Duct smoke detectors in Ontario: what the code requires and where they actually go

Last updated: September 2026

Quick answer

• A duct-type smoke detector exists to stop the air-handling system pushing smoke around a building. It is not a fire detector for the inside of a duct.
• In Ontario the trigger is Article 3.2.4.12 of the current Building Code: a fire alarm system is installed, and the air-handling system serves more than one storey, more than one suite in a storey, more than one of the fire compartments the article names, or is built without fire dampers where the Code would permit that.
• Part 9 carries its own version for recirculating systems at 9.10.18.5. Part 6 points straight back at 3.2.4.12 through Article 6.9.2.2.
• There is no litre-per-second or cfm capacity threshold in any of those provisions. The capacity numbers circulating online are American.
• On actuation the detector signals the fire alarm system and the air-handling system must stop circulating smoke, by fan shutdown, damper closure, or both. It is an alarm or alert device, not a supervisory one.
• The detector is listed to CAN/ULC-S529 and installed to CAN/ULC-S524. Neither Ontario code names S529 directly: it arrives through S524.
• Run it outside its listed air velocity range and it stops being a listed device, whatever the sticker says.

What a duct detector is actually for

Most people, including a fair number installing them, describe a duct smoke detector as a device that finds a fire in the ductwork. It is not. Ducts do not usually burn, and no code is written around the case where one does. The purpose runs the other direction. An air-handling system is a distribution network, and left running during a fire it takes smoke from the floor of origin and delivers it efficiently to floors and suites that are not on fire yet. The National Research Council puts the intent plainly: the HVAC system is to be shut down during a fire to minimise the circulation of smoke through the building. NRC traces the idea to a 1939 National Board of Fire Underwriters recommendation.

So you are not trying to catch smoke early. You are trying to catch smoke already in the air stream, before the fan hands it to the rest of the building.

When the Ontario Building Code actually requires one

Ontario's Building Code is now O. Reg. 163/24, which adopts the National Building Code of Canada 2020 plus an Ontario amendment document. The old full-length regulation, O. Reg. 332/12, was revoked on 1 January 2025 and the fire alarm articles were renumbered when it went. Most free code-index sites still show the revoked numbering.

Article 3.2.4.12, Prevention of Smoke Circulation, is the governing provision. Where a fire alarm system is installed, the air-handling system has to be designed to prevent the circulation of smoke on a signal from a duct-type smoke detector if the system:

• serves more than one storey,
• serves more than one suite in a storey,
• serves more than one of the fire compartments the article cross references, or
• has no fire dampers where the Code would otherwise permit them to be left out.

In the revoked 332/12 this was Article 3.2.4.13. A specification or website quoting 3.2.4.13 is citing a code that no longer exists, and the renumbering is not uniform across the subsection, so check the article title, not the number.

Part 9 has its own provision, Article 9.10.18.5, Smoke Detectors in Recirculating Air Handling Systems. It applies where a fire alarm system is required, exempts a recirculating system serving not more than one dwelling unit, and bites where the system supplies more than one suite on the same floor or serves more than one storey.

Part 6, where the mechanical designer looks, is one sentence: Article 6.9.2.2 says air-handling systems are to incorporate smoke detectors where and as required by Article 3.2.4.12. There is no separate mechanical requirement to find.

Two sourcing cautions. Free consolidated Ontario text exists only in the 2024 Building Code Compendium, the January 2025 consolidation, and the operative Ontario amendment document is later and unpublished. And the wording above was read from a secondary code index, not the Compendium itself, though its numbering matches the Compendium wherever this site has checked directly. Treat the wording as reliable and the currency as something to confirm with your authority having jurisdiction.

The capacity threshold that is not there

Search this topic and you will land on American pages tying the requirement to the air handler's capacity in cubic feet per minute. That threshold belongs to American standards and model codes. It does not appear in 3.2.4.12, in 9.10.18.5, or in 6.9.2.2. Ontario's test is what the system serves, not how much air it moves, so a small air handler feeding two suites is caught here and would not be caught by a cfm test. For why the American installation standard is not interchangeable with ours, see NFPA 72 vs CAN/ULC-S524.

Where it goes in the duct

Code says a duct detector is required. It does not say where in the duct. That comes from CAN/ULC-S524, the installation standard the Building Code calls up at Article 3.2.4.5.(1), and from the manufacturer's instructions, which a listed device is conditioned on being installed in accordance with.

The physics is simple, and it makes the common error obvious. A sampling-tube duct detector does not sit in the air stream. It sits outside the duct in a housing, and air is driven through that housing by a pressure difference between two tubes. The inlet tube faces into the airflow, so moving air is forced into it, and the exhaust tube faces downstream, into the lower pressure behind it. That differential is the only thing pulling a sample across the sensing chamber.

Reverse the tubes and the housing sees a negative differential. The detector powers up, the LED is green, the panel reports it normal, and no useful quantity of duct air crosses the sensing chamber. It looks installed. It is not.

Mircom's installation instructions for its addressable duct detector, as manufacturer documentation rather than code, say to install the unit pointing towards the direction of air flow, and give a placement rule of three duct diameters before a damper, filter or change of direction and five diameters after one. Those are one manufacturer's figures for one product. Others publish different numbers, and if your specification names a product, its instructions govern.

NRC's testing does not support the largest of those distances. It found no justification for placing duct detectors three to ten duct diameters from bends, while still recommending the mid-length of a straight run. NRC also found that in a horizontal duct carrying warm smoke, stratification puts the highest concentration at the top at low velocity, evening out as velocity rises. A vertical sampling tube mounted at the top of the duct samples the right layer and keeps dust out.

Air velocity, and why the listed range is a hard edge

Every duct detector carries a listed air velocity range. Mircom's addressable unit is listed for 300 to 4,000 ft/min, roughly 1.5 to 20 m/s, and its low-flow series goes down to 100 ft/min. These are not advisory numbers.

Below the range there is not enough pressure differential to move a sample through the housing. Above it, transit time collapses and the sample may be too dilute, or moving too fast, for the chamber to register. Either way the device sits there reporting normal.

The figure that should worry you is NRC's field survey of Canadian HVAC systems, which found velocities from roughly 2 to 40 m/s. The top of that is double the top of the range these detectors are tested over, so there are installed systems whose duct velocity is outside anything the detector was listed for. The fix is not clever: measure velocity with an anemometer at the intended location with the air handler running, before you cut the hole.

Alarm, alert or supervisory: what the panel should do

This produces the most argument on Canadian sites, partly because American practice differs and partly because the answer is split across two ideas. A duct detector required by Article 3.2.4.12 is an initiating device on the building fire alarm system. Writing in IAEI's Canadian column, code consultant Ark Tsisserev is explicit that the detector has to be part of a building fire alarm system, that it transmits a fire alarm signal on a single-stage system or an alert signal on a two-stage system, and that the response includes closing recirculation dampers, signalling the panel and shutting down the fan.

So: not a supervisory signal. Program a code-required duct detector into supervisory and the building will not evacuate when that detector is the first device to see smoke. If the distinction is hazy, start with alarm, supervisory and trouble signals and the supervisory device entry.

The other half of the answer is the fan. The relay, damper actuator or fan control interface that carries out the shutdown is an ancillary device: connected to the fire alarm system, but not part of it, and so not subject to the same electrical supervision as detection and notification circuits. That is why a duct detector can be healthy while the shutdown it commands has quietly stopped working, and why proving the detector alarms proves nothing about the fan. Demonstrating that interconnection is what integrated systems testing to CAN/ULC-S1001 exists for.

A duct detector installed voluntarily, where no provision requires one, is a design decision, and how it is annunciated is for the designer and the AHJ to agree. We could not find a Canadian provision authorising a supervisory connection, and we would not assume one exists. Separately, smoke dampers must close on a signal from a smoke detector near the duct opening: a different detector doing a different job.

Remote test stations, remote indicators and access

Duct detectors end up in mechanical rooms, above ceilings, on roofs and inside penthouses, and a device you cannot see or reach will not get tested properly. The trade answer is a remote station: a plate with an alarm LED, a power LED and a key-operated test and reset switch, mounted where a technician can stand. Mircom lists that accessory for its four-wire duct detectors and a remote LED annunciator for the addressable unit.

Be honest about the limits here. The requirement for remote indication where a detector is concealed from view, and for accessibility generally, sits inside CAN/ULC-S524, which is copyrighted and paywalled. We could not verify the clause, and we are not going to invent one. The code hook is Article 3.2.4.5.(1). If you need to argue it with a contractor, quote the standard, not a website.

Ontario's Fire Code, O. Reg. 213/07, requires a fire alarm system to be inspected and tested in conformance with CAN/ULC-S536 at Article 6.3.2.2.(1). The annual frequency comes from the surrounding provisions, not that sentence. Repairs, replacements and alterations go to S524 through Article 6.3.1.8. If the difference between annual work and a one-time verification is unclear, verification vs annual inspection sets them side by side.

None of it helps if the detector is above a hard ceiling with no hatch. The technician has three options, all bad: skip it and mark the report, cut the ceiling, or bill for a lift and a drywall repair. The Building Code does address duct access at Article 6.8.1.2, which requires access openings so material accumulating in plenums and ducts can be removed, but that is a cleaning provision, not a detector-servicing one. Useful at design review, not a clean answer. The reliable fix is upstream: get the access panel onto the reflected ceiling plan before the ceiling goes in.

What fails at verification

A CAN/ULC-S537 verification is where the shortcuts surface. The recurring duct detector findings are consistent and mostly avoidable:

• Sampling tube reversed, because air flows the other way to the direction assumed on the shop drawing.
• Sampling tube the wrong length, so it samples one corner of the air stream instead of spanning the duct.
• No measured air velocity, so nobody knows whether the device is inside its listed range.
• Alarm proven, shutdown not proven. The device reports at the panel, but the fan interface was never demonstrated.
• Detector unreachable, so testing was done from the panel and no smoke or aerosol entered the sampling tube.
• Programmed as supervisory on a code-required device.
• Tested with the fan off, which proves the sensing chamber works and nothing else.

That last one is the one that passes. A duct detector tested with the air handler off is being tested as an ordinary spot detector in a plastic box.

Dust, and the unwanted alarm problem

A duct detector spends its life in the one place in a building where air moves past it continuously, carrying whatever the filters missed, so the sensing chamber loads up over the years. On a conventional device that drift ends in an alarm nobody caused. On an addressable device with drift compensation it usually produces a trouble or maintenance signal first, which is better, but only if somebody acts on it.

Duct detectors have a strong reputation in the trade as the leading source of unwanted alarms in commercial buildings. We could not find Canadian data ranking causes of unwanted alarms, so we will not state that as a fact. The consequence is verifiable: the City of Toronto charges owners for false fire alarms caused by equipment malfunction or by improper maintenance or installation, while alarms during authorised testing are exempt.

The maintenance points are unglamorous. Clean the sampling tube, not just the chamber, and check the holes are still open and still facing the right way after somebody else has worked in the duct. Record sensitivity readings year over year: a drifting detector shows up in the trend long before it shows up in the lobby at three in the morning.

Which standards apply, and which edition

Two things matter about the product standard, CAN/ULC-S529. Neither Ontario code references it directly. Searches of both volumes of the free Compendium and of the Fire Code's referenced-documents table return nothing for it: it reaches your project through S524, which requires listed devices. The designation has also moved, with development transferred to UL Standards and Engagement and the "S" being dropped.

What the Ontario codes reference is CAN/ULC-S524:2019 and CAN/ULC-S537:2019, even though a 2024 edition of the installation standard exists. That gap is covered in which ULC edition applies, and it is why "I have the newest standard" is not an answer on an Ontario job.

What to confirm before you rely on this

• The current article wording. Free consolidated Ontario text stops at the January 2025 Compendium, and a later Ontario amendment document exists that is not published online.
• Whether your AHJ wants a return-air detector as well as a supply-air one. Article 3.2.4.12 requires the system to be designed to prevent smoke circulation. It does not prescribe supply, return or both.
• The S524 clause on accessibility and remote indication. Paywalled, and we could not read it.
• Whether your detector carries a ULC listing. Some Canadian suppliers' literature cites American product standards, which is a documentation habit rather than a statement about your building.

---

Fire Alarm Academy provides educational content only. It does not confer any certification, licence or credential, and it is not affiliated with, endorsed by, or accredited by the Canadian Fire Alarm Association, the Electrical Contractors Association of Ontario, ULC Standards, UL Standards and Engagement, CSA Group, the Electrical Safety Authority, any authority having jurisdiction, or any manufacturer. Always work from the edition of any standard adopted by your authority having jurisdiction.