ok so duct smoke detectors.
Not the hardest thing on a job, usually. But there's enough going on between the IMC and NFPA 72 that stuff gets missed on submittals, and the AHJ catches it every time. Let me walk through the parts that actually trip people up.
The 2,000 CFM threshold
NFPA 72 doesn't tell you when to put in a duct detector. That's the mechanical code's job. The International Mechanical Code, section 606.2, requires duct smoke detectors on air-handling systems with a capacity greater than 2,000 CFM. Most states that haven't written their own mechanical code adopt the IMC, so 2,000 CFM is your working number in most places. Check local amendments though. Some jurisdictions have moved it.
The IMC wants a detector on the supply side for systems over 2,000 CFM. Return-air detectors come in at 15,000 CFM or when the return is drawing from multiple stories. The supply detector is usually the one that generates the most submittal questions.
Rooftop unit example with real airflow numbers
Here's a breakdown from a recent commercial office job (anonymized). Five separate RTUs on one roof, mixed tonnage:
| Unit | Nominal Tons | Approximate CFM | Duct Detector Required? |
|---|---|---|---|
| RTU-1 | 2 ton | ~800 CFM | No |
| RTU-2 | 3 ton | ~1,200 CFM | No |
| RTU-3 | 5 ton | ~2,000 CFM | Borderline, verify AHJ |
| RTU-4 | 7.5 ton | ~3,000 CFM | Yes |
| RTU-5 | 10 ton | ~4,000 CFM | Yes |
RTU-3 was the one that started the argument. Nameplate said 5 tons, which puts it right at 2,000 CFM nominal. Actual tested airflow at high speed came in around 2,150 CFM. AHJ wanted the detector. If you've got a unit sitting on the threshold, use actual airflow data from the mechanical schedule, not just the nominal tonnage. The nominal is just an approximation anyway.
Sampling tube sizing
Once you've confirmed a duct detector is required, the inlet and exhaust sampling tubes need to be sized to actually sample the airstream. NFPA 72 won't give you a depth chart here. The manufacturer's listing is where the real rules live.
General idea: the inlet tube needs to pull a representative sample from across the duct cross-section. For ducts narrower than about 3 feet, a single tube extending roughly to the center is typical. Wider ducts may need a perforated tube spanning the full width, or multiple sampling points, depending on the listing.
Don't guess on this. Pull the manufacturer's installation guide, check the listed coverage table, and match it to your actual duct dimensions. Some manufacturers give you a chart based on duct width, some give a formula. Either way, it shows up in the cut sheets you submit, and plan checkers do cross-reference it. I've seen submittals kicked back because the tube depth shown on the detail didn't match the listed coverage for that duct width.
Remote test and reset station placement
NFPA 72 requires duct detectors to be accessible for testing and resetting. If the detector is mounted at 14 feet in a mechanical room where you'd need a ladder every time you wanted to test it, you need a remote test and reset station.
"Readily accessible" is the code language. In practice, most AHJs interpret that as: if it takes a ladder or a tool to reach the detector head directly, install a remote station. Some inspectors are stricter than others on that call.
Where should it go? Near the air handler, or at the mechanical room entrance, low enough to reach without a ladder, visible from the unit when possible. Confirm with your specific AHJ before the rough-in. On the submittal, label it explicitly and tie it back to the device it serves on the riser or system diagram. Don't just note it on the mechanical plan and call it done.
IMC vs NFPA 72, and what to do when they overlap
Here's the part that creates the most paperwork headaches. The IMC tells you where and when to install duct detectors. NFPA 72 governs how they have to perform, be listed, and tie into the fire alarm system. Both apply. On most jobs they don't conflict, but occasionally the IMC specifies a physical location that the listing requirements complicate.
Common example: the IMC location puts the detector too close to an elbow or in a section of duct that won't provide a representative air sample per the listing. When that happens, document why you deviated and how you resolved it. The AHJ appreciates seeing that you thought it through. Ignoring one code because it's inconvenient is not a real answer.
General rule when two codes overlap: the stricter requirement governs.
What the AHJ actually wants to see on the riser
This one I'll just say plainly. Most AHJs want duct detectors shown on the riser diagram as a distinct device, even though they connect to the panel as a control or supervisory input. Don't bury them in a general note or assume the mechanical plan covers it.
Show the device symbol. Show the remote test and reset station. Show the relay output if it's shutting down the air handler on alarm. Show which zone or circuit it lands on. A complete riser means fewer questions back from the plan checker, and fewer questions means a faster permit.
It's also worth adding a short note on the cover sheet calling out your duct detector design basis: the CFM basis, the code section you're designing to, and the detector model. Saves a lot of back-and-forth emails.
Anyway. If you're regularly doing the submittal assembly by hand after doing all this work, that's the part I got sick of. Full disclosure, I built a tool called FireDeck that handles the battery calc, voltage drop, device schedule, and cut sheet side of the package. Free trial at firedeck.app if you want to see if it saves you time. Obviously biased.