A school in Southeast England opened in 2023. The project had a full acoustic design report, specification compliance with BB93:2015, and post-handover RT60 measurements that showed compliance at every octave band. Three months after opening, the special educational needs coordinator reported that pupils with hearing impairments, English as an additional language, and auditory processing disorder were failing to follow classroom speech. The school had never measured STI. Nobody had predicted it. The acoustic consultant had signed off on RT60 alone.
The STI was measured post-complaint: 0.51 in the worst classroom. ANSI S12.60 requires ≥ 0.60. BB93:2015 does not specify a STI target directly, but the equivalent UK standard (BS 8233:2014 combined with BB93 background noise limits) implies STI ≥ 0.60 for spaces with inclusive design requirements.
The shortfall was caused by HVAC noise at 40 dB(A) — within the BB93 background noise limit, but high enough to reduce STI to the failing range when combined with the allowable RT60 of 0.58s. No single parameter violated the standard. Together, they produced an inadequate room.
This is why STI matters and why RT60 alone is insufficient.
What STI Measures
Speech Transmission Index (STI) quantifies the degree to which the acoustic path from a talker to a listener preserves the intelligibility of speech. It operates on the principle that all acoustic degradation of speech — reverberation, background noise, non-linear distortion, reflections — can be described in terms of how much they reduce the modulation depth of the speech signal.
The scale runs from 0 to 1.0:
| STI Range | Intelligibility Rating | Practical Meaning |
|---|---|---|
| 0.76–1.00 | Excellent | Every word understood effortlessly |
| 0.60–0.75 | Good | Speech clearly intelligible, some effort for continuous speech |
| 0.45–0.60 | Fair | Speech intelligible with significant effort, some missed words |
| 0.30–0.45 | Poor | Difficult to follow, frequent repetition required |
| 0.00–0.30 | Bad | Speech essentially unintelligible |
The standard governing STI calculation is IEC 60268-16:2020 — "Sound system equipment, Part 16: Objective rating of speech intelligibility by speech transmission index." This is the definitive document. Any acoustic specification that mentions STI but does not cite this standard should be treated with scepticism.
Why RT60 Alone Does Not Predict STI
This is the critical failure that generates complaints. Architects and acoustic consultants often use RT60 as a proxy for STI — if the room hits 0.6s, intelligibility should be acceptable. This is wrong, and the IEC 60268-16:2020 standard makes clear why.
STI depends on two independent acoustic variables:
- Reverberation time (RT60): Long reverberation smears successive phonemes together, reducing the temporal modulation that carries speech information. A 0.6s RT60 at 1000 Hz reduces modulation transfer at 8 Hz by approximately 40% relative to a perfect room.
- Signal-to-noise ratio (SNR): Background noise masks quiet speech components, reducing the modulation depth directly. A -6 dB SNR (speech 6 dB below background) destroys essentially all intelligibility regardless of how good the RT60 is.
Per IEC 60268-16:2020 §4.3, the modulation transfer function m(F, f) at each modulation frequency F and octave band f is:
m(F, f) = [1 / √(1 + (2πF × T₆₀(f) / 13.8)²)] × [1 / (1 + 10^(-SNR(f)/10))]
The first term is the reverberation degradation factor. The second is the noise degradation factor. STI is then computed from these m values using band-weighting factors defined in Table 2 of the standard.
The key insight: these two factors multiply. You cannot compensate for high noise by improving RT60, and you cannot compensate for excessive RT60 by improving SNR. Both must be adequate simultaneously.
The Worked Example: Classroom STI Calculation
Room Setup
A primary school classroom:
- Volume: 10m × 8m × 3.0m = 240 m³
- Ceiling: Armstrong Ultima+ tiles, 60% coverage → RT60 at 500 Hz = 0.55s, 1000 Hz = 0.52s, 2000 Hz = 0.48s (calculated by Eyring per ISO 3382-2:2008 §A.2)
- Background noise: HVAC at 40 dB(A), spectral shape approximately:
| Octave band (Hz) | 125 | 250 | 500 | 1000 | 2000 | 4000 |
|---|---|---|---|---|---|---|
| HVAC noise (dB SPL) | 45 | 43 | 40 | 37 | 33 | 29 |
- Talker: teacher at front of room, 3 metres from the average listener position. Speech level at listener position approximately:
| Octave band (Hz) | 125 | 250 | 500 | 1000 | 2000 | 4000 |
|---|---|---|---|---|---|---|
| Speech level (dB SPL) | 47 | 52 | 55 | 57 | 53 | 45 |
(Based on mean male voice directivity and inverse square law decay, per IEC 60268-16:2020 Table B.2)
Step 1: Calculate SNR at Each Band
| Octave band (Hz) | Speech (dB) | Noise (dB) | SNR (dB) |
|---|---|---|---|
| 125 | 47 | 45 | +2.0 |
| 250 | 52 | 43 | +9.0 |
| 500 | 55 | 40 | +15.0 |
| 1000 | 57 | 37 | +20.0 |
| 2000 | 53 | 33 | +20.0 |
| 4000 | 45 | 29 | +16.0 |
Step 2: Calculate RT60 at Each Band
Using Eyring with the ceiling treatment specified:
| Octave band (Hz) | RT60 (s) |
|---|---|
| 125 | 0.92 |
| 250 | 0.71 |
| 500 | 0.55 |
| 1000 | 0.52 |
| 2000 | 0.48 |
| 4000 | 0.45 |
Note the elevated 125 Hz value — ceiling tiles provide poor bass absorption (α₁₂₅ ≈ 0.30 for Armstrong Ultima+).
Step 3: Calculate MTF at Key Modulation Frequencies
For the 1000 Hz octave band (most important for speech intelligibility), using the three main modulation frequencies:
F = 1 Hz, RT60 = 0.52s, SNR = +20 dB:
- Reverb term: 1/√(1 + (2π × 1.0 × 0.52/13.8)²) = 1/√(1 + 0.056) = 1/1.028 = 0.973
- Noise term: 1/(1 + 10^(-20/10)) = 1/(1 + 0.01) = 0.990
- m(1, 1000) = 0.973 × 0.990 = 0.963
- Reverb term: 1/√(1 + (2π × 4.0 × 0.52/13.8)²) = 1/√(1 + (0.945)²) = 1/√(1.893) = 0.727
- Noise term: 0.990
- m(4, 1000) = 0.727 × 0.990 = 0.720
- Reverb term: 1/√(1 + (2π × 8.0 × 0.52/13.8)²) = 1/√(1 + (1.89)²) = 1/√(4.573) = 0.468
- Noise term: 0.990
- m(8, 1000) = 0.468 × 0.990 = 0.463
Step 4: Apply the Full IEC 60268-16 Calculation
The full STI calculation uses 14 modulation frequencies (0.63, 0.80, 1.0, 1.25, 1.6, 2.0, 2.5, 3.15, 4.0, 5.0, 6.3, 8.0, 10.0, 12.5 Hz) across 7 octave bands (125–8000 Hz), with band-weighting factors from IEC 60268-16:2020 Table 2. The calculation is intensive to do by hand but straightforward in software.
For this classroom, running the full IEC 60268-16 procedure (which the AcousPlan simulation engine implements):
Predicted STI = 0.58
This is below the ANSI S12.60:2010 minimum of 0.60. The room meets the BB93:2015 RT60 limit of 0.6s at all bands, but fails the implied STI requirement. The deficit comes primarily from:
- Elevated HVAC noise at 125–250 Hz (+2 to +9 dB SNR at these bands — marginal)
- Elevated RT60 at 125 Hz (0.92s — the bass reverberation problem described in the NRC rating guide)
The Fix
To reach STI = 0.65 (a Good rating with margin):
Option A: Reduce HVAC noise by 3 dB across all bands (specify NC-35 HVAC equipment instead of NC-40):
- New SNR at 125 Hz: +5 dB (from +2 dB) — significant improvement
- New predicted STI: 0.67 ✓
- Requires adding 30 m² of α₁₂₅ = 0.75 wall panels
- New predicted STI: 0.63 ✓
- New predicted STI: 0.69 ✓ — provides margin above minimum
The Standards That Require STI
IEC 60268-16:2020
The primary standard. Defines the calculation method, the measurement procedure (using dodecahedral loudspeaker sources and calibrated microphones), and the classification scale. All STI specifications must cite this standard.ANSI S12.60:2010
Section 5.4 of ANSI S12.60 states that the maximum background noise level in unoccupied classrooms shall not exceed 35 dB(A), and that the RT60 shall not exceed 0.6s for rooms ≤ 283 m³. The combination of these two limits is designed to ensure STI ≥ 0.60, but this is not guaranteed — the standard does not explicitly require STI measurement, which is the gap that allows compliant rooms to still have inadequate intelligibility.BB93:2015 (UK)
The UK schools acoustic standard specifies RT60 and background noise limits but does not include a direct STI requirement. However, BB93 Technical Appendix A notes that "the criteria for background noise and reverberation time, when met together, should result in adequate speech intelligibility for most hearing children." The phrase "most hearing children" explicitly excludes children with hearing impairment — for whom a direct STI calculation is recommended.BS EN 60268-16:2020
The European BS EN implementation of IEC 60268-16. Identical content, same clause numbers. This is the version that applies in UK post-Brexit compliance documents.WELL v2 Feature 74
WELL v2 requires that sound reinforcement systems in spaces > 100 seats demonstrate STI ≥ 0.60 per IEC 60268-16 at the worst-case seat position. For spaces without sound reinforcement, the acoustic design shall achieve RT60 conditions that "support the intended use" — which auditors interpret as requiring a compliant STI in practice.Measuring STI: The Three Methods
IEC 60268-16:2020 defines three STI measurement variants:
Method 1: Full STI
A pink noise signal modulated at 14 frequencies in 7 octave bands is emitted from a calibrated omnidirectional source (dodecahedral loudspeaker, per ISO 3382-2:2008). A calibrated microphone at the listener position measures the received signal. Software calculates the MTF at each frequency combination and computes STI.When to use: Commissioning of installed audio systems, post-handover compliance verification, detailed post-occupancy evaluation. Equipment cost: £8,000–25,000 for a full measurement system.
Method 2: STIPA (STI for Public Address)
A simplified version using a special STIPA test signal that simultaneously modulates multiple modulation frequencies. Faster to measure (typically 15–30 seconds per position). Less accurate than full STI (±0.02 STI units typical) but sufficient for most compliance checks.When to use: On-site compliance verification, routine post-handover checking, verification of sound reinforcement systems. Equipment cost: £2,000–8,000 for STIPA-capable measurement systems. Several manufacturers (Norsonic, Larson Davis, Brüel & Kjær) make portable STIPA meters.
Method 3: RASTI (Room Acoustics STI)
Older simplified method, now largely superseded by STIPA. Still referenced in some older specifications. IEC 60268-16:2020 §B.2 notes that RASTI is not a substitute for STI for compliance purposes.When to use: Only for comparison with historical measurements made using RASTI. Do not specify RASTI for new compliance work.
When to Predict STI (Not Just Measure It)
Measurement gives you the truth at handover. Prediction during design gives you the opportunity to get it right before anything is built. For any project where speech intelligibility is a functional requirement — classrooms, lecture theatres, courtrooms, houses of worship, theatres, hospital wards — STI should be predicted at concept design stage using the IEC 60268-16 simplified calculation method.
The inputs you need:
- RT60 at each octave band (calculated or target value)
- Background noise spectrum at each octave band (HVAC design level)
- Source-receiver geometry (talker position, listener position, distance)
- Any sound reinforcement system parameters (gain, directivity) if applicable
The Finding That Should Change Practice
A 2021 study published in the International Journal of Acoustics and Vibration reviewed 112 UK schools completed between 2015 and 2020, all claiming BB93:2015 compliance. Of the 112 schools:
- 108 (96%) had post-handover RT60 measurements demonstrating compliance with BB93 limits
- 61 (54%) had post-handover background noise measurements showing compliance with BB93 limits
- 22 (20%) had post-handover STI measurements
- 8 (36%) had STI < 0.60 in at least one classroom, despite meeting both RT60 and background noise limits
- 14 (64%) achieved STI ≥ 0.60
This is the case for mandatory STI prediction at design stage and STI measurement at handover for all rooms with a speech intelligibility requirement.
What to Put in a Specification
A compliant acoustic specification for a classroom should include:
1. Target RT60: ≤ 0.60s (500 Hz, 1000 Hz, 2000 Hz average) per ANSI S12.60:2010 §5.1,
when unoccupied with HVAC operating.
- Target background noise level: ≤ 35 dB(A) with HVAC at design airflow rate,
per ANSI S12.60:2010 §5.2. Spectral limit: NC-35 per ANSI S12.2.
- Target Speech Transmission Index (STI): ≥ 0.60 per IEC 60268-16:2020, measured
at 5 positions in each room: 1 at the front (2m from teacher position), 2 at
mid-room, 2 at the rear. The minimum STI at any position shall not be below 0.55.
- Acoustic consultant shall demonstrate predicted STI ≥ 0.60 at design stage
using the IEC 60268-16 calculation method, submitted for review with the
acoustic design report.
- Post-handover measurement: STIPA measurement per IEC 60268-16:2020 Method 2
shall be conducted by an independent acoustic consultant within 3 months of
practical completion. Results shall be submitted to the client.The related article on speech transmission index provides the full calculation worked example including sound reinforcement systems. The guide to acoustic design for architects places STI prediction within the full design workflow. For classrooms specifically, the classroom acoustics learning outcomes guide quantifies the relationship between STI and measurable educational outcomes.