Skip to main content

Round Robin 3 case study

There have been a series of room acoustic simulation round robin tests organized by Physikalisch-Technische Bundesanstalt National Meterology Institute (PTB institute). The first international round robin was conducted in 1994, followed by the second in 1996-1998. The round robin 2 was organized with 16 participants from 9 different countries using 13 different computer programs. That time the object under examination was a concert hall in Jönköping/Sweden named ELMIA-hall. The results of this round robin had been much better than those of Round Robin 1 but the complex design of this hall generated many troubles. In 1999 the PTB decided in agreement with leading room acousticians in the technical committee of the European Acoustical Association to start another round robin 3" with a relatively simple structured room to have a good control over the scene and boundary materials. In this study over 20 simulation tools were used to simulate the studio room of PTB institute [1]. The room was measured and the material properties of all surfaces in the room were obtained in octave bands 125 Hz - 4k Hz. The room itself is a relatively small (volume of 400 m2m^2 ) but complex room with special diffusors. A strong standing wave was present in the room and a large part of the wall area could be covered with curtains. Results were given both with open curtains and closed curtains.

image desc

Fig. 1 Model of PTB studio in Sketchup

Setting up the simulation in Treble​

From the coordinates given by PTB [1]. A model was drawn up in Sketchup and imported to Treble. Following the guidelines of Treble of modeling hanging surfaces for the curtain surfaces were merged to the walls behind them, as that is preferred for the wave-based solver and the mesher. Two sources are shown as green dots while the receivers are indicated as blue dots.

image desc

Fig. 2 Model of PTB studio in Treble web application, with sources and receivers places, highlighting curtain material.

Materials, Scattering and Settings​

All the materials specified were created using the material fitter of Treble. The material fitter converts the absorption coefficients to the most likely surface impedance [2] and the specified scattering was applied to each surface. The wave-based solver is using the surface impedance (phase-included input data), in the GA solver the pressure based image source method uses reflection coefficient and the ray-tracer uses the absorption coefficient.

image desc

Fig. 3 Absorption coefficients for all materials used in the simulation as given by PTB

The given scattering coefficients in Fig 2 are frequency dependent, but currently a single-valued scattering coefficient is used in Treble as of June 2023. For each surface, we inputted the averaged scattering coefficient over the 500 and 1000 Hz bands. Then this single valued scattering coefficient per surface is converted to an s-shaped pre-defined curve. Please note that newer versions of Treble will accept frequency-dependent scattering coefficients from mid September 2023, after this study was made.

image desc

Fig. 4 Scattering coefficients for all materials used in the simulation as given by PTB

The transition frequency chosen for the simulation was 710 Hz, meaning the WB solver is used for the 1/1 octave bands up to and including 500 Hz, and the GA solver is used from the 1 kHz octave band and higher. In the geometrical acoustics, the image source of order 2 and using 5000 rays are used.

Results​

PTB provided measured objective parameters, EDT, T30, D50, C80, Ts, G, LF, LFC, and IACC per octave bands for 6 source-receiver pairs. Here are two comparisons of T30 at S2R3 and the T30 distribution over the source and receiver combinations at all measured bands. In Fig. 5, the T30 estimated by Treble agrees well with the measurement, note that the 250 Hz band is simulated using the wave based solver in Treble, where the exact solution to the wave equation is obtained. Particularly the variation due to source and receiver positions is better captured, while the other results show virtually no inter-position variation. The standard deviation of T30 for the measurement is 0.008, that for Treble is 0.015, and that for the other mean is 0.005, which is substantially smaller. This is one of the weaknesses of GA also described in [3]

T30​

image desc [Fig. 5, T30 results for source and receiver pairs at frequency band 250]

image desc [Fig. 6, T30 results for source and receiver pairs at frequency band 500]

image desc [Fig. 7, T30 results for source and receiver pairs at frequency band 1000]

image desc [Fig. 8, T30 results for source and receiver pairs at frequency band 2000]

Looking at a single receiver​

In Fig. 9, we can see that the GA simulation results (above 1 kHz) underestimate the reverberation time (which is also observed in the mean of the other software that is all GA). However T30 prediction gets more accurate using the wave-based solver, outperforming the other GA software that is limited by the model error, causes for the overall underestimation of T30 from the other simulations above 500 Hz could indicate that there is an error in the material absorption properties given at these frequencies and/or that the air absorption conditions, temperature and humidity have deviated from what is used in the simulations [4].

image desc [Fig. 9, T30 for all measured frequencies at Source 2, Receiver 3]

EDT​

Figures 10-13 show the EDT values across receiver, source combinations at all measured bands. The results for EDT values tell a similar story as the T30 results. At the 250 Hz frequency band Trebles wave-based solver captures the EDT of the measurements almost spot on. At 500 Hz the fit has diminished and at 1000 Hz and 2000 Hz and underestimation compared to the measured results is observed.

image desc [Fig. 10, EDT results for source and receiver pairs at frequency band 250]

image desc [Fig. 11, EDT results for source and receiver pairs at frequency band 500]

image desc [Fig. 12, EDT results for source and receiver pairs at frequency band 1000]

image desc [Fig. 13, EDT results for source and receiver pairs at frequency band 2000]

References​

[1][Homepage of round robin studies of PTB institute](https://www.ptb.de/cms/ptb/fachabteilungen/abt1/fb-16/ag-163/round-robin-in-room-acoustics.html)

[2] Mondet, B., Brunskog, J., Jeong, C. H., & Rindel, J. H. (2020). From absorption to impedance: Enhancing boundary conditions in room acoustic simulations. Applied Acoustics, 157, 106884.

[3] Marbjerg, G., Brunskog, J., & Jeong, C. H. (2018). The difficulties of simulating the acoustics of an empty rectangular room with an absorbing ceiling. Applied Acoustics, 141, 35-45.

[4] Aspöck, L., Brinkmann, F., Ackermann, D., Weinzierl, S., & Vorländer, M. (2019). A Round Robin on room acoustical simulation and auralization: Results of the simple scenes. Universitätsbibliothek der RWTH Aachen.