Guest post by Deniss Mironovs, acoustic consultant at Akukon / Researcher at RTU / PhD
Introduction
Speech Intelligibility is one of the main attributes of a well-functioning classroom or auditorium. It is widely known, that learning outcomes depend on the acoustical conditions of the classrooms. Common practice is to use reverberation time RT as a convenient and easily estimated metric. However, researchers pointed out that for non-diffuse spaces, such as modern classrooms with an absorptive ceiling, Sabine equation estimates do not fit empirical data. What is more, RT alone does not fully describe speech intelligibility, as RT is a global parameter, and speech intelligibility usually reduces with distance – the further away from the speaker you are, the worse the understanding of the spoken material.
This study proposes Speech Clarity C50 as a metric for simpler design of learning spaces with absorbent ceilings. C50 allows for validating whether speech intelligibility is enough, even in remote positions. It is generally understood by acoustical designers, that in order to calculate C50, one needs an impulse response of the room. Usual practice is to model the room in a 3D environment and perform simulations in Odeon, CATT, Treble and other room acoustics simulation software. For designers, this can be a setback, as the price of the design increases. Sometimes an analytical formula is enough to check the validity of the design. An empirical formula for C50 estimation based on distance for a room with absorbent ceilings is proposed in this study as an alternative practical tool for acoustical designers.
Measurements
The measurements were done in Riga Technical University auditoria with a regular rectangular shape. All rooms have Ecophon “Class A” acoustic ceiling tiles and “Class A” Akusto wall panels on the back wall, a conventional design for teaching premises. The room is furnished, providing a decent amount of scattering. All rooms have an average ceiling height of 2.66 m and a width 5.78 m with insignificant variation. The length of the rooms varies from 8.8 m to 27 m. The measurements were done according to ISO 3382-1. All source-receiver positions were measured, so that the distances between sources and receivers are known. In 8 of the 9 rooms, there were 3 sources and 5 to 10 individual receivers for each source. Only one room had a single sound source. In total, there are 181 separate measurements.
Results
The reverberation time T30 for all rooms varies between 0.4 and 0.7 seconds, depending on the room size. This variance is not at all significant. The T30 values themselves satisfy most national standards. Clarity, on the other hand, shows a very significant variance from 0 to 12 dB at 500 Hz and 1000 Hz octave bands across all rooms and all distances. The average absolute Pearson correlation coefficient for C50 and T30 at mid frequencies is around 0.50, whereas for C50 and d it was around 0.75. This shows that in already non-reverberant rooms, clarity is almost not dependent on reverberation time, but on the distance between the source and receiver.
Regression analysis
C50 values were analysed in relation to source–receiver distance d for mid frequencies (500–1000 Hz) using different regression models. Among the tested options, the 2nd order polynomial provided the most reliable and practical fit across frequency bands, making it the preferred choice over the 1st and 3rd order models. An average C50 model was then derived by combining coefficients across all octave bands, which aligned well with the measurements, showing only minor deviations at 500 Hz and 1000 Hz. This empirical model enables C50 to be estimated directly from distance in classrooms with ceiling and backwall absorption:
Concluding remarks
This study investigated the relationships between C50, T30, and source–receiver distance in university classrooms with absorptive surfaces. A clear negative correlation was found between C50 and distance, especially at mid to high frequencies, while T30 had less influence due to the low reverberation times. Overall, the results suggest that C50 can be reliably estimated from distance in such environments. Future work will expand the dataset, refine regression models, and include T30 as a predictor to further simplify classroom acoustic design without relying on 3D modelling.
One interesting artefact was observed – in the longest room of 27 m C50 at 125 Hz and 250 Hz increased slightly beyond 20 m. Other scholars also mentioned this effect.
Acknowledgements
This work has been supported by a postdoctoral grant No. RTU-PG-2024/1-0037 under the EU Recovery and Resilience Facility funded project No. 5.2.1.1.i.0/2/24/I/CFLA/003 “Implementation of consolidation and management changes at Riga Technical University, Liepaja University, Rezekne Academy of Technology, Latvian Maritime Academy and Liepaja Maritime College for the progress towards excellence in higher education, science, and innovation”.
Many thanks to Deniss for this guest post. Deniss is based in the Akukon Riga office in Latvia.
Other related articles about speech clarity:
