Public-address system design: STI targets, zoning, and the line-array hang nobody calculates

A public-address system is one of the most consistently mis-specified ELV layers in commercial and civic buildings. The brief usually arrives as a wattage figure ('we need 1500 watts of PA') or a brand preference ('JBL or Yamaha'), neither of which decides the actual outcome. The deliverable is speech intelligibility at every seat — measured as STI (Speech Transmission Index) — and the wattage and brand are downstream consequences of designing for that target.

## STI is the design target; everything else is implementation

STI is a single number between 0 and 1 that captures how well a listener can understand spoken words at a given location. STI < 0.45 is poor; 0.45–0.55 is fair; 0.55–0.65 is good; > 0.65 is excellent. We design every PA project to STI ≥ 0.55 at every seat (the IS-16102 voice-alarm threshold for life-safety) and we typically deliver STI 0.62–0.68 across audience planes when the brief is performance.

STI is degraded by three things: reverberation (room reflections that smear consonants), noise (HVAC, audience, external), and direct-to-reverberant ratio (the amount of direct sound versus room-reflected sound at the listener). The design loop is to model the room's acoustic behaviour, place the speakers to deliver high direct-to-reverberant ratio at the listening plane, treat the room to control reverberation, and check the noise floor at the worst-case use condition. We model in EASE for any audience plane above 200 seats; below that, calibrated experience and a measured impulse test are usually sufficient.

## Line-array hang geometry — the part nobody calculates

Above 400 seats, the standard answer is a line-array hang. The mistake most integrators make is to specify the line-array off the manufacturer's catalogue without engineering the hang geometry to the room. A line-array's coverage pattern is set by the splay angles between elements, the J-curve at the bottom of the array, and the DSP shading that compensates for distance differences between the front and back rows.

We model every line-array hang in the manufacturer's design tool (K-array K-Framework3, JBL Line Array Calculator, d&b ArrayCalc) before procurement. The output is a per-element splay angle, a per-element DSP shading filter, and a measured SPL coverage map across the audience plane that should fall within ±2 dB front-to-back and ±3 dB seat-to-seat. We commission against the model and we publish the measured coverage map in the commissioning report.

Most line-arrays we audit have been hung off the catalogue specification with uniform splay angles and no DSP shading, producing the characteristic 'hot front rows, dead back rows' pattern that audiences experience as 'the PA is too loud at the front and too quiet at the back'. That is not a wattage problem. It is a geometry problem, and the fix is a half-day re-hang with calculated splay angles and DSP shading.

## Zoning and emergency over-ride

Routine paging — meeting calls, lunch announcements, lost-child broadcasts at a mall — runs on the PA's primary zone configuration. Emergency announcements — fire-alarm voice-evacuation, controlled lockdown — must pre-empt routine paging without manual operator intervention, and the routing has to be hardware-enforced rather than software-only.

We zone amplifiers so each zone has a hardware priority input that can override the routine paging from a higher-priority source. The fire-alarm panel's voice-evac module is wired into the highest-priority input on every zone amplifier; on a fire-alarm trigger, the zone amplifiers switch to the voice-evac source regardless of what the routine PA is broadcasting. This is the IS 16102 / IEC 60849 architecture for voice-alarm systems, and it is what separates a building's PA from its life-safety system.

## Voice-alarm overlay where the building requires it

For occupancies above the threshold defined in IS 16102 (auditoria above 250 seats, hospitals, hotels above defined room counts, malls above 5,000 sq m), the PA system has to be designed as a voice-alarm system to IS 16102 / IEC 60849 / EN 54-16. That standard adds requirements: supervised cabling, redundant amplifiers, battery-backed power for 30 minutes minimum, signal monitoring, and per-zone fault reporting. The voice-alarm overlay is engineered as one with the routine PA, not as a separate system, because the same speakers, amplifiers and zone topology serve both.

## Callout — what buyers most miss

**STI per seat is the deliverable. Wattage per zone is the cost line.** Most PA tenders spec wattage and brand and leave the STI question to commissioning. By that point, the geometry is fixed and the only lever left is more wattage — which does not improve STI in a reverberant room. Tender for STI per seat as the acceptance criterion; the wattage will be whatever it needs to be.

## Reference deployment context

The Capital Cultural Hall PA was designed with JBL VLA-series line-array hangs at the proscenium on Crown DCI amplification and a BSS BLU100 DSP profile, with calibrated splay angles and per-element shading published in the commissioning report. STI was measured at 0.64 mean across the 1,800-seat audience plane with a 0.58 minimum at the rear corners. The amplifier zones are voice-alarm-rated to IS 16102 and the fire-alarm panel pre-empts routine paging on a hardware-enforced priority input. The Town Hall Dimapur PA, on a smaller compact line-array hang, hit STI 0.61 at every seat across the 400-seat hall.

## References

1. IS 16102:2014 — *Voice Alarm Systems — Sound systems for emergency purposes*.

2. IEC 60849 (now IEC 7240-19) — *Sound systems for emergency purposes*.

3. AES-15id-1991 — *Sound system design — Speech intelligibility*.

4. ISO 3382-1:2009 — *Acoustics — Measurement of room acoustic parameters*.

5. EN 54-16:2008 — *Voice alarm control and indicating equipment*.