Event announcements: intelligible, multilingual and resilient

An announcement is only as good as the sentence the audience actually hears. On a festival site, in an exhibition hall or in front of a stage, it competes with music, conversation, wind and reverberation. If you need to say “Please leave this area using the exits on your left”, you have to be confident that people will understand it – including visitors who do not speak the local language.

This article sets out what intelligibility depends on, which figures the standards and technical literature give, and what that means for planning and running an event.

The problem: loud is not the same as clear

Plenty of systems are loud enough and still hard to follow. The causes are nearly always the same: too little margin over ambient noise, echoes between several loudspeakers, reverberation off hall walls, and sentences that are too long or spoken too quickly. Organisational gaps add to this – a message is drafted only once the incident is under way, a translation is missing, or the system relies on an internet connection that cannot be trusted on site.

The fundamentals in figures

Measuring intelligibility: STI

The Speech Transmission Index (STI) is a standardised measure of speech intelligibility on a scale from 0 (unintelligible) to 1 (perfect). Emergency sound systems are expected to reach at least 0.50, which corresponds to “fair” intelligibility. The main factors are the signal-to-noise ratio, reverberation time, echoes and distortion. STI is measured on site with the short-form STIPA method described in IEC 60268-16, because the result always depends on the space, the noise and the distance as well as on the equipment.

For speech, the frequency range that matters most is roughly 250 to 4,000 Hz (ITU-T P.50).

Level and margin over noise

The standards for emergency sound systems (EN 50849, published in Germany as DIN EN 50849 / VDE 0828-1) give two simple reference values:

• at least 75 dB sound pressure level at the listener’s position;
• at least 6 dB above ambient noise, with at least 10 dB recommended for critical messages such as evacuation or fire.

Typical noise levels:

Type of event	Typical ambient noise
Conference	55–65 dB
Trade fair	65–75 dB
Concert	80–100 dB
Open-air festival	85–110 dB

At a concert, talking over the music is rarely enough. The music has to be turned down or paused for the announcement.

Range: 6 dB per doubling of distance

In the free field, sound pressure level falls by 6 dB every time the distance doubles (Kuttruff; Kinsler et al.). In practice, wind, temperature gradients, reflections and absorption shorten that distance considerably. Covering a large area therefore calls for several distributed loudspeakers rather than one very loud one.

Several loudspeakers and echoes

If people hear the same sentence from two loudspeakers more than about 50 ms apart, they perceive a separate echo – the Haas effect. At 20 °C sound travels at roughly 343 metres per second, so 50 ms equates to a path difference of just over 17 metres. Overlapping coverage needs a suitable delay or clearly separated directions of throw.

What this means in practice

Put loudspeakers where people are

Plan coverage around where the audience actually stands, not around where power happens to be available. Entrances, the area in front of the stage, walkways and queuing areas all need coverage. Where there is no camera mast, visitor announcement systems (BDS) fill the gap: free-standing loudspeaker columns on telescopic masts. Direction, tilt and mounting height decide whether the sound reaches the audience or the neighbouring buildings.

Think in zones

Not every announcement belongs everywhere. A request to ease pressure at an entrance concerns that entrance; an evacuation concerns everyone. Divide the site into areas that can be addressed individually or together, so that people in unaffected parts are not unsettled unnecessarily.

Prepare messages – do not improvise

There is no time to draft wording during an incident. Keep a library of prepared messages for typical situations – evacuation, fire, medical emergency, security, severe weather, area closure, end of event – arranged as an escalation ladder from mild to serious. Use short sentences, one instruction each, a calm tone and a clear direction. Slower speech with pauses between sentences holds up better against reverberation and echo.

Several languages – with checked translations

An international audience needs announcements in several languages, played one after another in a fixed order. Machine translation is usually serviceable but can distort individual sentences: “Please stay calm” can come out as something closer to “Do not move”. Safety messages should therefore be checked by a person in advance.

Bring music and rides into the system

At fairs and festivals, stages and rides run their own sound. A safety announcement has to be able to interrupt that music, otherwise it is lost – this follows directly from the requirement for a 6 to 10 dB margin over ambient noise.

Resilience: plan for offline operation

Mobile networks and internet access are often overloaded at large events. Announcements must not depend on them. Text, translations and speech output should be held on site and generated before doors open, so that nothing has to be computed in an emergency.

Keep a record

Log every announcement with the time, text, languages, loudspeakers reached and the person who triggered it.

How MOVETOS OPS handles announcements

MOVETOS OPS plans loudspeakers on the site map, showing each one’s position, direction, dispersion angle and range as a sound cone. The platform calculates range from rated and target sound levels and warns when the margin over ambient noise falls below 6 dB. The layout proposal places BDS units where only announcements are needed.

In operation, staff choose one or more areas – masts and BDS – and the languages. The platform offers German, English, Spanish, French and Italian; the standard messages come with hand-written translations, and your own translations can be reviewed and corrected. If a translation is missing, that language is skipped – German text is never read out in a foreign voice. Prepared messages are triggered by pressing and holding, so nothing starts by accident, and a stop button ends a running announcement everywhere.

Speech is generated without an internet connection on the event’s own server: sentence by sentence, with pauses and at a slightly slower pace. When an event starts, every selected message is produced in every language in advance, so in an emergency the audio is ready instantly and sounds exactly as it did in testing. A recorded message can be used instead of the synthetic voice. Ride sound systems can be connected so that safety announcements interrupt their music.

When the platform detects a critical situation, it suggests a suitable announcement – but nothing is played until a person approves it. The system suggests; a person decides. Every announcement is recorded automatically in the operations log with its text, languages, loudspeakers and the person responsible.

Checklist for your next event

• Assess the noise: what ambient level should you expect, and do you have a 6 dB margin – 10 dB for evacuation?
• Plan coverage: do loudspeakers reach every entrance, walkway and audience area at 75 dB or more?
• Avoid echoes: do any loudspeakers overlap with a path difference of more than about 17 metres?
• Define zones: which areas must be addressable on their own?
• Prepare messages: is there short, unambiguous wording for every typical situation?
• Agree languages: which languages does your audience need, and have the translations been checked?
• Interrupt the music: can stages and rides be lowered or muted for safety announcements?
• Test offline: does everything still work without the internet?
• Measure STI: has intelligibility been verified on site with STIPA?
• Keep the log: is every announcement recorded with a timestamp?

Conclusion

Intelligible announcements are made during planning, not during the incident: enough level above the noise, distributed loudspeakers without disruptive echoes, clear zones, prepared and checked messages in the audience’s languages – and equipment that keeps working without the internet. The STI target of 0.50 is the benchmark every installation should be measured against on site.