Hi-
I would like to do a fairly large 3D simulation with a heterogeneous medium with 3D arrays representing c, rho, and alpha_power. This works fine with the kspaceFirstOrder3D function, but gives me garbage output for kspaceFirstOrder3DC. If I change medium.alpha_power to, say, 0.75 (per the example), it works just fine. medium.sound_speed and medium.density are heterogeneous for both. Is there something obvious that I'm doing wrong? I can share an example, if that's helpful.
Thanks,
Nick
Hi Nick,
The input for medium.alpha_coeff can be a matrix, but the input for medium.alpha_power must be a single value (for both the matlab and C++ codes).
Brad.
Hi Brad,
Thanks for the quick reply. I made a mistake in my previous post. My problem arises for a matrix medium.alpha_coeff. Admittedly, some of my medium.alpha_coeff get quite high, but the matlab code seems to handle it without difficulty while the C code produces some sort of numerical instability which spreads away from the voxels with high alpha.
I uploaded an example to my dropbox:
https://dl.dropboxusercontent.com/u/12851069/kwaveExample.zip has code and an example of my array inputs (6.5 MB)
https://dl.dropboxusercontent.com/u/12851069/exampleoutput.mat has my output from the script in the previous (126 MB)
Please let me know if you see the same problem!
Thanks,
Nick
Hi Nick,
The reason is that the optional input medium.alpha_mode = 'no_dispersion' is not supported by the C++ code. If you remove this, the MATLAB simulation will also be unstable. This is because the absorption values are very high, so the sound speed dispersion is also significant. You can alleviate this by using a smaller time step (CFL number).
Hope that helps,
Brad.
Thanks very much! That explains the difference.
Best,
Nick
You must log in to post.