen 6.9 percent, carbon monoxide 41 ppm, stack temperature 372 F, ambient 66 F. Probe centered between the heat exchanger outlet and the hood, hole sealed, recorded after the stack temperature held inside 2 F over a minute.

Attributing the difference. All three of Tech A's departures push stack temperature the same way, down: hood dilution adds room-temperature air, the unsealed hole adds more of it, and 90 seconds in the exchanger is still cold. That is why the 104 F gap between 268 F and 372 F is the strongest single piece of evidence that the two techs sampled different things rather than watched an appliance change.

Two of the three push oxygen up, the hood and the unsealed hole. The early reading is roughly neutral on oxygen.

On carbon monoxide the three conflict: hood dilution and the unsealed hole both push it down, the early sample pushes it up. Two against one, in opposite directions, which is exactly why the raw carbon monoxide numbers, 24 against 41 ppm, cannot arbitrate anything. That is a 71 percent difference in the raw value and it is mostly bookkeeping.

What the correction recovers, and what it does not. Air-free carbon monoxide for Tech A is 24 times 20.9 over 8.8, about 57 ppm. For Tech B it is 41 times 20.9 over 14.0, about 61 ppm. The two land within about 7 percent of each other. The correction did its job on carbon monoxide, because both samples were diluted by air and that is the one thing the correction is built to remove.

It rescued nothing else. Tech A's oxygen, carbon dioxide, excess air, stack temperature and efficiency figures still describe a mixture. And Tech A's correction multiplied his reading by 2.375 against Tech B's 1.493, so his corrected value carries a proportionally wider band in absolute ppm - the sibling card on the air-free correction owns that arithmetic and states the gate, which is that the raw reading should sit at least five times above the cell's stated floor. At 24 and 41 ppm against a floor on the order of a few ppm, both readings clear it, so the comparison above is licensed. On a 4 ppm reading it would not have been.

The failure mode. Tech A's file now contains a documented 24 ppm on this appliance. Next season, someone compares a properly taken 41 ppm against it, sees carbon monoxide production apparently rising 71 percent, and condemns a heat exchanger that never changed. The defect is not in either reading. It is that the first one was filed without the four lines from step 8 that would have made it uncomparable on sight.

How to verify you got this right

• Take a second sample at a second location. Upstream and downstream of the hood on the same appliance, in the same minute. The gap between them is your dilution, measured rather than assumed.
• Pull the probe out into the room and watch it return to ambient. A path that returns to room oxygen quickly is clear; one that drifts back slowly has a restriction, and your flue reading was taken through that same restriction.
• Re-check the fresh air zero after the job, not just before. Drift during the visit means the numbers you recorded carry that drift.
• Have a second person repeat the sample without seeing your numbers. Two techs following step 1 through step 6 should land inside the instrument's stated accuracy. If they do not, the difference is in the procedure, and the log from step 8 is what lets you find which line.

References

• 29 CFR 1910.147 and 29 CFR 1910.333(b)(2), with the carve-out for electric utilization installations at 1910.147(a)(1)(ii)(C)
• NFPA 70E-2021, 120.5, as adopted through your employer's electrical safety program
• Appliance manufacturer installation and service instructions for the specified sampling port and warm-up requirement
• Analyser manufacturer documentation for calibration, cell accuracy, trap and filter service
• See related: What a Combustion Analyser Is Actually Computing; What an Air-Free Correction Is Doing; Combustion Fundamentals