Consulting services / PAGA and voice alarm testing

PAGA and voice alarm speech intelligibility testing (STIPA)

Geonoise Asia measures whether people can understand the announcements from public address and general alarm (PAGA), public address and voice alarm (PA/VA) and fire-alarm evacuation (EVAC) systems. We carry out STIPA measurements to IEC 60268-16 on the live system, check alarm-tone audibility against plant noise, and report the results position by position against your project's criterion. The work covers offshore and onshore oil and gas, petrochemical, power and mining sites, and transit, airport, hospital and high-rise projects across South-East Asia.

What we test

• STIPA measurement on live systems. Speech Transmission Index measured with the STIPA method on live PA/VA and PAGA systems, indoors and outdoors, at the positions where people listen. Background noise is measured and accounted for as IEC 60268-16 describes, and results are given position by position against the criterion in your project specification.
• Alarm audibility. The sound level of alarm tones measured against ambient noise in noisy plant areas, to show whether the alarm stands far enough above the background for the standard the project follows.
• Design-stage prediction and review. Prediction or review of loudspeaker layouts for speech intelligibility and coverage before the system is installed, when changes cost least.
• Diagnosis and retest. When a system fails, we establish why (reverberation, background noise, coverage, delays and echoes between loudspeakers, or gain settings) and retest after the contractor's corrections.
• Reporting. A report suitable for handover, client acceptance and review by the authority or insurer.

Why independent verification matters

A contractor who designs, supplies and commissions a PA/VA system is also the party that benefits when it passes. An independent measurement removes that conflict. Because Geonoise has no product to sell and no installation to defend, the owner, the main contractor and the system supplier can rely on the same result, and a failing position is reported as it was measured.

We do not recommend suppliers or products. Corrective measures are described in terms of what has to change acoustically; the system contractor decides how to achieve it.

Where speech intelligibility testing is needed

• Offshore and onshore oil and gas: PAGA coverage of process and utility areas, compressor and generator rooms, accommodation, walkways and muster points.
• Petrochemical plants and refineries: process units, pump and compressor houses, control buildings and workshops.
• Power stations and mining: turbine halls, boiler and crusher areas, conveyors and maintenance workshops.
• Rail, metro and transit: platforms, concourses and tunnels. See also our railway and MRT commissioning test standards.
• Airports: terminals, gate lounges and baggage halls.
• Hospitals and high-rise buildings: voice alarm and phased evacuation (EVAC) systems in stair cores, corridors, atria and car parks.

The difficult positions tend to share the same features: high background noise from running equipment, long reverberation in hard-surfaced halls and enclosures, and areas reached only by distant loudspeakers whose sound arrives late.

Standards and criteria

Which document governs depends on the project specification, the national code and, offshore, the operator's own requirements. The documents we most often test against:

Voice-alarm criteria in these documents commonly centre on an STI of 0.50, with individual positions not below about 0.45 — the governing figure is the one your project specification names.

For alarm tones the requirement is a margin above ambient noise, commonly 10 to 15 dB depending on the standard the project follows.

How a STIPA test is carried out

1. Agree the criterion and the positions. We take the target from the project specification and select positions where people will listen: work areas, walkways, muster points, platforms, corridors and rooms. Positions with the highest noise and the furthest loudspeakers are included, not only the easy ones.
2. Play the STIPA test signal through the system itself. The modulated test signal is fed in at the input that live or recorded announcements use, so the result includes the whole chain: amplifiers, loudspeakers and the room or plant area around the listener.
3. Measure STIPA at each listener position with a Class 1 analyser, repeating readings where they scatter.
4. Measure background noise at the same positions under representative operating conditions, and account for it as IEC 60268-16 describes.
5. Check alarm-tone audibility where the system also sounds alarm tones.
6. Report position by position against the criterion: STI value, qualification band, pass or fail, and the probable cause at each failing position.

The background to the method is in our guide STI and STIPA explained.

When a system fails

Speech intelligibility depends on the signal-to-noise ratio at the listener, reverberation, echoes and late arrivals from distant loudspeakers, loudspeaker coverage, and distortion in the system. A low result usually traces back to one or two of these, and the measured data shows which:

• Background noise: announcements too quiet relative to the plant or crowd noise at the listener.
• Reverberation: hard, enclosed spaces such as halls, stairwells, tunnels and modules where sound decays slowly and smears speech.
• Echoes and delays: sound from distant loudspeakers arriving noticeably later than from the nearest one.
• Coverage: positions between, behind or too far from loudspeakers.
• Gain, level and distortion: amplifier and zone settings, clipping or equalisation.

We describe what has to change acoustically at each failing position. The system contractor makes the corrections, and we retest the same positions so the results before and after can be compared directly.

Instruments, calibration and the status of our reports

Measurements are made with Class 1 sound level meters and analysers capable of STIPA, calibrated by an ISO/IEC 17025 accredited laboratory. Geonoise itself is not ISO/IEC 17025 accredited: our reports are independent engineering reports, not accredited test reports. Each report contains:

• the criterion and the document it comes from;
• the measurement positions, on a plan or in a list;
• the STI result and qualification band at each position, with the background noise;
• alarm audibility results, where measured;
• pass or fail per position, and the probable cause of each failure;
• instrument and calibration details;
• retest results after corrections, where applicable.

Where we work

We work from our offices in Bangkok, Kuala Lumpur and Jakarta, and deliver projects in Vietnam, India and the Philippines with local partner firms.

Frequently asked questions

What is STIPA, and how is it different from STI?

The Speech Transmission Index (STI) rates speech intelligibility on a scale from 0 to 1, as defined in IEC 60268-16. STIPA is the simplified method in that standard for public address systems: a modulated test signal is played through the system itself and analysed at each listener position. Full STI and indirect methods based on a measured impulse response also exist, but STIPA is the method normally used to commission and verify PA/VA and PAGA systems.

What STI does a voice alarm or PAGA system have to achieve?

Voice-alarm criteria in these documents commonly centre on an STI of 0.50, with individual positions not below about 0.45 — the governing figure is the one your project specification names. On offshore and industrial projects the figure may come from the operator's own specification, so we confirm the criterion before measuring.

Can STIPA be measured in a running plant or on an offshore facility?

Yes. Background noise is measured at each listener position under representative operating conditions and accounted for as IEC 60268-16 describes, so the result reflects the noise the system actually has to work against. Testing only in a quiet, shut-down plant can give results that do not hold once the equipment is running.

The alarm is loud enough. Why would speech still fail?

Audibility and intelligibility are different tests. An alarm tone only has to stand out above the ambient noise, with a margin above ambient noise, commonly 10 to 15 dB depending on the standard the project follows. Speech also has to survive reverberation, late sound from distant loudspeakers, gaps in coverage and distortion in the system, so a system can pass the audibility check and still fail on STI.

Do you supply or install PA/VA systems?

No. Geonoise is an independent acoustic consultancy. We do not sell, supply or install PA/VA systems, loudspeakers or instruments, and we do not recommend suppliers. That is what allows us to witness and verify the contractor's system impartially, at commissioning, at client acceptance or in a dispute.

Is your report an accredited test report?

No. Our reports are independent engineering reports. The measurements are made with Class 1 sound level meters and analysers capable of STIPA, calibrated by an ISO/IEC 17025 accredited laboratory, but Geonoise itself is not ISO/IEC 17025 accredited. The report states the method, positions, conditions, instruments and calibration so that a client, authority or insurer can review how each result was obtained.

What happens if the system fails?

We use the measured data to find the cause at each failing position: background noise, reverberation, echoes or delays between loudspeakers, coverage, or gain and level settings. The system contractor makes the corrections, and we retest the same positions so the results before and after can be compared directly.

Which countries do you cover?

We work from our own offices in Thailand, Malaysia and Indonesia, and deliver projects in Vietnam, India and the Philippines with local partner firms.

Request a test or a review

Send the project specification, the type of system and the areas to be tested through the enquiry form or by e-mail to inquiries@geonoise.asia.

Related: industrial noise · noise prediction for oil and gas, power and process plants · why noise measurements fail when they matter