Speech Transmission Index (STI): What It Is and Why Your Building Might Fail

TL;DR

Speech Transmission Index (STI) is a 0-to-1 metric that quantifies how well speech modulation patterns survive the journey from talker to listener. Defined by IEC 60268-16:2020, it accounts for reverberation, background noise, and signal processing degradation. An STI below 0.50 means most listeners will struggle to understand speech — and for emergency PA systems, regulatory failure. Yet STI is routinely ignored during building design. Architects specify RT60 targets and material absorption coefficients but rarely verify that the resulting room will actually deliver intelligible speech at the back row. This article explains the physics behind STI, walks through the calculation method, and presents three case studies of buildings that failed speech intelligibility testing at commissioning — including a hospital where emergency announcements were unintelligible in 40% of ward areas.

The Hospital Where Nobody Could Hear the Fire Alarm

In 2023, a newly completed 280-bed hospital in the Midlands failed its BS 5839-8 voice alarm commissioning test. STIPA measurements across 340 test points showed that 38% of ward areas and 52% of corridor junctions fell below the 0.50 minimum. The voice alarm speakers had been specified and installed correctly. The problem was the building itself.

Ward ceilings were exposed concrete soffits with integrated services — a deliberate architectural choice for the industrial aesthetic that the trust's design brief demanded. These hard, parallel surfaces created reverberation times of 1.4-1.8 seconds at mid-frequencies. Combined with background noise from medical equipment and HVAC (measured at NC-42 in occupied conditions), the acoustic environment destroyed speech modulation patterns faster than the PA system could deliver them.

The remediation cost £340,000: 2,800 m² of suspended acoustic rafts, additional speakers to reduce throw distances, and a complete re-commissioning programme. Had STI been modelled during RIBA Stage 3, the acoustic treatment would have cost approximately £85,000 as part of the original ceiling package.

What STI Actually Measures

Speech carries information through temporal modulations — the patterns of loud and quiet syllables that encode meaning. When you speak the word "cat," the consonant onset (hard 'c'), the vowel ('a'), and the stop consonant ('t') create a specific amplitude modulation pattern. For a listener to understand the word, these modulation patterns must arrive intact.

Reverberation smears modulations by adding reflected energy that fills the gaps between syllables. Background noise masks modulations by reducing the signal-to-noise ratio. Together, they reduce the modulation depth at the listener position — and reduced modulation depth means reduced intelligibility.

STI quantifies this by measuring the Modulation Transfer Function (MTF) across seven octave bands (125 Hz to 8 kHz) and fourteen modulation frequencies (0.63 to 12.5 Hz). The MTF value m(F, fm) at each combination represents the fraction of original modulation depth that survives:

m(F, fm) = 1 / √(1 + (2π × fm × T/13.8)²) × 1 / (1 + 10^(-SNR/10))

Where T is RT60 at frequency F, fm is the modulation frequency, and SNR is the signal-to-noise ratio. Per IEC 60268-16:2020 §4.2, this indirect calculation method yields a close approximation to the full STI measurement.

STI Rating Scale

The Two Enemies of Speech Intelligibility

Enemy 1: Reverberation

Every 0.1-second increase in RT60 reduces STI by approximately 0.03-0.05 points (the relationship is nonlinear but this approximation holds for typical rooms). The mechanism is straightforward: longer reverberation means more reflected energy arriving late, which fills the temporal gaps between syllables and reduces modulation depth.

These values assume negligible background noise. In practice, background noise reduces STI further.

Enemy 2: Background Noise

Background noise reduces the signal-to-noise ratio (SNR), which directly reduces modulation depth. The critical threshold is approximately SNR = 15 dB — below this, STI degrades rapidly. A typical unamplified talker produces about 60 dB(A) at 1 metre. At 4 metres (typical classroom distance), the direct sound level drops to approximately 48 dB(A). If background noise is 40 dB(A) (NC-35), the SNR is only 8 dB — marginal for intelligibility.

Calculate STI for your room design → AcousPlan Calculator

STIPA: The Practical Measurement Method

Full STI measurement requires injecting a specially modulated test signal and measuring the received modulation depth at each frequency-modulation combination — 98 measurements per position. This is impractical for routine site testing.

STIPA (Speech Transmission Index for Public Address), defined in IEC 60268-16:2020 §5, uses a simplified test signal that modulates seven octave bands simultaneously with two modulation frequencies each (14 modulation frequencies total). A single 15-second measurement yields an STI value with ±0.03 accuracy. STIPA is the standard method for commissioning voice alarm systems per BS 7827:2011 and EN 54-16:2008.

STIPA Measurement Procedure

1. Set up the STIPA signal generator connected to the PA system amplifier input
2. Adjust output level to match the design speech level (typically 65-75 dB(A) at 1 m from speaker)
3. Position the STIPA analyser at the measurement point (typically 1.2-1.5 m above floor)
4. Record background noise level with PA system off
5. Play STIPA test signal for at least 15 seconds
6. Read STI value from analyser display
7. Repeat at each measurement point (grid spacing typically 2-3 m)

Three Case Studies: Buildings That Failed

Case 1: The Hospital (described above)

• Root cause: Hard parallel surfaces, high background noise (NC-42)
• STI range: 0.31-0.58 across ward areas
• Fix cost: £340,000 (acoustic rafts + additional speakers)
• Prevention cost: £85,000 (ceiling treatment in original fit-out)

Case 2: Multi-purpose school hall (Leeds, 2024)

• Root cause: 1,200 m³ volume with glazed walls and timber floor
• STI at rear: 0.38 (Poor)
• RT60: 2.2 seconds at 1 kHz
• BB93:2015 requires: STI ≥ 0.60 for speech rooms
• Fix: Wall-mounted absorptive panels (120 m² at £95/m²) + ceiling clouds (80 m²)
• Total fix cost: £23,600

Case 3: Open-plan office (Canary Wharf, 2022)

• Root cause: Not an STI failure — STI was too high (0.72)
• Problem: Speech privacy was non-existent; every phone call was audible 6 desks away
• ISO 3382-3 requires: Spatial decay rate D₂,S ≥ 5 dB for normal privacy
• Fix: Sound masking system (42 dB(A) pink noise) + desk screens
• Lesson: High STI is not always desirable. Open offices need controlled intelligibility — high enough for direct conversation, low enough for privacy at distance.

Designing for STI: The Practical Checklist

Step 1: Establish the STI target. For speech rooms (classrooms, meeting rooms, courtrooms), target STI ≥ 0.60. For PA/VA systems, target STI ≥ 0.55 minimum. For open-plan offices, target STI 0.45-0.55 at the workstation and below 0.40 at 8 m distance.

Step 2: Control reverberation. RT60 should not exceed 0.8 seconds for speech rooms under 500 m³. Use the Sabine or Eyring equation to verify.

Step 3: Control background noise. Target NC-30 or lower for classrooms (per ANSI S12.60-2010 §5). Target NC-35 for offices. Target NC-25 for courtrooms and lecture theatres.

Step 4: Verify critical distance. Ensure the farthest listener is within 1.5× the critical distance for unamplified speech, or provide reinforcement.

Step 5: Model STI. Use the indirect calculation method (RT60 + noise) for design-stage verification. AcousPlan calculates STI automatically from your room parameters.

Summary

STI is the single most important metric for any room where speech must be understood — which includes nearly every occupied building. It is more informative than RT60 alone because it integrates reverberation and noise into a single intelligibility prediction. Yet it remains absent from most architectural specifications, discovered only at commissioning when remediation is expensive. Model STI during design. Verify at commissioning. The cost of getting it wrong is measured in hundreds of thousands of pounds and, in emergency situations, human safety.

Check your room's STI score → AcousPlan Speech Intelligibility Calculator