7.5 Combustion Testing and Analysis: CO, Oxygen, Stack Temperature, Draft, and Efficiency

Why Combustion Analysis Is the Only Objective Proof

A gas appliance can look perfect — blue flame, warm supply air, no odor, no complaints — and still be producing dangerous carbon monoxide (CO), running thirty percent overfired, or spilling flue gas into the room every time the dryer runs. Flame color, supply-air temperature, and the customer's impression are opinions. A combustion analysis is a measurement. It is the only way to demonstrate, on paper, that the appliance you installed or serviced is burning fuel completely, venting properly, and delivering its rated input. The TDLR content outline lists Testing — Combustion as its own sub-topic, and Modern Refrigeration and Air Conditioning, the approved reference on the exam's list, covers the same procedure.

Stoichiometric air and why excess air is unavoidable

Stoichiometric combustion is the exact air-to-fuel ratio at which every fuel molecule finds precisely the oxygen it needs — no leftover fuel, no leftover oxygen. For natural gas that ratio is approximately 10:1 by volume, about 10 cubic feet of air per cubic foot of gas, which works out to roughly 17:1 by mass. Complete combustion of methane yields carbon dioxide, water vapor, and heat:

$\mathrm{CH_4} + 2,\mathrm{O_2} \rightarrow \mathrm{CO_2} + 2,\mathrm{H_2O} + \text{heat}$

No field burner achieves that ratio. Fuel and air cannot blend perfectly in the fraction of a second available in the combustion zone, so a burner set at exactly 10:1 would leave unburned fuel and generate CO. Every practical burner therefore runs with excess air — commonly 20 to 50 percent above stoichiometric on residential gas equipment. Excess air is the safety margin that guarantees complete combustion, and its cost is efficiency, because every extra cubic foot of air is heated and thrown up the flue. Tuning combustion is the search for the lowest excess air that still holds CO down. That is exactly why the analyzer reports oxygen and CO side by side: one is the cost, the other is the risk.

What the Analyzer Measures Versus What It Calculates

A modern residential analyzer contains electrochemical sensors and thermocouples. On most instruments it measures only four things directly:

1. Oxygen (O2) in the flue gas, as a percent by volume, from an O2 cell. (A few instruments carry a direct CO2 sensor instead.)
2. Stack temperature, from the thermocouple in the probe tip.
3. Ambient or combustion air temperature, from a second thermocouple at the analyzer body.
4. Carbon monoxide (CO), in parts per million, as it exists in the sample drawn through the probe.

Everything else on the display is calculated: carbon dioxide (CO2) from the measured O2 and a fuel factor, excess air, net stack temperature (stack minus ambient), air-free CO, and combustion efficiency. Combustion efficiency is a steady-state number derived from net stack temperature and O2 or CO2. It is not AFUE and it is not system efficiency — an 80 percent AFUE furnace commonly displays 78 to 82 percent combustion efficiency, while a condensing appliance displays numbers in the 90s only because its stack temperature is so much lower.

Reference table: measurement, instrument, typical range, and interpretation

Say this out loud before you use any of those numbers: they are field benchmarks published for typical residential gas equipment, not code limits. The appliance rating plate, the manufacturer's service instructions, and the authority having jurisdiction govern every one of them. A condensing modulating furnace, a power burner, and a 1960s atmospheric unit will not share a target window. When a manufacturer publishes a setup table, that table wins.

As-Measured CO Versus Air-Free CO

This distinction decides whether a furnace gets red-tagged, so understand it rather than memorizing it. As-measured CO is what the sensor sees in the sample. If a burner is producing CO but the flue also contains a great deal of excess or dilution air, the CO is spread through more total gas volume and the reading drops — even though the appliance is generating exactly the same amount of poison per minute. Air-free CO mathematically removes that dilution and reports the CO concentration as if combustion had occurred with no excess air, giving one number that can be compared appliance to appliance:

$\mathrm{CO_{air\text{-}free}} = \mathrm{CO_{measured}} \times \dfrac{20.9}{20.9 - \mathrm{O_2%}}$

The 20.9 is the oxygen content of ambient air by volume. As the flue O2 rises toward 20.9 percent, the multiplier climbs steeply — which is why a 40 ppm reading taken downstream of a draft hood can represent a badly failing appliance.

Thresholds. ANSI Z21.47, the standard for central furnaces, limits vented equipment to 400 ppm air-free (0.04 percent) in the flue, and ANSI/BPI-1200 uses that same 400 ppm air-free figure as the CO threshold for central furnaces of all categories: above it, the appliance is unsafe and is shut down. Best-practice service standards are far tighter — the National Comfort Institute uses 100 ppm as-measured as its maximum for vented residential equipment, and a healthy, well-tuned furnace normally runs under 50 ppm. Practical framing for the field and for the exam: treat roughly 100 ppm as find-and-fix, and treat 400 ppm air-free as shut it off. Both figures are program and standard specific; your jurisdiction or the manufacturer may impose a lower number, and the lower number always wins.

Where to place the probe

Sample undiluted flue gas, and only undiluted flue gas.

• On a Category I atmospheric appliance fitted with a draft hood or draft diverter, sample between the heat exchanger outlet and the draft hood — in the flue gas outlet ports where gases leave the heat exchanger. Everything downstream of the diverter is mixed with room air and will report artificially low CO and artificially high O2.
• On induced-draft Category I equipment and on Category III and IV appliances with no diverter, sample in the vent connector downstream of the inducer, at the test port the manufacturer designates. Where a general port is used, keep the probe roughly 12 inches or more away from a barometric damper or diverter.
• Let the appliance reach steady state before recording anything. Numbers taken in the first minute of a cold start are not the numbers the appliance runs at.

Draft, Spillage, Manifold Pressure, and Clocking the Meter

Draft is the pressure difference that pulls flue gas out of the appliance and up the vent. It is measured in inches of water column (in. w.c.) with a manometer or the analyzer's draft port. A natural-draft atmospheric appliance should read negative — below room pressure — in the vent connector once it is warm. Field practice places that at roughly -0.02 to -0.04 in. w.c., and standards-based protocols use a band running from about 0 to -0.040 in. w.c. A reading of zero or positive means flue gas is not leaving: a blocked or disconnected vent, an oversized or cold chimney, or a depressurized combustion appliance zone.

Confirm spillage directly. Hold a smoke source at the draft hood relief opening shortly after startup — the smoke should be drawn in, not pushed out — or hold a cool mirror at the relief and look for condensation. Test under worst case: every exhaust fan, clothes dryer, and range hood running, and interior doors positioned to depressurize the space containing the appliance. Under recognized protocols, a vented appliance that continues to spill flue gas more than 60 seconds after startup fails the spillage test.

Manifold gas pressure is read with a manometer at the gas valve's outlet pressure tap while the burner fires. Common values are 3.5 in. w.c. for natural gas and 10 to 11 in. w.c. for LP, but the rating plate governs, and two-stage or modulating valves have separate low-fire and high-fire setpoints. Adjusting manifold pressure by feel, or to chase a temperature rise, is malpractice. Note also that manifold pressure alone does not prove input rate — orifice size, gas heating value, and gas pipe sizing all affect the actual firing rate.

Clocking the meter

$\mathrm{CFH} = \dfrac{3600 \times \text{dial size in ft}^3}{\text{seconds per revolution}} \qquad \text{then} \qquad \mathrm{BTU/h} = \mathrm{CFH} \times \text{heating value in BTU/ft}^3$

Shut off every other gas appliance in the building, fire the unit alone, and time one full revolution of the smallest test dial. Suppose a 1 cubic foot dial takes 36 seconds and the local utility reports 1,030 BTU per cubic foot:

$\mathrm{CFH} = \dfrac{3600 \times 1}{36} = 100 \qquad \mathrm{BTU/h} = 100 \times 1{,}030 = 103{,}000$

Against a 100,000 BTU/h nameplate that is about three percent high — acceptable. Now suppose the same dial turns in 28 seconds: 128.6 CFH, roughly 132,000 BTU/h, about 32 percent overfired. That appliance is cooking its heat exchanger, and the manifold pressure or orifice must be corrected before anything else is touched. Always use the heating value your gas supplier publishes rather than assuming 1,000 BTU per cubic foot; the value varies by region and season. And in Texas, the high-altitude derate only comes into play in the far west and the Panhandle, around El Paso and Amarillo, not on the Gulf Coast.

Read the derate rule carefully, because its two numbers do different jobs. NFPA 54 Section 11.1.2 (reproduced as UMC Appendix B 101.2) says gas input ratings are used as published for elevations up to 2,000 feet, and that above 2,000 feet the input rating is reduced at the rate of 4 percent for each 1,000 feet above sea level. The 2,000-foot figure is only the trigger that decides whether you derate at all; once you are over it, the multiplier is measured from sea level, not from 2,000 feet.

El Paso sits near 3,760 feet, so the derate is $3.76 \times 4% \approx 15%$ — not the 7 percent you get by wrongly counting only the 1,760 feet above the trigger. A 100,000 BTU/h furnace is therefore rated near 85,000 BTU/h in El Paso. Amarillo, near 3,600 feet, derates about 14 percent. Subtracting from the wrong baseline roughly halves the correction and leaves the equipment undersized for the design heat loss.

Reading Failure Patterns in the Numbers

The value of an analyzer is not any single reading; it is the pattern across readings, and how that pattern changes as the appliance runs.

• Stack temperature climbing over minutes while efficiency falls. Heat that should move into the air stream is going up the flue instead: a heat exchanger fouled on the fire side, restricted air flow across the exchanger, a plugged filter, or overfiring. Verify input by clocking before condemning anything.
• CO climbing while O2 falls. Classic under-aerated burner: flame impingement, dirty or misaligned burners, a partially blocked flue, or insufficient combustion air reaching the appliance. Combustion air openings and vent condition come first here.
• CO climbing while O2 also climbs. Air is entering somewhere it should not. Suspect a cracked or perforated heat exchanger, a loose vent connector, or an unsealed inspection panel. The signature of a cracked heat exchanger is readings that change when the circulating blower starts or stops — CO spikes or drops on blower start, O2 shifts, and the flame visibly lifts, rolls, or leans. That blower-linked change is the diagnostic, not the absolute number.
• Draft near zero with acceptable CO measured at the heat exchanger. The combustion is fine; the venting is not. Look at vent sizing and slope, the termination, the connector joints, and building depressurization from exhaust fans, dryers, and range hoods.
• Everything nominal but capacity complaints. Clock the meter. An underfired appliance produces textbook combustion numbers and still fails to heat the house.

Required Actions When Unsafe CO Is Found

When air-free CO exceeds the shutdown threshold for the appliance, or ambient CO in the space reaches the action level of the protocol or jurisdiction you are working under, the sequence is not negotiable:

1. Shut the appliance off at the gas valve and lock or tag it out. Do not leave it in an operable condition.
2. Ventilate the space and move occupants outside if ambient levels warrant it. Elevated ambient CO is a call to the fire department or the gas utility, not a service ticket.
3. Document the readings — as-measured CO, air-free CO, O2, draft, stack temperature, manifold pressure, clocked input, and the operating conditions under which each was taken.
4. Notify the owner in writing, in plain language, describing what failed and what must be repaired.
5. Repair, then retest. The job is not complete until a fresh combustion analysis shows the appliance back inside the manufacturer's limits.

That final step is the exam-level idea worth carrying out of this section: combustion testing is a before-and-after discipline. A reading taken only after a repair proves nothing about what was wrong, and a reading taken only before proves nothing about what you left running in someone's home.