What Notification Appliances Have to Achieve

Why this matters

Notification appliances are the only part of a fire alarm system whose requirement is written at the occupant rather than at the device. A detector is judged by whether it is listed and correctly placed. A horn is judged by a level somebody has to stand in the room and measure, against an ambient level somebody else measured in the same room earlier. Two site measurements bracket the job, and when either is missing the design was a guess. This card covers the two performance criteria, the five labels every level has to carry, why adding appliances is a weak lever, and what the same criteria do in a quiet office and a noisy shop.

The two jobs, and they are not the same job

Audible notification has to be heard above whatever the room is already doing. It is a level comparison, and it fails as a building gets louder, gets partitioned, or hands its occupants hearing protection. Visible notification has to put light into every part of the space an occupant can be in. It is a geometry problem, and it fails when a room gets bigger, gets an L in it, or gets a partition that casts a shadow. It is also the criterion that governs for occupants who are deaf or hard of hearing.

NFPA 72, the National Fire Alarm and Signaling Code, sets both, in the edition your authority having jurisdiction has adopted and amended, and it binds the building owner and reaches the trades through the permit and the acceptance test. The AHJ is a named role with authority to approve, commonly the fire marshal or the building official, and not a synonym for whoever inspects.

The federal floor is separate and narrower. 29 CFR 1910.165 requires an employee alarm system to be perceived above ambient noise or light levels by all employees in the affected portions of the workplace, and requires tactile devices where they would not otherwise perceive it. That duty runs to your employees and is not a substitute for the adopted code, nor the reverse. When you quote a number, say which document you are quoting.

A level is meaningless without its five labels

The criterion is a subtraction between two numbers, so both have to be on the same basis. A sound level carries all five of these or it is not a measurement.

1. The quantity. Sound pressure level, not sound power level. Pressure is what an ear or a meter at a place experiences; power is a property of the source and has no location.
2. The reference. Sound pressure level is decibels referenced to 20 micropascals; sound power level is decibels referenced to 1 picowatt. The same word over two references gives numbers roughly 10 to 20 apart for ordinary sources, more than the entire margin the criterion asks for.
3. The weighting. A-weighted, C-weighted or unweighted. The criterion is written in A-weighted decibels, dBA, and a C-weighted reading of the same room runs well above it where the noise is low frequency. A shop with a big fan is exactly that room.
4. The bandwidth. Overall broadband, or per octave band. The criterion is broadband; octave band data is for diagnosing why an appliance is inaudible.
5. The time basis. Equivalent continuous level over a stated period, written Leq, or an instantaneous reading on fast or slow response, or a maximum. The criterion needs an average ambient and a maximum that persists, two different readings from the same walk.

Write it in full on every report: "82 dBA, A-weighted sound pressure level re 20 micropascals, broadband, Leq over 10 minutes, meter at 5 ft above finished floor." A reader six years from now can act on that. They cannot act on "82 dB."

The audible criterion, stated the way the code states it

In public mode, meaning notification to the general occupants, NFPA 72 in the adopted edition sets the required signal at the greater of two figures: 15 dBA above the average ambient sound level, or 5 dBA above the maximum sound level having a duration of at least 60 seconds, measured 5 ft above the floor in the occupiable area. In private mode, notification to people trained to respond, the first figure drops to 10 dBA. In sleeping areas the criterion is an absolute level at the pillow rather than a margin, commonly 75 dBA. The code also caps the signal at 110 dBA at the minimum hearing distance.

The "whichever is greater" is a real fork and it tells you something. Where the average and the sustained maximum are close, the 15 dBA branch governs and the room is steady. Where a loud intermittent source dominates, the 5 dBA over maximum branch can govern, and one machine is setting your design.

Once the criterion is met, the ambient stops mattering to the total. Levels combine on a power basis, not by adding decibels. A signal 15 dB above ambient carries 31.6 times the ambient's acoustic power, so the combined level is the signal plus about 0.1 dB. At 15 dB up the ambient is acoustically invisible, which is the whole reason the criterion is written as a margin.

Why you cannot fix a shortfall by adding appliances

Independent sources of equal level combine on a power basis, so doubling the appliance count at the same distance adds 3 dB, not 6. Closing a 15 dB gap that way needs 10 raised to 1.5, about 32 times the count. Nobody installs 32 horns in a bay.

Distance is the lever that works. For a point source in a free field, with no reflecting surfaces other than the one it is mounted on, level falls 6 dB per doubling of distance, so moving an appliance from 40 ft to 20 ft buys the same 6 dB as quadrupling the count, with one device instead of four. That relationship holds only in the free field condition just named; in a hard reverberant room the level flattens beyond a critical distance instead of continuing to fall, so the free field model under-predicts there and over-predicts in an absorptive room. The other levers are a higher output appliance or tap setting, recorded on the drawings, and reducing the ambient.

Note what the appliance rating already contains. Audible output is published as a level at 10 ft, produced under a standardized test method in a reverberant chamber, so it already contains that chamber's acoustics. Applying free field inverse square decay to it is not adding a term, it is re-basing the number onto a different acoustic condition, and it will be wrong in both directions depending on the room. That is why the criterion is verified by measurement at acceptance, not by a spreadsheet.

The visible criterion is geometry, not loudness

Visible appliances are rated in candela of effective intensity, and NFPA 72 in the adopted edition assigns required candela through a spacing table keyed to room size and mounting height rather than a calculation you perform. Wall mounted appliances go in a stated height band, commonly not less than 80 in and not more than 96 in above finished floor, and the wall mounted table assigns 15 cd to a 20 ft by 20 ft room and 30 cd to a 30 ft by 30 ft room, continuing up in the adopted edition.

Three rules travel with that table. Non-square rooms are handled by the largest dimension, or by subdividing into squares each covered by its own appliance, which is usually cheaper in candela and always cheaper in glare. Obstructions cast shadows, so a partition blocking line of sight leaves that point uncovered whatever the room dimensions say. And appliances in the same field of view must flash in synchronization.

The same requirement run against two rooms

Room one: an open plan office, 60 ft by 40 ft, no full height partitions. Ambient measured 48 dBA, A-weighted sound pressure level re 20 micropascals, broadband, Leq over 30 minutes at 5 ft above finished floor. Maximum level with a duration of at least 60 seconds was 55 dBA, from the rooftop unit's high speed stage.

• 15 dBA over average ambient: 48 + 15 = 63 dBA.
• 5 dBA over the sustained maximum: 55 + 5 = 60 dBA.
• Greater governs: 63 dBA, and the 15 dBA branch won, which says the room is steady.

An appliance published at 88 dBA at 10 ft, under free field decay, reaches 82 at 20 ft, 76 at 40 ft and 70 at 80 ft. The worst listener is about 36 ft from a centrally located appliance, the half diagonal of a 60 by 40 room, and free field decay from 88 dBA at 10 ft puts about 77 dBA there, so the audible criterion is satisfied with margin by one device. Subdividing 60 by 40 into 30 ft squares gives two squares along the length and, because 40 divided by 30 is 1.33, two across the width: four appliances at 30 cd each. Audible needed one, visible needs four. The count here is set by geometry, and an estimator who priced it off the audible calculation is short by three devices.

Room two: a fabrication bay, 80 ft by 50 ft, concrete floor, metal deck, block walls. Ambient measured 82 dBA, same five labels, Leq over a 10 minute representative production period at 5 ft above finished floor. Maximum with a duration of at least 60 seconds was 88 dBA, the dust collector.

• 15 dBA over average ambient: 82 + 15 = 97 dBA.
• 5 dBA over the sustained maximum: 88 + 5 = 93 dBA.
• Greater governs: 97 dBA, leaving 110 - 97 = 13 dB of legal headroom, which is the entire design space.

The specified appliance is published at 95 dBA at 10 ft, already 2 dB below the requirement at the closest listener, before any distance is applied.

Correction, printed. The 95 dBA figure is a reverberant chamber result at 10 ft. Re-basing it onto this bay substitutes one acoustic condition for another rather than adding a room term to a free field number, and in a hard block and deck bay the real level beyond a few appliance-distances will sit above the free field prediction. That direction is favourable and it is still not a design basis, because nobody knows the critical distance in this room without measuring it. What the design needs is a measured level at the worst listener at acceptance, at 5 ft above finished floor, reported with all five labels.

The finding that changes the means, not the count. Employees in this bay wear hearing protection under a hearing conservation program per 29 CFR 1910.95, so whatever the meter reads at 5 ft above the floor, the protected ear receives that level minus the protector's attenuation and the 97 dBA target does not describe what the occupant experiences. Under 29 CFR 1910.165 the alarm must be perceived above ambient noise levels by all employees in the affected portion, which in a room where everyone is protected cannot be answered by the horn alone. Visible notification carries the bay, sized off its geometry, and any employee whose station sits behind an obstruction gets a device with line of sight or a tactile means.

A hazard the acceptance test itself creates. Sounding evacuation signals unannounced starts a real evacuation and can dispatch the fire department. Notify the occupants and the supervising station before the test and confirm both, and treat the system as impaired for the duration with a start, a compensating measure and a named end.

Sibling rule check, printed. Against the supervisory signal card: these appliances answer alarm signals only, so nothing in either room is expected to sound on a closed valve or a ground fault. Against the duct detector card: the rooftop unit's duct detector is a supervisory point and correctly appears in neither calculation. Against the panel trouble card: the standby battery calculation must carry the alarm current of all four office appliances plus the bay's devices, exactly the load that card warns gets left out when appliances are added after commissioning.

How to verify you got this right

Two measurements and one drawing, and the job is not closed without all three. The ambient survey taken before design, with all five labels and the location of every reading. The achieved level survey at acceptance, at the worst listener in every notification zone, same five labels, against the criterion that governed. And a plan showing every visible appliance with its candela rating and the square it covers, the only document that survives the next fit-out when a new partition puts an occupiable corner in shadow.

References

• NFPA 72, National Fire Alarm and Signaling Code, in the edition your authority having jurisdiction has adopted and amended, which sets the audibility criteria and the visible appliance spacing tables
• 29 CFR 1910.165, employee alarm systems, requiring the alarm be perceived above ambient noise or light levels by all employees and requiring tactile devices where it would not be
• 29 CFR 1910.95, occupational noise exposure, where hearing protection changes what the occupant receives
• Manufacturer published output data, which states the level at 10 ft and the test condition it came from
• See related: What a Supervisory Signal Is and Why It Is Not an Alarm; How to Work Out Why a Panel Keeps Going Into Trouble