The Car Warm-Up That Set Off the CO Alarm Upstairs

Why this matters

A carbon monoxide alarm that sounds on winter mornings and never in the afternoon is telling you about a schedule, and the schedule in this house was a car warming up in the attached garage. The fire department was right to ventilate and leave; the HVAC shop was right to test the furnace first. What made the case worth writing up is that the garage air reached a bedroom upstairs through three separate routes, each owned by a different party, and the fix only held when all three were closed.

First: what anyone in this house should do when the alarm sounds

The owner's first call to the HVAC shop came the morning after the second alarm, and the shop's dispatcher started where every call like this has to start:

• When a CO alarm sounds, everyone leaves the house and calls 911 or the fire department from outside. Nobody goes back in to silence it, to open windows, or to find the source.
• If a vehicle is running in the garage, nobody goes into the garage to turn it off if that means walking into the exhaust. The fire department handles it.
• Do not reset the alarm and wait to see if it happens again. An alarm that has sounded twice has already said it will.
• Never warm up or idle a vehicle in an attached garage, with the overhead door open or shut. An open door helps less than it seems; exhaust still builds in the garage and the house still draws on it.

The owner had done the first part right both times. The second time, the fire department measured CO in the upstairs bedroom on arrival, ventilated the house, found the furnace running, and advised the owners to have it checked before using it again.

The house

Two stories over a basement, built in the 1990s, with an attached two-car garage on the first floor and a bedroom built over the garage. The gas furnace is in the basement. The bedroom over the garage is served by its own supply and return ducts, which run from the basement through the garage: up the inside of the garage wall and along the garage ceiling in a sheet metal return trunk and a supply duct, then up through the garage ceiling into the joist space under the bedroom floor.

On cold mornings, one of the owners started the car in the garage with the overhead door shut, let it run about 15 minutes to clear the frost, then opened the door and left. Both alarms sounded in the bedroom over the garage, 20 to 30 minutes after the car was started.

Ruling out the furnace

The shop's tech tested the furnace first, because the fire department had found it running and a cracked heat exchanger or a blocked vent would produce the same alarm. Flue gas CO at normal levels for the appliance, no CO in the supply air at the registers with the furnace firing, vent clear and drafting, heat exchanger inspected without findings. The tech wrote the furnace off as the source, put it back in service for the owners only after watching a full heat cycle with the burner drawing and no CO at the registers, pressed the test button on both bedroom CO alarms to confirm they still sounded, and asked the question that ended the furnace hypothesis: "What happens in this house on the mornings the alarm goes off that doesn't happen in the afternoon?"

The car.

Tracing the routes without running the car

The tech did not reproduce the event. Running a car in the garage to see where the CO goes would put a known lethal concentration next to an occupied house, and it is the one test this call never needs. Air carries the CO; the tech traced the air instead, with a manometer and a smoke pencil, on a cold morning at about 18 F outdoors, with the overhead door shut and nothing running in the garage.

Route 1: the house pulls on the garage (stack effect, blower off). A manometer tube under the house-to-garage door read the first floor about 2.0 pascals below the garage. In winter, warm air rising in the house pushes out through the upper floors and pulls the lower floor negative against everything attached to it, the garage included. Smoke at the door's bottom corner and at a shared-wall outlet drifted into the house. Once inside, that air rises with the stack: up the stair to the second floor, where it leaves through the upper leaks, passing the bedrooms on the way. See related: Stack Effect and the Neutral Pressure Plane.

Route 2: the return duct in the garage pulls garage air in (blower on). With the furnace blower running, smoke held at the return trunk's seams along the garage ceiling was drawn into the duct, and drawn hard at a filter access door on the trunk whose cover had been left loose. Every cfm the return takes from the garage goes through the furnace and out of every supply in the house, the bedroom over the garage included. The IRC does not allow this: ducts in the garage must be at least 26 gauge sheet steel or other approved material and may have no openings into the garage (R302.5.2 in the 2021 edition, as adopted by your jurisdiction).

Route 3: the garage ceiling holes feed the joist space (blower on). Three openings in the garage ceiling drywall: a gap about an inch wide around the supply and return where they rise into the joist space, a 4-inch hole left from a light fixture that had been removed, and a cable hole for the door opener. The return duct and its boot in the joist space under the bedroom had unsealed joints of their own, so with the blower on, the joist space ran slightly negative and smoke at the ceiling holes went up into it. The holes are also breaches of the fire separation: a garage ceiling under a habitable room is required to be 5/8 inch Type X gypsum board under IRC Table R302.6, and penetrations through it are to be protected (R302.5.3, pointing to the fireblocking provision), both as adopted.

The reading that looked better and was worse. With the blower on, the door reading moved from about -2.0 to about -1.1 pascals. The return was pulling the garage itself down, so the pressure across the door shrank. A tech reading only the door would have concluded the blower helped. In fact, with the blower on, the garage air stopped needing the door: the duct was carrying it straight to the supplies.

Why the bedroom over the garage alarmed first

The bedroom sat at the end of both blower-on routes and at the top of the blower-off one. On a cold morning the furnace cycles often: while it ran, the return took garage air through its seams and the ceiling holes and delivered it to the supplies, and the supply nearest the garage end was this bedroom's. Between cycles, the stack drew garage air in at the first floor and carried it upstairs. The alarm's timing, 20 to 30 minutes after the car started, fits a concentration building over that time rather than an instant spike, and the alarm's peak-level memory, which the tech read on the first visit, showed a peak in the low hundreds of ppm on the second morning. That memory is evidence worth asking for on every CO call, because by the time a tech arrives the house has been aired out and the number is gone. Under UL 2034, the standard residential CO alarms are listed to, an alarm responds faster to a higher concentration, sounding within minutes at several hundred ppm and over hours at concentrations below 100.

Three parties, three fixes

The HVAC shop: the duct. Sealed every seam of the return trunk and supply in the garage with mastic, removed the filter access door from the garage trunk and moved filtration to a rack at the furnace in the basement, and sealed the boots and the joints it could reach in the joist space from below. The shop's written recommendation also covered moving the bedroom's ducts out of the garage at the next remodel, since a sealed duct in a garage is a duct that can be unsealed by the next person who works on it.

The drywaller, hired by the owners: the ceiling. Patched the light fixture hole and the opener cable hole with 5/8 inch Type X gypsum to match the existing ceiling, closed the gap around the ducts with a material accepted for fireblocking, and taped and finished the patches so the ceiling is continuous again. The HVAC shop's tech marked every hole with tape before leaving, so the drywaller found all three.

The owners: the car and the door. No more idling in the garage; on cold mornings the car is backed out and the overhead door shut before it warms up. The house-to-garage door got new weatherstripping and a new door sweep, and its closer was adjusted so it latches. CO alarms went in outside the sleeping areas, which the IRC requires in a house with an attached garage that communicates with the dwelling (R315 as adopted), plus one on each level as the shop recommends, and the bedroom over the garage kept its own.

The proof, again without the car

On the next cold morning, same conditions: with the blower on, smoke at the former return seams and at the ceiling patches went nowhere; the door read about -2.0 pascals with the blower on and with it off, because the return no longer pulled on the garage. The pressure across the door did not go away, because the stack is still there every winter; what changed is that the door's weatherstripping now leaves almost no hole for it to work through. Smoke at the door's bottom corner held still.

What would have changed the read. If the alarms had come at night with no car involved, the furnace would have stayed the main suspect and a heat exchanger test under the manufacturer's procedure would have gone deeper. If the bedroom over the garage had no ducts in the garage, Route 2 would not exist and the fix would have been the ceiling and the door alone. See related: The Attached Garage as a Pressure Neighbor.

References

• See related: The Attached Garage as a Pressure Neighbor; The Garage to House Fire Separation; Stack Effect and the Neutral Pressure Plane
• See related: How to Check Whether Garage Air Reaches the House
• IRC R302.5, R302.6 and Table R302.6, and R315, 2021 edition, as adopted by your jurisdiction
• UL 2034, single and multiple station carbon monoxide alarms