Voice Evacuation & Emergency Communications
EVAC, ECS, and Mass Notification
Voice evacuation turns a generic horn tone into a building-specific instruction. NFPA 72 Chapter 24 governs the whole ECS ecosystem — from in-building fire voice/alarm through mass notification and responder radio.
EVAC vs ECS vs MNS — The Terminology
Fire alarm voice systems go by many names, and the names actually matter. NFPA 72 Chapter 24 organizes them into a hierarchy:
When Voice Evacuation Is Required
IBC §907.5.2.2 and NFPA 72 Ch 24 define when EVAC is mandated:
Anatomy of an EVAC System
A modern in-building EVAC has eight functional blocks. Each is supervised by the panel and most are UL 864 listed as fire alarm control equipment:
Standard Voice Messages
A well-designed EVAC system has 4–6 pre-recorded digital voice messages. Each plays in response to a specific panel condition and loops on a tone-message-tone cadence:
Message Design Best Practices
Use a calm, adult voice matched to the building demographics. Avoid robotic/synthetic TTS for primary evacuation — human voice is 10–15% more effective at moving occupants. Include a short beep/tone before each message so occupants lock onto the audio. Repeat the message on a 30–45 second cycle. For multilingual buildings, sequence languages (English → Spanish → …) rather than playing them simultaneously.
Intelligibility: CIS and STI
Sound-pressure level alone does not prove voice evacuation works. You can have 95 dBA at every seat in an auditorium and still have unintelligible speech because of reverberation, echoes, or speaker overlap. NFPA 72 §18.4.11 requires quantified intelligibility in every Acoustically Distinguishable Space (ADS):
What Kills Intelligibility
- Long reverb time (RT60 > 1.5 s) — hard surfaces, high ceilings, minimal acoustic treatment
- High ambient noise — mechanical rooms, manufacturing floors, airport terminals, casinos
- Excessive speaker overlap — multiple speakers reaching the same listener at similar amplitude create comb-filter cancellations
- Speakers too far apart — occupants in "dead zones" between speakers rely on reflected sound, which is always less intelligible
- Wrong tap settings — too-high taps cause clipping; too-low taps fall below the ambient threshold
- Wall/ceiling changes after commissioning — renovations that remove carpet, add glass walls, or remove acoustic treatment invalidate the original intelligibility test
ADS Mapping Is Not Optional
NFPA 72 requires the design submittal to identify every Acoustically Distinguishable Space (ADS) — a single acoustic zone with uniform finish, ceiling height, and reverb characteristics. Each ADS gets its own STI/CIS target. A 100,000 sq ft building with 40 different acoustic environments has 40 ADS measurements at acceptance. Skipping ADS mapping is a common design-submittal deficiency and a frequent reason for AHJ rejection.
Survivability for EVAC Risers
In a high-rise, the EVAC speaker riser passes through every floor — including the fire floor. If the riser burns through in 8 minutes, everything above that floor goes silent. NFPA 72 §24.4 defines four survivability levels; most high-rise EVAC risers must meet Level 2: the circuit integrity survives 2 hours of direct fire exposure, either through UL 2196 listed cable (CI or MI construction) or through enclosure in a 2-hour rated shaft.
Survivability Level by Application
- Level 0 (no fire rating) — low-rise tone-only systems, small voice systems where alarm reaches all occupants before fire reaches the wiring
- Level 1 (mechanical protection) — standard commercial EVAC in the protected shaft
- Level 2 (2-hour rated cable or shaft) — high-rise EVAC risers, partial evacuation systems, most mass notification
- Level 3 (Level 2 + geographic separation) — rare; effectively replaced by Level 2 with diverse routing
See our Fire Alarm Pathway Classes article for the full treatment of Class A/B/X/N plus survivability levels — the two concepts are complementary and must both be specified.
Mass Notification Override
Where an MNS and an EVAC share speakers, NFPA 72 §24.4.1 requires a priority hierarchy governed by a formal Risk Analysis. Non-fire MNS messages can override fire alarm messages if the risk analysis justifies it. Typical hierarchy:
The rationale: during an active shooter event, an "evacuate the building" announcement would actively endanger occupants. An MNS shelter-in-place message must be able to override the fire alarm — but only after a formal risk analysis, documented in the ECS design package, and approved by the AHJ. Without the risk analysis, fire alarm always takes priority. NFPA 72 §24.4.1
Acceptance Testing
Voice evacuation acceptance testing is substantially more involved than tone-only acceptance:
- Silence the background — acceptance happens in the unoccupied building with HVAC at typical noise
- Verify message integrity — every pre-recorded message plays correctly on every zone
- Verify live paging — FF mic delivers clear speech to every zone with correct channel select
- Measure SPL — dB at every representative listener location per NFPA 72 §18.4
- Measure STI / CIS — with calibrated analyzer in each Acoustically Distinguishable Space
- Verify priority / override — if MNS is installed, confirm hierarchy with simulated multi-input conditions
- Verify backup amplifier transfer — intentionally fail the primary amp and measure time to backup engagement
- Verify battery run-time — cut AC, run system through full 24-hour standby + 15-minute voice alarm at representative load
- Document ADS map, speaker tap settings, amp loading, and STI results — submit to AHJ and retain for future ITM baseline
ITM for EVAC Systems
NFPA 72 Chapter 14 adds specific requirements on top of standard fire alarm ITM:
- Quarterly: test each pre-recorded message plays correctly. Verify FF mic silences tones and broadcasts live speech.
- Annual: full functional test of every speaker — amp, NAC, speaker. Verify backup amp auto-transfer. Verify priority hierarchy.
- Annual: battery load test of the EVAC amplifier battery set at full voice load (the 15-minute alarm duration, not 5 minutes).
- After any renovation — repeat STI/CIS measurement in affected ADS if room finishes, walls, or speaker layout changed.
- 5-year: full intelligibility re-measurement in all ADS — baseline may have drifted with furniture changes, HVAC replacements, or minor renovations.
Frequently Asked Questions
What is the difference between EVAC and ECS?
When is voice evacuation required?
What is the CIS / STI intelligibility score I need to hit?
What are the standard EVAC messages?
Can I use a PA system as an EVAC system?
Do firefighter phones still exist?
References
1. NFPA 72 (2022), Chapter 24 — Emergency Communications Systems.
2. NFPA 72 (2022), §18.4.11 — Speech intelligibility.
3. NFPA 72 (2022), §24.4 — Survivability levels.
4. NFPA 72 (2022), §24.5 — Two-way communication.
5. IBC §907.5.2.2 — Emergency voice/alarm communication.
6. IEC 60268-16 / IEC 61672 — STI measurement methodology.
7. ANSI S3.2 — Common Intelligibility Scale.
8. UL 2572 — Mass Notification Systems.
Was this article helpful?
Rate this article to help us improve
Discussion (3)
The single biggest cause of EVAC intelligibility failure is late-stage design changes to the room finishes. The engineer models STI assuming carpeted floors and acoustic ceiling tile. Owner value-engineers to polished concrete and exposed deck. Reverb time doubles, intelligibility drops from CIS 0.78 to 0.52, and we are adding 30% more speakers at commissioning to pass the test. Always re-run the acoustic model when finish packages change.
This is why we push a coordination meeting between the FA designer and the interiors architect during design development — not just at CD review. A change to polished concrete is an opportunity to add acoustic baffles or sound masking as part of the finish package, not fight about it after construction.
Pre-recorded messages are where owners try to save money and end up in trouble. Default factory voice prompts are fine for a small retail space — but for hotels, schools, and high-rise, we strongly recommend professionally recorded messages in the voice that matches the building demographics. Generic robotic voice on a K-12 system is less effective at moving students than a human voice they recognize. The cost delta is maybe $500. Worth it.
When reviewing a high-rise EVAC design, I always ask for the speaker layout stamped with dB and CIS predicted values at each measurement grid point, plus the assumed acoustic finish package. If the design package does not include acoustic modeling, I reject it. Intelligibility is not something you hope for — it is something you engineer.