What a Smoke Control System Is Trying to Achieve

Why this matters

Smoke control is the one fire protection system a mechanical contractor can break with a balancing damper. It shares fans, shafts, dampers, and controls with the comfort system, it gets touched during every tenant fit-out, and its acceptance criterion is a number almost nobody on site can see. Crews who would never touch a sprinkler head will happily re-aim a makeup air louver, and the building loses a life safety system that afternoon. The correction is understanding what the system is actually trying to do, because it is not what the name suggests.

It does not remove smoke, it holds a boundary

A smoke control system's deliverable is a boundary held in a stated place for a stated time. Nothing in the design promises clear air.

Two families do it two different ways, and they have different acceptance criteria.

Smoke containment holds a pressure difference across a barrier so smoke cannot cross it: a pressurized stair, a pressurized elevator hoistway, a zoned system that pressurizes the floors above and below the fire floor relative to the fire floor. Its criterion is a differential pressure across a specific barrier, measured in inches of water column, in a stated direction, with the doors in a stated position. A pressure with no direction and no door position stated is not a measurement.

Smoke management exhausts an unseparated volume, usually an atrium or a large covered mall, so that the interface between the buoyant smoke layer above and the relatively clear layer below stays above the highest walking surface people are using. Its criterion is an interface height above a named surface, held for a stated design time, under a design fire that is an input from the design documents.

The unifying idea: the criterion is a boundary and a duration, never an absence of smoke. A reader who takes "the atrium looked clear" as a pass has measured nothing, and a reader who takes "there was smoke at level three" as a fail has also measured nothing until they know where the design put the interface.

Where the design numbers come from, and who owns them

Design fire size, design time, required interface height, required pressure differences, and required exhaust rates all live in the design documents prepared by the engineer of record and approved under the adopted building code. NFPA 92, in the edition your authority having jurisdiction has adopted and amended, is the standard those documents work under, and the AHJ is a named role with authority to interpret and enforce the adopted edition, not a synonym for whoever inspects. Periodic testing intervals are set by the adopted building code and commonly differ between dedicated systems and systems shared with the comfort HVAC, so read the interval out of the adopted edition rather than assuming a year.

Two relationships are worth carrying in your head, both with their held-constant conditions attached.

Plume mass flow grows with clear height. For an axisymmetric plume at a constant design fire heat release rate, with no balcony spill and no ceiling jet effects, the mass of air entrained into the plume grows roughly with the five-thirds power of the clear height above the fire. Raising a required interface from 34 ft to 40 ft above the floor, at that same constant fire size and plume geometry, raises the mass to be exhausted by roughly 30 percent. Change the plume type, to a balcony spill plume or a window plume, and the relationship changes with it.

Makeup air is a limit, not a leftover. Exhaust cannot exceed what comes in, and the air that comes in cannot be moving fast enough to blow the plume apart. NFPA 92, in the adopted edition, commonly caps makeup air velocity at the plume at 200 ft/min unless a specific engineering analysis supports a higher value. The velocity is measured at the opening's free area, which is the clear area after louver blades, screens, and door leaves are subtracted, not the rough opening.

A narrative case: the atrium that passed on fans and failed on air

Three-level retail atrium, dedicated smoke exhaust, semiannual test under the adopted building code. The design requires the smoke layer interface held at least 6 ft above the highest walking surface, which is the level three balcony at 34 ft above the atrium floor, so the interface target is 40 ft, held for 20 minutes.

The signal. The previous test had passed. This one was ordered after a tenant fit-out on level one that replaced a section of storefront and re-hung the automatic entrance doors. The technician's first reading looked fine and the test still failed.

Before anything ran. The sequence was initiated from the firefighters' smoke control panel, which drives dampers and fans to their extremes. That command is itself a hazard and it gets handled before it is given: the building was cleared of occupants for the test window, the two fuel-fired rooftop units served by an affected shaft were shut down and their gas cocks closed and tagged so a depressurized shaft could not pull products of combustion back down a vent, and the outside air preheat coils in the affected air handlers were verified drained because the sequence drives those units to full outside air and the test ran in February. Fan room work was done with the starter locked and tagged under 29 CFR 1910.147 before any guard came off; no belt guard was removed with a fan able to start. Duct velocity readings were taken through test ports, never by reaching into a running duct.

Reading one, exhaust. Four fans, design total 180,000 cfm, design documents accept measured exhaust not less than 90 percent of design. Traverse total at the fan discharges: 172,000 cfm, or 95.6 percent. That reads as a comfortable pass.

Correction, printed, and it is a re-basing not an addition. The traverse plane is at the fan, downstream of roughly 90 ft of shaft and plenum. What the fan moves already includes air that leaks into that duct from outside the smoke zone, so the figure to compare against design is not the traverse reading plus a leakage term, it is the traverse reading re-based downward by the leakage share. Measured duct leakage at test pressure came to about 3 percent of the traverse. Exhaust actually drawn from the atrium volume: 172,000 times 0.97, about 166,800 cfm, which is 92.7 percent of the 180,000 cfm design. Still a pass, and still worth printing, because the same correction on a system starting at 91 percent turns a pass into a fail and it runs in the unflattering direction every time.

Reading two, makeup. Design makeup is 90 percent of design exhaust, 162,000 cfm, split by design between two motorized louver banks and the level one storefront openings, with half the makeup, 81,000 cfm, coming through the storefront. The storefront's measured free area after the fit-out is 240 sq ft. That gives 81,000 divided by 240, about 337 ft/min at the opening.

Against the 200 ft/min limit in the adopted NFPA 92 edition, 337 ft/min is 1.7 times the cap. The maximum flow that opening can carry at the cap is 200 times 240, which is 48,000 cfm. The storefront was carrying 81,000, so 33,000 cfm has to move to the louver banks or the free area has to grow.

What the tracer smoke showed, and what it did not prove. A cold tracer release put the visible interface at roughly balcony level rather than 6 ft above it. That observation is qualitative only. The adopted NFPA 92 edition does not accept a tracer or smoke-bomb demonstration as a substitute for verifying the design quantities, and it should never be reported as if it were the acceptance test. It is useful as a symptom, and here it pointed the same direction as the velocity reading: the makeup jet was punching into the plume and dragging the layer down. The tracer fluid's own SDS was read first, the fire alarm system was placed on test with the monitoring station and the AHJ notified before release and taken off test the same visit, and the release was made from the atrium floor with the level occupied only by the test crew wearing the respiratory protection the SDS specified under a written program per 29 CFR 1910.134.

The correction. Rebalance the makeup split so the storefront carries not more than 48,000 cfm, with the remaining 114,000 cfm through the louver banks, then re-run the test and re-measure velocity at the storefront free area. The fit-out contractor had reduced the storefront's free area by re-hanging the doors with a wider center mullion, and nobody's scope of work said the word smoke.

Sibling-rule check. The building was cleared before the sequence was driven, so no egress path was narrowed or blocked while the test ran, which the egress-path card requires. No door hardware was altered and no door was held closed by the test, so the locked-door card's one-motion gate held throughout; the pressure differences involved here are across an atrium volume, not across a stair door, and the door-force interaction belongs to the stair pressurization card. The alarm system's on-test and off-test times, the gas shutoff and its restoration, and the makeup deficiency all carry a start time, an owner, and an end, which the fire protection records card requires of any bounded out-of-service state. Every level and pressure in this case carries its plane, its direction, and its door position.

What flips the answer

Change the plume type and the arithmetic above stops applying. A balcony spill plume from a fire under a projecting level, or a window plume from a fire in a room venting into the atrium, entrains differently from the axisymmetric case the 30 percent figure was derived under, and the design documents will say which one the building was designed for.

Change the system from dedicated to non-dedicated and the failure surface widens sharply. A dedicated smoke exhaust fan does nothing else, so it either runs or it does not. A shared comfort air handler pressed into smoke control duty inherits every filter change, every belt, every damper actuator, and every balancing decision made by somebody solving a comfort complaint. On a shared system the question after any HVAC work is not whether comfort was restored, it is whether the smoke control sequence still produces the design quantities, and only a test answers that.

References

• NFPA 92, Standard for Smoke Control Systems, in the edition adopted and amended by your authority having jurisdiction, for design criteria, makeup air velocity limits, and acceptance testing
• The adopted building code chapter on smoke control systems, for special inspection and periodic testing intervals
• 29 CFR 1910.147 for lockout and tagout of mechanical and stored energy before working on fan equipment; 29 CFR 1910.333(b)(2) where the work is electrical
• 29 CFR 1910.134 for respiratory protection under a written program
• See related: Why a Stair Pressurization System Fails Quietly; What a Fire Protection Record Has to Show a Year Later