Less is More?

Is the more accurate model always better?

- Once more a "techy" post -

dBsea can be quite resource-intensive to run, especially if your model demands calculation of many sources and spatially detailed outputs.
Today I'll make a case for always running simple simulations until all parameters are set as wanted, or you are ready to run your model through the night. Even though dBSea is heavily optimised with respect to resource consumption, it is easy to make a model that demands huge amounts of calculations.
A comparison of calculation time versus accuracy of output seems to be the logical way forward.

Figure 1. A comparison of a coarse solve VS a detailed one, 5 minutes VS 3 hours. Top panes are the maximum levels projected to the surface, the middle shows exclusion zones and the bottom shows the 3D version with the calculation grid superimposed.

While the above solves are by no means identical, but sound levels are within circa 10 dB (see the spectrograms in the top panes), and exclusion zones are within a factor of 2. While this might seem like big differences, when you are designing you model, and want a bit off feedback on the changes you've made, a rapid update of the results is very desirable.

Another example is the below example (Figure 2), where two ray trace solves are very close to identical. Scenario "A" took over 30 min to compute, with thousands of outgoing rays from an approximated line source. Scenario "B" was very quick to solve with no calculation of attenuation, and only 20 outgoing rays. Also no frequencies over 1 kHz were included as attenuation means these frequencies do not propagate as far.

Figure 2. Comparison of a high accuracy solve VS a quick solve. dB-levels are only for comparison.

For quick evaluation and feedback during scenario design, simplifying your solves will radically improve this phase, making sure you only need to do the "big" solve once.

Thanks for reading, please don't hesitate to comment.