IEC 60268-16: Measuring Speech Intelligibility (STI/STIPA)

TLDR: What STI Measures and Why It Matters

The Speech Transmission Index (STI) is the internationally accepted metric for quantifying how well a room and its sound system deliver intelligible speech to listeners. Defined in IEC 60268-16:2020, STI produces a single number between 0 and 1 that indicates the proportion of speech information preserved between talker and listener. An STI of 0.75 means speech is "excellent" — virtually every word is understood. An STI of 0.30 means "bad" — listeners catch fewer than half the sentences and must guess context constantly.

STI matters because speech intelligibility is the primary acoustic performance criterion in most occupied buildings. Classrooms need it for learning outcomes. Hospitals need it to prevent medication errors from misheard instructions. Transport hubs need it for safety announcements. Courts need it for legal proceedings. And open-plan offices need to control it — not maximise it — to maintain speech privacy between workstations.

The standard defines two measurement methods: full STI (98 modulation transfer values, 7 octave bands, 14 modulation frequencies) and STIPA (14 modulation values, simplified for field use on PA/VA systems). Both methods quantify how the room's reverberation and background noise degrade the amplitude modulations that carry speech information. This guide covers the measurement physics, practical field procedures, equipment requirements, and the interpretation errors that lead to incorrect pass/fail decisions.

The Field Story: When Nobody Can Hear the Announcements

In 2024, a new international terminal at a major Southeast Asian airport opened to immediate complaints about public address intelligibility. Passengers could not understand gate change announcements, boarding calls, or safety information. The terminal had been designed by a leading international architecture practice and featured a dramatic vaulted ceiling reaching 28 metres at the apex, exposed structural steel, full-height curtain walling, and polished stone floors — a visually stunning space that was an acoustic disaster.

Post-opening acoustic measurements revealed the problem in stark numbers. Reverberation time measured RT60 3.2 seconds at 500 Hz and 4.1 seconds at 1 kHz — typical of a large reverberant volume with minimal absorption. STI measurements using the installed PA system showed values ranging from 0.32 to 0.42 across the terminal seating areas, categorised as "poor" per IEC 60268-16 Table 2. At gate areas furthest from the nearest loudspeaker cluster, STI dropped to 0.28 — "bad." Background noise from HVAC, crowd babble, and retail units measured 62 dBA, further degrading the signal-to-noise ratio.

The PA system itself was well-designed — high-quality line arrays with digital signal processing, directional control, and adequate SPL coverage. The problem was not the electroacoustics; it was the room acoustics. No PA system can deliver intelligible speech when the room adds 3.2 seconds of reverberation to every syllable. Each word smears into the next, and the direct-to-reverberant ratio collapses beyond 10 metres from the loudspeaker.

Remediation involved installing 3,200 m2 of suspended acoustic baffles in the roof void and 800 m2 of acoustic wall panels on the check-in hall back wall. Cost: USD $2.8 million. The terminal partially closed for eight weeks during installation. Post-remediation measurements showed RT60 reduced to 1.6 seconds and STI improved to 0.52-0.58 — "fair" to borderline "good." Still below the BS 7827 target of 0.50 minimum (met) and the STIPA 0.60 preferred target (not met at all positions), but a transformative improvement in passenger experience.

The acoustic consultant's original design report had recommended a maximum RT60 of 1.5 seconds and 5,000 m2 of ceiling absorption. The architect rejected the recommendation as incompatible with the design vision. The consultant's STI prediction of 0.38 at worst-case seats was documented but overruled. The airport authority did not understand what "STI 0.38" meant until passengers started missing flights.

The Modulation Transfer Function: How STI Works

STI is based on the modulation transfer function (MTF), which describes how a room and its sound system preserve or degrade the amplitude modulations in speech. Human speech consists of rapid amplitude fluctuations — syllables, consonants, pauses — that carry the information content. A room with long reverberation "fills in" the pauses between syllables, reducing modulation depth. Background noise masks the quieter parts of the speech signal, also reducing modulation depth.

The MTF is measured or calculated at 14 modulation frequencies (0.63, 0.80, 1.00, 1.25, 1.60, 2.00, 2.50, 3.15, 4.00, 5.00, 6.30, 8.00, 10.0, 12.5 Hz) across 7 octave bands (125, 250, 500, 1000, 2000, 4000, 8000 Hz), producing a 14 x 7 matrix of 98 values. Each value m(F,f) represents the reduction in modulation depth at modulation frequency F in octave band f.

For a room with reverberation time T and signal-to-noise ratio S/N (in dB), the MTF can be estimated theoretically:

m(F,f) = [1 / sqrt(1 + (2 pi F T/13.8)^2)] [1 / (1 + 10^(-S/N / 10))]

The first term captures reverberation degradation; the second captures noise masking. The 98 MTF values are then converted to apparent signal-to-noise ratios, averaged across bands with speech-weighting factors, and mapped to the 0-1 STI scale.

The STI Intelligibility Categories

IEC 60268-16:2020 Table 2 maps STI values to qualitative categories:

These categories correlate with the Common Intelligibility Scale (CIS) and with subjective listening tests. The transitions are not sharp — an STI of 0.59 and 0.61 are perceptually almost identical — but the category boundaries are widely used in specifications and building codes.

Calculate Now: Use AcousPlan's free calculator to predict STI from your room's reverberation time and background noise levels before construction.

STIPA: The Field Measurement Method

Full STI measurement requires specialised equipment and careful procedure. STIPA (Speech Transmission Index for Public Address systems) is the simplified variant designed for practical field measurement of installed PA/VA systems. Defined in IEC 60268-16:2020 Section 5, STIPA uses a specially modulated test signal containing two modulation frequencies per octave band (14 total), allowing simultaneous measurement across all bands in approximately 15-20 seconds.

Equipment Requirements

• STIPA signal source: A standardised test signal (available as WAV file or generated by the measurement instrument). The signal sounds like modulated noise. It is played through the PA system under test.
• STIPA analyser: A measurement microphone and analyser that demodulates the received signal and calculates the modulation transfer values. Dedicated instruments include the NTi Audio XL2, Bedrock Audio SM90, and Gold Line TEF25. Software solutions running on calibrated measurement microphones are also available.
• Calibrated microphone: Class 1 per IEC 61672 recommended, Class 2 acceptable for survey-grade measurements.

Measurement Procedure

1. System configuration: Set the PA system to its normal operating condition — normal gain, EQ, delay settings. Do not use special "test" modes.
2. Source signal: Route the STIPA test signal into the PA system at normal speech level (approximately 65-70 dBA at 1 m from the loudspeaker).
3. Measurement positions: Measure at representative listener positions throughout the coverage area. Minimum positions depend on the standard being applied (BS 7827 specifies measurement at the furthest listener position from each loudspeaker zone). Microphone at 1.2 m height (seated ear height) or 1.5 m (standing).
4. Averaging: Each measurement takes 15-20 seconds. Take at least 3 measurements per position and average. Ensure no transient noise events (doors slamming, phones ringing) during measurement.
5. Background noise: Measure background noise levels (no signal) at each position in octave bands. Background noise should be representative of normal operating conditions — HVAC running, but not crowd noise for an unoccupied test.

Occupied vs Unoccupied Measurement

A critical consideration is whether to measure occupied or unoccupied. People absorb sound and reduce reverberation time, which improves STI. But people also generate babble noise, which degrades STI. The net effect depends on the space type:

Most standards specify measurement in unoccupied conditions with HVAC running. If the specification requires occupied STI, this must be stated explicitly and the measurement conditions documented.

STI in Room Design: Predicting Before Building

While IEC 60268-16 defines the measurement method, acoustic designers need to predict STI at the design stage. The standard provides the theoretical MTF formula (Section 4.2) that enables calculation from:

• Room volume and surface areas (to predict RT60 via Sabine/Eyring)
• Absorption coefficients at each octave band
• Background noise spectrum (from HVAC design or site survey)
• Source-receiver distance and room geometry

1. Calculate RT60 at 125, 250, 500, 1000, 2000, 4000, 8000 Hz using Sabine or Eyring equations per ISO 3382-2.
2. Estimate background noise levels at each octave band from mechanical services design data or site measurement.
3. Define source level (human speech: approximately 60 dBA at 1 m for normal voice, per ANSI S3.5).
4. Calculate signal-to-noise ratio at the receiver position accounting for distance attenuation.
5. Apply the MTF formula at all 98 frequency-modulation combinations.
6. Weight and average per IEC 60268-16 Section 4.3 to obtain predicted STI.

Applications and Target Values

Different building types and codes require different STI levels:

Note that open-plan offices invert the requirement: STI should be low enough to prevent overhearing conversations at adjacent workstations. This is achieved through a combination of masking sound systems, absorbent ceilings, and screens — the opposite of the strategies used to maximise STI in classrooms.

The Relationship Between RT60 and STI

For rooms without electronic sound reinforcement (natural speech only), reverberation time is the dominant factor affecting STI. The following table shows approximate STI values for natural speech in rooms with different RT60 values and background noise levels:

These values assume normal voice level (60 dBA at 1 m) and 4 m source-to-receiver distance in a diffuse sound field. Two important observations:

First, at low background noise (30 dBA), RT60 dominates. Reducing RT60 from 2.0 to 0.6 seconds improves STI by 0.30 — a shift from "fair" to "good/excellent."

Second, at high background noise (50 dBA), both RT60 and noise matter. Even with excellent RT60 of 0.4 seconds, STI only reaches 0.60 if background noise is 50 dBA. Noise control is as important as absorption treatment in many real buildings.

Common Mistakes

Mistake 1: Measuring STIPA with the PA system at maximum volume. STI is signal-to-noise ratio dependent. Turning the PA system to maximum increases the signal but creates non-linear distortion, which artificially degrades modulation depth and reduces measured STI. Always measure at normal operating level.

Mistake 2: Ignoring masking correction factors. IEC 60268-16 Section 4.4 defines auditory masking corrections that account for the upward spread of masking in the human auditory system. Software that omits these corrections will overestimate STI by 0.02-0.05. Ensure your measurement instrument or calculation tool applies auditory masking per the standard.

Mistake 3: Measuring in an empty room and claiming compliance. If the specification requires STI with typical background noise (e.g., HVAC at normal operating levels), measuring in a completely silent building at night is not valid. HVAC systems must be running at design operating conditions during measurement.

Mistake 4: Confusing STI with STIPA accuracy limits. STIPA agrees with full STI within approximately plus or minus 0.03 for typical rooms and PA systems. However, for rooms with strong discrete echoes or for systems with time-variant processing (certain digital effects, feedback suppressors), STIPA can deviate by up to plus or minus 0.10. Use full STI measurement in these special cases.

Mistake 5: Treating STI 0.60 as a cliff edge. STI categories are ranges, not thresholds. There is no perceptual difference between STI 0.59 and 0.61. When a measurement returns 0.58 against a 0.60 target, the appropriate response is to consider measurement uncertainty (typically plus or minus 0.03 for STIPA) rather than declaring outright failure. IEC 60268-16 Annex E discusses measurement uncertainty.

Summary

IEC 60268-16 provides the definitive framework for measuring and predicting speech intelligibility. The STI metric distils the complex interaction of room acoustics, background noise, and sound system performance into a single actionable number. For designers, the standard enables prediction of intelligibility from room acoustic parameters. For commissioning engineers, STIPA provides a practical field measurement method. For building owners and operators, STI categories translate technical measurements into meaningful quality assessments.

The airport terminal case demonstrates the cost of ignoring STI predictions: USD $2.8 million in remediation, eight weeks of disruption, and immeasurable passenger frustration. The acoustic consultant's prediction of STI 0.38 was documented but overruled by architectural vision. The eventual measured STI of 0.32-0.42 confirmed the prediction almost exactly.

Start your acoustic design by defining the STI target, then work backward to the RT60 and background noise limits that will achieve it. AcousPlan's acoustic calculator automates this prediction from room dimensions and material selections, giving you STI estimates before a single material is specified.

Predict speech intelligibility now: Use AcousPlan's free STI calculator to model your room and verify compliance with IEC 60268-16 requirements.