Sound Simulation_ Reverberation Time calculation_Inconsistant Results

Good morning all,

I am currently working on the calculation of the reverberation time of different closed spaces. I have a closed geometry (simplified model of the Hypogeum in Malta), and what I did was change the scale of it, so I had 3 versiones in: 1:1 / 1:2 / 1:5. What caught my attention was that the reverberation time in the 3 cases was quite similar even when the scales were so different and I don’t believe this makes sense at least from a theoretical point of view. The longer the distance for the sound waves to travel, the longer the reverberation time should be correct? But the results I am getting don’t show any significant changes. I did 4 tests in each geometry and I believe I might have something wrong in my definition, In my Ray tracing component I set it to make the calculation with " Trace Specified Number of Rays", not with the “Minimum Convergence”, since it would take much longer time for computing. I AM BASICALLY TRYING TO UNDERSTAND THE RELATION BETWEEN THE SCALE OF THE INTERIOR SPACE AND THE REVERBERATION TIME. If anyone could take a look I would truly appreciate it since I have no idea what I am doing wrong...Thank you all!

https://drive.google.com/drive/folders/17b_cNcm1JMljTkX8iiH39E_G8SN...

Replies to This Discussion

Permalink Reply by Michael Dubby on December 21, 2024 at 10:23am

Reverberation time is an important acoustic measurement that refers to the time it takes for sound to decay by 60 decibels after the sound source stops. This measurement is essential in spaces like concert halls, auditoriums, and even outdoor environments. It is often calculated using the Sabine or Eyring formula, which takes into account the volume of the space and the surface materials within it. For example, during the Upper Mustang Trek trekkers might notice different reverberation times in the unique natural acoustics of the mountain canyons and caves they pass through, influenced by the surrounding terrain and atmospheric conditions.

Permalink Reply by Arthur van der Harten on July 9, 2025 at 12:06pm

Hi,

For reference, this forum doesn't send us everything, but we did receive this one. I'll take a look this evening.

Arthur

Permalink Reply by Arthur van der Harten yesterday

Hi,

So why did you assign 10 percent transmission to all surfaces in grasshopper and then hide the wires to that input? You lose 40 percent of the energy in your simulation that way. It took me forever to find that, but I'm pretty sure that's your problem.

Arthur

Permalink Reply by Arthur van der Harten yesterday

Hi. Here is what I got for this - much closer to the paper:

{0;0}
125. 0.763982
250. 0.965731
500. 0.84799
1000. 0.857607
2000. 1.126713
4000. 1.442056
{0;1}
125. 0.771627
250. 0.979936
500. 0.846216
1000. 0.856514
2000. 1.137329
4000. 1.44285
{0;2}
125. 0.775357
250. 0.98732
500. 0.860795
1000. 0.871164
2000. 1.156493
4000. 1.456735
{0;3}
125. 0.783173
250. 1.002847
500. 0.863712
1000. 0.8734
2000. 1.163863
4000. 1.461097

It is very close to the results from the paper, except at 2000 and 4000 hertz. Now a perfect match is unlikely because of the effects of monte carlo algorithms. Still, the divergence at high frequencies suggests a difference in algorithms. Pachyderm follows non-diffuse effects very deep into the impulse response, so if this is happening in this room, and we know for certain that what is in the paper is correct, then I would argue that there is more scattering at those frequencies, and Pachyderm is more sensitive to the discrepancy in input data.

I hope this helps. Please watch your grasshopper file more carefully in the future. If the transmission on all surfaces was an attempt at a shortcut - don't ever do that again. It amounts to having the wrong absorption coefficients.

Thanks.

Arthur

RSS