The Appliance That Ran Clean Until the Door Closed
Why this matters
A combustion appliance is not a self-contained machine. It is one half of a system whose other half is the building, and the building supplies the air. A tech who tests the appliance in the room as they found it, with the door propped and every fan off, has tested a condition the appliance almost never operates in. This case is about a water heater that produced a clean flue reading and put carbon monoxide in a hallway forty minutes later, with nothing wrong inside it.
Stop the exposure before you diagnose anything
The first reading of the visit was ambient carbon monoxide in the hallway outside the mechanical closet at 12 ppm, taken on a personal monitor that was already running when the tech walked in. That is a finding, not background.
The sequence, in order, before any diagnosis: occupants out of the house and into fresh air, including the two people who said they felt fine. Fuel shut off at the appliance shutoff valves for both appliances. Exterior doors and windows opened to ventilate. Only then does anyone start thinking about causes. Anybody reporting headache, nausea, dizziness, or confusion goes to emergency medical care, and the reason is that those symptoms do not reliably scale with exposure and a person who feels better outside can still be carrying a significant blood level.
Do not relight anything to "see if it does it again" while people are inside. The test that reproduces this fault is run with the house empty of anyone who does not have a monitor on them.
The call
Two-year-old complaint file, nothing dramatic in it. The homeowner called because the upstairs hall smelled faintly of exhaust on cold mornings, and their new plug-in detector had chirped once and reset itself. No pilot outages, no soot, no rollout, no yellow flame reported.
The mechanical closet holds a 100,000 Btu/hr fan-assisted furnace and a 50,000 Btu/hr atmospheric water heater with a draft hood, common-vented into a lined chimney. Total connected input 150,000 Btu/hr. The closet is 3 feet by 7 feet with an 8-foot ceiling, 168 cubic feet, and it has a solid door with a single stamped louver panel.
Why the as-found test was not a test
With the closet door open, the water heater fired, draft established within seconds, and the flue read 25 ppm carbon monoxide on an air-free basis. That is a clean appliance. A tech who writes that number on the ticket and leaves has not made a mistake in measurement; they have made a mistake in choosing the condition to measure under.
So the second run was made under worst case: closet door closed, every interior door in the house closed, clothes dryer running, range hood on high, both bath fans on. That is not an artificial stress test. It is a Tuesday morning.
Under that condition the water heater's draft hood spilled within about thirty seconds of the burner lighting, visible on a smoke source held at the relief opening. Flue carbon monoxide climbed and settled near 340 ppm air-free. Ambient carbon monoxide inside the closet reached 45 ppm and the hallway reading rose from 12 to 30 ppm over several minutes.
Worth sitting with: 340 ppm air-free is still under the 400 ppm limit the appliance certification standards allow in the products of combustion. By the appliance's own standard it was passing. It was passing while dumping its products into a hallway, which is why the ambient reading is the one that decides the call and a flue reading alone never can.
Putting a number on the room
Room pressure with respect to outdoors was measured through the closet door, with a digital manometer and a tube run outside.
- Door open, no exhaust running: about -1.5 pascals. Normal winter stack effect on a low floor.
- Door closed, dryer and range hood and both bath fans running: about -9 pascals.
An atmospheric appliance with a draft hood works on buoyancy alone. The draft it can generate is small, and the field protocols that publish worst-case depressurization limits for this appliance type put them in the low single digits of pascals for exactly that reason. Use whatever limit table your jurisdiction or certification program specifies, because they differ. In this case the argument never got that far: the appliance spilled, visibly and repeatably, and observed spillage under worst case ends the discussion regardless of what any pressure limit says.
The important structural point is that the number describing the fault was a room pressure. Nothing measured at the burner would have found it.
What the code method says this closet needed
Run the National Fuel Gas Code's combustion air methods against this room and the shortfall is not marginal.
All indoor air, the standard method: the space needs at least 50 cubic feet per 1,000 Btu/hr of total input. At 150,000 Btu/hr that is 7,500 cubic feet. The closet has 168. It fails by a factor of about 45, which is expected; a closet was never going to pass on its own volume and was always meant to draw from the house.
Openings to an adjacent indoor space: two permanent openings, each sized at 1 square inch per 1,000 Btu/hr of total input, with a minimum of 100 square inches each. At 150,000 Btu/hr that is 150 square inches each, 300 square inches total. The existing single louver panel measured roughly 24 inches by 12 inches gross, and a stamped metal louver's free area runs around a fifth of its gross area, so call it 58 square inches. One opening where two are required, at under a fifth of the required area for a single one.
Openings to outdoors: two permanent openings, each 1 square inch per 4,000 Btu/hr where they communicate directly with the outdoors or through vertical ducts, so 37.5 square inches each here. Through horizontal ducts the requirement doubles to 1 square inch per 2,000 Btu/hr, so 75 square inches each. Outdoor openings are far smaller than indoor ones for the obvious reason: outdoors is not a reservoir that runs out.
So the closet was non-compliant before anyone touched the house. It nonetheless worked for years, which is the part that traps people. A leaky house with modest exhaust equipment supplied enough air by accident that the appliance never needed the openings it did not have.
Three entries in the file, none of them on the appliance
The account history is where this case actually resolves. Three entries, in order:
- Water heater replaced about two years ago, same input rating, same vent.
- Insulation and air sealing work about eight months ago, done by another contractor.
- Range hood replaced about five months ago, with a substantially higher rated airflow.
First cold-weather complaint arrived three weeks ago.
Nothing on the appliance changed. What changed is that the accidental air supply got sealed off and the largest exhaust device in the house got bigger. The appliance had been operating on a margin nobody had ever measured, and two unrelated jobs by two other trades spent that margin. That is the general shape of this fault: the combustion appliance is the thing that shows the symptom, and the cause sits in work orders that were never on the same job.
Containment, and what each option costs
The house needed heat and hot water that night, so containment was a real decision, not a formality.
- Shut both appliances down until the permanent repair. Safest, and it puts a family without heat in cold weather, which creates its own risks and pushes the customer toward a space heater with its own hazards.
- Shut only the water heater and leave the furnace. Rejected. They share a vent, and a common vent is not a place to reason about one appliance in isolation without retesting the whole system under worst case.
- Remove the closet door entirely, install a low-level carbon monoxide monitor in the hallway, and retest under worst case with the door off. Chosen, with two conditions written on the ticket: the door stays off until the repair, and the permanent work is scheduled within days, not weeks. The retest with the door off showed draft holding and no spillage under the same fan load, and ambient carbon monoxide back to zero after ventilation.
The door removal is not a repair, and calling it one is how an interim measure becomes a permanent condition. It buys days by breaking the pressure coupling between the closet and the rest of the house.
The repair, and why the cheaper one was rejected
Two permanent options were quoted.
Outdoor combustion air openings sized by the code method above, ducted to the closet. Fewer labor hours, no appliance changes, and it brings the room into compliance. It was rejected as the sole measure for one reason: the house is on a trajectory. It has already been tightened once, and any future exhaust appliance, a bigger dryer, a second range hood, a whole-house fan, changes the pressure regime again. Passive openings compete with powered exhaust, and they do not always win.
Direct-vent sealed combustion for the water heater, which draws its combustion air from outdoors through its own pipe and vents under its own power, plus combustion air provision for the furnace. More hours, and it removes the atmospheric appliance, which is the only one in the room that depends on buoyancy. After that change, the room's pressure stops being a combustion safety variable.
The customer took the sealed-combustion path for the water heater. The reasoning that sold it was not efficiency. It was that the other option has to be re-verified every time somebody else works on the house, and nobody is going to make that call.
Confirming it under the condition that broke it
The verification run repeated the exact worst case: every interior door closed, dryer running, range hood on high, both bath fans on, from a cold start on both appliances. Spillage checked with a smoke source at one minute. Ambient carbon monoxide monitored in the closet and the hallway throughout. Flue readings taken at steady state on the furnace.
Two additional steps that are easy to skip and worth the fifteen minutes. A low-level carbon monoxide monitor was left in the hallway, because a plug-in alarm built to the residential alarm standard is designed not to sound at levels that still matter over a long exposure. And a follow-up reading was scheduled for the first genuinely cold stretch, because stack effect scales with indoor-to-outdoor temperature difference and a mild day understates the driving pressure.
The last thing on the ticket was the pressure number. Door closed, everything running, minus 9 pascals before, and the reading after. That number, not the flue reading, is the one that tells the next tech what this house does.
References
- NFPA 54 / ANSI Z223.1, National Fuel Gas Code, for combustion air methods, opening sizing, and appliance venting requirements
- ANSI Z21 appliance certification standards, for the 400 ppm air-free carbon monoxide limit in products of combustion
- Field protocols for worst-case depressurization testing and combustion appliance zone testing, as adopted by the certifying program or jurisdiction
- See related: Why Venting Has to Match the Appliance; How Combustion Actually Works; How to Read a Flue Gas Result