Question 1

What STI score is considered acceptable for speech intelligibility?

Accepted Answer

IEC 60268-16 defines five qualitative categories: Excellent (>0.75), Good (0.60-0.75), Fair (0.45-0.60), Poor (0.30-0.45), and Bad (<0.30). For most applications, a minimum STI of 0.50 (Fair/Good boundary) is the practical target. Voice alarm systems (EN 54-24) require STI of at least 0.50 in all areas. Transport hubs and critical communication environments typically require 0.50-0.60. Classrooms and conference rooms should achieve 0.60 or above. Performance spaces for unamplified speech need 0.65+. The STI value represents the percentage of syllables that would be correctly understood: STI of 0.50 corresponds to roughly 80% syllable recognition. SonaVyx displays the qualitative rating alongside the numeric STI value.

Question 2

What STI does a voice alarm system need to achieve?

Accepted Answer

Voice alarm systems are governed by EN 54-24 (fire detection and alarm components for voice alarm) and BS 5839-8 (fire detection code of practice for voice alarm). The typical requirement is STI of at least 0.50 in all areas where voice messages must be intelligible. Some authorities require 0.45 as an absolute minimum with 0.50 as the average. The measurement must be performed with the ambient noise conditions representative of normal occupancy. In areas with high background noise (plant rooms, kitchens), a higher STI threshold or louder system output may be needed. Note that voice alarm testing also requires a minimum SPL of 65 dBA or 5 dB above ambient (whichever is greater). SonaVyx tracks both STI and SPL simultaneously during measurement to verify both criteria.

Question 3

What is CIS and how does it relate to STI?

Accepted Answer

CIS (Common Intelligibility Scale) is an alternative metric derived from STI that maps nonlinearly to a 0-1 scale. The formula is CIS = 1 + log10(STI). CIS was developed to provide a more intuitive scale where equal increments correspond to equal perceptual changes in intelligibility. However, CIS has not gained wide adoption and most standards and specifications use STI directly. The mapping: STI 0.50 = CIS 0.69, STI 0.60 = CIS 0.78, STI 0.70 = CIS 0.85, STI 0.80 = CIS 0.90. The main use of CIS is in some European military and NATO specifications. SonaVyx calculates and displays both STI and CIS values so you can report whichever is required by your project specification.

Question 4

How does background noise affect STI scores?

Accepted Answer

Background noise reduces STI by masking the modulated speech signal. The effect depends on the signal-to-noise ratio (SNR) at the listener position. As a rough guide: SNR above 25 dB means noise has negligible effect on STI. SNR of 15-25 dB causes moderate STI reduction (0.05-0.15 reduction). SNR of 5-15 dB causes significant STI reduction (0.15-0.30 reduction). SNR below 5 dB makes intelligibility poor regardless of room acoustics. The noise effect is frequency-dependent: low-frequency noise (HVAC rumble) affects the 125-250 Hz bands most, while high-frequency noise (ventilation hiss) affects the 2-8 kHz bands. SonaVyx ambient noise correction applies the measured noise spectrum to each octave band independently, giving an accurate STI that accounts for the spectral character of the noise.

Question 5

What is the relationship between RT60 and STI?

Accepted Answer

RT60 and STI are inversely related: longer reverberation reduces intelligibility. As a rough approximation, STI ≈ 0.9 × exp(-0.6 × RT60) for an ideal system with no noise. At RT60 = 0.5 s, STI ≈ 0.67 (Good). At RT60 = 1.0 s, STI ≈ 0.49 (Fair). At RT60 = 2.0 s, STI ≈ 0.27 (Poor). However, this simplified relationship assumes no background noise and uniform reverberation. In practice, noise often has a larger impact than reverberation, especially in industrial or transportation environments. A room with RT60 of 0.8 s and 55 dBA background noise may have lower STI than a room with RT60 of 1.2 s and 35 dBA background noise. SonaVyx STI prediction model uses both measured RT60 and noise spectrum for accurate estimation.

Question 6

Can STI vary significantly across different positions in the same room?

Accepted Answer

Yes, STI can vary by 0.10-0.20 or more across different listener positions. Positions close to a speaker and away from reflective surfaces tend to have the highest STI. Positions far from speakers, under balconies, or in reverberant areas have lower STI. Coverage gaps where speakers do not direct sufficient energy result in poor SNR and low STI. Positions near hard rear walls may have strong late reflections that reduce STI. The spatial variation is why standards require measurements at multiple positions and typically specify the minimum STI found across all positions as the criterion for pass/fail. SonaVyx records STI per position and highlights the worst-case location, which is the most critical for system design verification.

Question 7

How is STI affected by loudspeaker distortion?

Accepted Answer

Loudspeaker distortion reduces STI by adding energy that is not correlated with the speech modulation pattern. The effect depends on the type and amount of distortion. THD of less than 5% has negligible effect on STI. THD of 5-15% may reduce STI by 0.02-0.05. THD above 15% can significantly degrade STI. Clipping distortion (from an overdriven amplifier) is particularly harmful because it flattens the waveform peaks, directly reducing the modulation depth. Intermodulation distortion redistributes energy across frequencies, smearing the spectral envelope. In practice, system distortion is rarely the dominant factor — reverberation and noise usually have much larger effects. The STIPA method captures distortion effects because the test signal passes through the entire electroacoustic chain. SonaVyx THD analyzer can be used alongside STI measurement to assess distortion contribution.

Question 8

What is the difference between STI and %ALcons?

Accepted Answer

%ALcons (Percentage Articulation Loss of Consonants) was the predecessor to STI, developed by Peutz in the 1970s. It estimates the percentage of consonants that would be lost (misheard) in a given acoustic environment. Lower %ALcons is better: below 5% is excellent, 5-10% is good, 10-15% is fair, above 15% is poor. %ALcons can be estimated from RT60 and the direct-to-reverberant ratio using empirical formulas. STI replaced %ALcons because it is more comprehensive (accounts for noise, distortion, and frequency-dependent effects) and is directly measurable. An approximate conversion: STI ≈ 0.9 - 0.3 × log10(%ALcons). SonaVyx uses the modern STI metric per IEC 60268-16 but can display an approximate %ALcons equivalent for comparison with legacy specifications.

STI Score Interpretation: What Do the Numbers Mean for Speech Intelligibility?

Frequently Asked Questions