Newbie Information Request on How to Use I-SIMPA

[Mark Kubis]

Hi,

I am learning how to use this powerful software. My acoustic room models will be smaller and more simple than the Elmia Hall tutorial and a very high level of accuracy is not required. 

I would welcome and appreciate some clarification of the following:

0) Is the 'SPPS' calculation method always more accurate than 'Classical Reverberation Theory' calculation method? Or are there some models where the Classical theory will give more accurate results?

1) When is it better to select 'Random' calculation method and when better to select 'Energetic' method?

2) When importing a 3D model, if you select 'Average Model Remesh' do you also need to select 'Repair model'?

Before running an 'SPPS' calculation:

3)  When do you check/uncheck the option 'Scene correction before meshing'?

4) When do you check/uncheck the option 'Surface receivers constraint'?

5) What is a good value to use for 'Surface receivers constraint' (m2)  when 4) is selected?

6) What is a good value to use for a Time Step?

7) What is a good value to use for 'Number of sound particles per source'?

Many thanks in advance,

Mark

[UMRAE Official]

Here some answers of your questions:

0) Classical theory is useful for room of homogeneous size with low and uniform wall absorption. In this case, this theory is relevant. Use SPPS instead.

1) I suggest you to use the 'random' methodn first, in order to test your room and the calculation parameters. When you consider your are ready to look for more precise results, change to 'Energetic'. Note that computational time will be longer.

2) Use 'Average Model Remesh' only if you are not able to produce and import a 'perfect' 3D scene. Sometime, I-Simpa can try to repair the model 3D; in this case check 'Repair model'. If you select  'Average Model Remesh' , you do not need to check 'Repair model', beacuse I-Simpa will create a new model

3) You can check this option. I-Simpa will try to correct some error in the 3D model before meshing.

4-5) You can check the option 'Surface receivers constraint' when you want more details on a surface receiver. Depending of the value 'Surface receivers constraint' (m2)', it will split the surface receiver in more surface elements.

6) If you want a 'relevant' echogram, you can use 0.005-0.002. If you only need room acoustics parameters, you can choose 0.01.

7) The computationnal time depends mainly of the Number of sound particles per source. Thus, in the SPPS code, each sound particle is followed, at each ‘Time step’ during its propagation inside the 3D domain, experiencing absorption, reflection, diffusion… and so on, during the ‘Simulation length’ and/or untill the particles disapears when its energy is to small.

Considering a small number of sound particles may not produced relevant results to proceed an interesting acouctic analazis of the 3D model. Thus, it is necessary to consider a large number of sound particles. Assigning the good value for large is difficult and depends mainly on the scene size and the absorbent nature of the scene (surface absorption and volume absorption).

Few recommandations to evaluate the computationnal time:

• the number of sound particles defined in SPPS (‘Number of sound particles per source’ parameter) is defined for each sound source. Considering N sound particles per source and Msound sources will generate N x M sound particles in the domain;
• one calculation is done for each frequency band: considering F frequency bands will be equivalent to a calculation with F x N sound particles for one source (or F x N x M for M sources). Note that, by default, the computation is paralleleized on the processor core, for each frequency band (a frequency band for a given core). So the increase of computional time with the nimber of frequency band is not linear;
• in the ‘Energetic’ mode, each sound particle is ‘alive’ during its propagation untill its energy is below a ratio of the initial energy (see ‘‘’Limit value of the particle extinction: ratio 10^n’ parameter). Increasing n will keep the sound particles longer in the domain, which increases the computational time;
• in the ‘Random’ mode, each physical phenomena (absorption, diffusion) is applied by considering probabilistic approaches. Using this mode, the computaional time is drastically decreased, but the quality of the results is aslo decrease. It is suggested to consider the ‘Random’ mode at the initial step of the study, and then to change to the ‘Energetic’ mode in order to obtain the final result.

Best