Gas Furnace Combustion Analysis and Tuning

Purpose

This standing instruction fixes what "tuning" means on a residential gas furnace. There is exactly one adjustment, manifold pressure, and it is set to the value on the rating plate. Everything else the analyzer tells you is a fault to correct, not a knob to turn.

The number that decides the visit is air-free carbon monoxide, not the raw ppm the probe reads. Raw CO moves with dilution, so a leaking vent connector or a probe in the wrong place can halve it while the appliance produces exactly as much CO as before. Two techs reporting raw CO on one furnace can disagree by a factor of two and both be reading correctly, which is why the shop records the corrected value with the oxygen it was corrected at.

Safety actions that gate the work

• Personal carbon monoxide monitor on before you enter the mechanical space, with the ambient reading recorded with the appliance off. That reading is the only one describing what the occupants were in, and it is destroyed the moment you fire the appliance or open a door.
• If ambient carbon monoxide is elevated with occupants present, the appliance goes off and people leave the space before any analysis begins.
• Prove the control circuit dead before opening a control compartment, proved on a known live source before and after per NFPA 70E-2021, 120.5, with work practices at 29 CFR 1910.333(b)(2), and close the appliance gas cock before any work on the gas train.
• Adjusting a regulator and clocking a meter both put you at a live gas train with the appliance firing. The cap comes off with the appliance running, so no ignition source beyond the appliance itself, ventilation available, and the joint re-checked with a listed leak-detection solution after it goes back on.
• Brushing or vacuuming burners as part of a correction liberates soot into your breathing zone, so HEPA-filtered vacuum at the deposit rather than compressed air, and respiratory protection under a written program per 29 CFR 1910.134 where it clouds.

Scope

Covers combustion analysis and firing-rate verification on residential and light-commercial gas furnaces and gas package units, from the first reading through the corrected appliance. Does not cover the blower-transition testing used to look for exchanger leakage, owned by the heat exchanger inspection methods SOP, or the vent itself, owned by the flue and venting inspection SOP. Does not cover boilers, oil equipment, or field conversion between gas families.

Roles and handoffs

Procedure

1. Zero the analyzer in clean outdoor air and confirm it is in calibration before it goes near the vent. Acceptance: sensor zero completed outdoors away from any vent termination or vehicle, calibration date inside the manufacturer's stated interval, and water trap and filter clear. Wrong looks like zeroing in a basement that already holds combustion products, which biases every reading of the visit low on oxygen. Stop rule: an analyzer past its calibration interval produces numbers that cannot be recorded as findings. Hazard: none at this step, it is done outdoors with the appliance off.

2. Record the appliance's declared numbers before you measure anything. Acceptance: rating plate input in Btu/h, manifold pressure, gas family, and any altitude derate the instructions state for the installed elevation, all written down. Wrong looks like judging an input against a remembered value for a furnace of that size. Stop rule: a plate you cannot read is a manufacturer lookup, not a guess. Hazard: none at this step, it is reading with the appliance off.

3. Run to steady state, then take the whole set together at the manufacturer's stated probe location. Acceptance: oxygen percent, measured CO in ppm, stack temperature and draft, all recorded from the same probe position after readings have stabilized. Wrong looks like sampling in the first minute, when a cold exchanger drives CO high and the number belongs to warmup rather than to operation. Stop rule: readings that will not stabilize point at a cycling fault, so stop and run the short cycling investigation SOP first. Hazard: you are beside a firing appliance with a probe in the vent, so keep the monitor in your breathing zone and hold the probe by its handle.

4. Convert measured CO to air-free and judge against the limits, not against a memory. The correction is measured CO multiplied by 20.9 divided by the quantity 20.9 minus measured oxygen percent, and it is derived on the assumption that the dilution is with ambient air at 20.9 percent oxygen. It corrects for dilution only; it does not forgive a rich flame. Acceptance: air-free CO recorded with the oxygen it was corrected at, judged against the 400 ppm air-free limit the ANSI Z21 series sets for listed vented appliances or the manufacturer's lower limit, and against the shop's own action level of 100 ppm air-free, which is a tunable operational default rather than a listing limit. Wrong looks like recording raw CO alone, which cannot be compared to anything. Stop rule: above the listing limit the appliance goes off now; between the action level and the limit it is investigated before the visit closes. Hazard: still beside a firing appliance, so the monitor stays on.

5. Verify the firing rate by clocking the gas meter rather than inferring it from pressure. With every other gas appliance off, time one full revolution of the smallest dial, compute cubic feet per hour, and multiply by the heating value your gas utility states for that service, which varies by supplier and season. Acceptance: computed input within the tolerance the manufacturer's instructions allow around the plate input, with the timing, the dial size and the heating value all recorded. Wrong looks like using a remembered heating value, which moves the answer by several percent on its own. Stop rule: input outside that tolerance is corrected at the manifold before any combustion number is accepted, and where a meter cannot be clocked, say so and rely on manifold pressure with that limitation noted. Hazard: this is done with the appliance firing and other appliances deliberately off, so relight any standing pilot you extinguished before you leave.

6. Correct causes first, then set manifold pressure to the plate and to nothing else. Acceptance: burners clean and seated, primary air openings clear, vent verified per the venting SOP, and manifold pressure at the rating plate value at steady firing with the regulator cap refitted and the joint leak-checked. Wrong looks like walking the regulator up or down until the CO number looks better, which changes the input the appliance was listed and vented for. Stop rule: if combustion will not come inside limits with manifold at the plate value and the burners clean, the fault is the appliance or the vent, and it goes off pending the exchanger or venting SOP. Hazard: the regulator cap comes off a live gas train while the appliance fires, so no other ignition source in the space and the cap joint gets a listed leak-detection solution afterward.

7. Re-run the full set, re-clock, and write the file. Acceptance: air-free CO with its oxygen, stack temperature, draft, clocked input against plate, manifold at plate, ambient CO with the appliance firing, and the blower door switch confirmed to kill the unit when the door is pulled. Wrong looks like recording only the improved CO and not the input that produced it. Stop rule: any value outside its bound and the appliance does not stay running. Hazard: this is where the appliance goes back into service with the family in the house, so stand to the side of the burner face and keep the monitor on until you leave.

The record this produces

One combustion record per visit, attached to the equipment:

• Analyzer zero location and calibration date
• Rating plate input, manifold pressure, gas family, and any altitude derate applied
• As-found oxygen percent, measured CO, air-free CO, stack temperature and draft, with the probe location
• Clocked input: dial size, seconds per revolution, heating value used and its source, and the computed Btu/h against plate
• What was corrected, and the final manifold pressure against the plate value
• Final oxygen, measured CO, air-free CO, stack temperature and draft, and ambient CO with the appliance firing

The next tech reads air-free CO across visits, the one number that means the same thing every time. The manager reads the clocked-input field, because a book full of appliances running over plate is a habit of tuning by feel, not a run of bad furnaces.

Worked pass: 100,000 Btu/h input furnace, elevated CO on an annual visit

Step 1: analyzer zeroed in the driveway away from the termination, calibration date inside the manufacturer's interval, trap clear.

Step 2: rating plate reads 100,000 Btu/h input, manifold 3.5 in w.c., natural gas, no altitude derate at this elevation.

Step 3: run to steady state, probe at the manufacturer's stated vent location. Oxygen 6.8 percent, measured CO 210 ppm, stack temperature and draft recorded alongside.

Step 4: air-free CO is 210 multiplied by 20.9 divided by 14.1, which is 210 times 1.482, so 311 ppm air-free at 6.8 percent oxygen. That is below the 400 ppm air-free listing limit but well above the shop's 100 ppm action level, so the appliance stays running only while it is being investigated in this visit.

Step 5 FAILS its acceptance and takes its stop rule. Every other gas appliance off, one revolution of the 1 cubic foot dial times at 34 seconds, so 3600 divided by 34 is 105.9 cubic feet per hour. The utility states 1,030 Btu per cubic foot for this service, giving 105.9 times 1,030, or 109,100 Btu/h. Against the plate's 100,000 Btu/h that is 9.1 percent over, outside the tolerance the instructions allow, so the input is corrected before any combustion number is accepted.

Step 6: burners pulled, cleaned with the HEPA vacuum at the deposit, primary air openings cleared. Manifold measured at 4.2 in w.c. against the plate's 3.5 and adjusted down to 3.5. At a fixed orifice, flow goes with the square root of manifold pressure, so 4.2 to 3.5 predicts about a 4.5 percent reduction in rate from that change alone, which does not by itself account for the 9.1 percent; the burner cleaning carries the rest, and the clocked rate governs over the estimate either way. Regulator cap refitted and the joint leak-checked with a listed solution, no indication.

Step 7: re-clocked at 37 seconds per revolution, so 3600 divided by 37 is 97.3 cubic feet per hour, times 1,030 is 100,200 Btu/h, which is 0.2 percent over the plate's 100,000. Re-run set: oxygen 8.1 percent, measured CO 48 ppm, air-free is 48 times 20.9 divided by 12.8, which is 48 times 1.633, so 78 ppm air-free, under the 100 ppm action level. Ambient reads 0 ppm with the appliance firing. Door switch pulled and the unit drops out.

Self-audit, criterion by criterion: step 4 demanded air-free CO recorded with the oxygen it was corrected at, and the run prints 311 ppm air-free at 6.8 percent and 78 ppm air-free at 8.1 percent, each with its correction shown inline; the raw figures 210 and 48 ppm are labelled raw and are never compared to a limit. Step 5 demanded the timing, dial size and heating value with the source, and the run prints 34 seconds, a 1 cubic foot dial, and 1,030 Btu per cubic foot attributed to the utility. Both clocked inputs use the same dial and the same heating value, so before and after are the same quantity. The square-root estimate in step 6 is stated as an estimate that the clocked rate overrides, rather than substituted for the measurement.

References

• The appliance rating plate for input, manifold pressure and gas family, and the manufacturer's instructions for the input tolerance, the probe location and any altitude derate, which are the only correct sources
• The gas utility for the heating value of the gas at that service, which varies by supplier and season and must not be assumed
• ANSI Z21 series listing limits for carbon monoxide in the flue products of listed vented appliances, at 400 ppm air-free, superseded by a lower manufacturer limit where stated
• 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5, for proving dead; 29 CFR 1910.134 for respiratory protection under a written program
• See related: the heat exchanger inspection methods SOP, the flue and venting inspection SOP, and the furnace short cycling investigation SOP