Simple Box Test

Replies to This Discussion

Permalink Reply by Arthur van der Harten on March 3, 2023 at 9:49am

Hello.

No worries. We are all here to learn, hopefully.

A fundamental property of the Sabine equation is that it assumes a diffuse field, so there will be many cases in which your Pachyderm simulation will differ with Sabine. The main reason for this is that Pachyderm will properly calculate strong, late, specular reflections that lead to things like flutter echoes. (Not all the tools out there will do that for you.) Flutter has the effect of severely lengthening reverberation, and is considered a defect in acoustic quality. If you raise the scattering to a higher number, you will find that Pachyderm's answer comes closer to the Sabine prediction.

I cracked open your model, and noticed that scattering was not set for any of the materials. This is why your impulse response came out so spiky. You must set scattering - very important. Even flat materials scatter a little bit. There are buttons in the scattering controls that allow you to set good pre-sets for gypsum board and glass, and help you set scattering for articulated surfaces).

Also, make sure that your calculation is set for a sufficiently long cut off time. If the Schroeder Integral (red line) is not linear for between 5 and 35 dB, then your calculation's simulated time frame was not long enough.

Hopefully this helps.

Arthur

Permalink Reply by rpict on March 8, 2023 at 10:52am

Hi Arthur

Thanks a lot for your explanations. Things are getting better with your suggestions.

But when I switch the calculation method to "Detailed Convergence", the calculation "never" (16h) ends. (See picture of the model in the attachement) Do the roof rafters have a significant impact on the reverb time anyhow? Or is it useless to model so high complexity?
What is the max. numbers of CPU's pachyderm can handle? I'm thinking of installing Rhino on a render blade with 56 CPU's.

-rpict

 

Attachments:

• Acoustic_Model.JPG, 290 KB

Permalink Reply by Arthur van der Harten on March 8, 2023 at 11:12am

Hi,
I'm glad it is working better for you.
Yes, the convergence algorithm is something I invented, and to be honest, I wasn't entirely certain where the threshold should be. Several years in, and having used it myself a number of times, I have some ideas to improve it, but for the most part, minimum conference l convergence is better than I thought it would be, and 'detailed' is as you say, not practical...

So, feel free to use minimum convergence. If you happen to have a model that you think minimum convergence isn't always working for, then take note of how many rays minimum tends to use, and click 'specify' instead, and enter a higher number of rays yourself. I would suggest doubling or tripling.

In a future version, I will hopefully have a better solution.

Arthur