http://www.k-wave.org/forum/topic/strange-behavior-with-sourceu_mask-and-dirichlet-boundary-conditions Support for the k-Wave MATLAB toolbox en-US Tue, 12 May 2026 23:07:58 +0000 http://bbpress.org/?v=1.0.2 q http://www.k-wave.org/forum/search.php http://www.k-wave.org/forum/topic/strange-behavior-with-sourceu_mask-and-dirichlet-boundary-conditions#post-6624 Tue, 30 Oct 2018 20:26:10 +0000 Bradley Treeby 6624@http://www.k-wave.org/forum/ <p>Hi NikiKeyz,</p> <p>Thanks for posting your solution. Using separate masks was something we considered when first writing k-Wave, but in the end we thought the sources would always be used at the same time. You're actually the first person to ask for it. In principle, it's easy to implement, but every option has to be rolled out across the code base and maintained, etc. We'll add it to the development list for discussion.</p> <p>Brad. </p> http://www.k-wave.org/forum/topic/strange-behavior-with-sourceu_mask-and-dirichlet-boundary-conditions#post-6619 Mon, 29 Oct 2018 13:27:34 +0000 NikiKeyz 6619@http://www.k-wave.org/forum/ <p>I managed to modify K-Wave source code to use independent ux_mask and uy_mask. It seems to work as desired now.<br /> This way time step can be significantly increased opposed to use of "real" density/speed. And simulation runs more than 10 times faster on my configuration this way.<br /> Would be nice to have this feature in future K-Wave releases. Especially for 3D C++/CUDA. Thanks! </p> http://www.k-wave.org/forum/topic/strange-behavior-with-sourceu_mask-and-dirichlet-boundary-conditions#post-6603 Tue, 23 Oct 2018 18:30:46 +0000 NikiKeyz 6603@http://www.k-wave.org/forum/ <p>Well, ok these are not bugs, sorry.</p> <p>1) source.ux and source.uy must be defined for every time step:<br /> <code>source.ux(1:nnz(source.u_mask),1:kgrid.Nt)=zeros(kgrid.Nt);</code></p> <p>2) Since u_mask is one for both ux and uy we can't set only ux to zero and leave uy without boundary condition at some grid points. That's why bottom and right sides scatter as intended and left and top sides give artifacts. Is there any way to overcome this?</p> <p>Thanks! </p> http://www.k-wave.org/forum/topic/strange-behavior-with-sourceu_mask-and-dirichlet-boundary-conditions#post-6602 Tue, 23 Oct 2018 17:42:54 +0000 NikiKeyz 6602@http://www.k-wave.org/forum/ <p>Hi!<br /> I am trying to model acoustic scattering in air from rigid surface. I've tried couple of methods before with small luck. I tried to set almost 0 sound speed for scattering surface which was fine and fast but there was significant phase shift. Second approach was to use density, sound speed and alpha coeff of real material. Works nice, but this way time step have to be quite small to avoid "explosion". And while I need to make thousands of evaluations this scheme is too slow.</p> <p>Finally I discovered idea to set dirichlet conditions for velocity to 0. This scheme works fast and seems to give realistic enough scattering with correct phase.<br /> But after some tests I discovered 2 strange things:<br /> 1) After some simulation time passes dirichlet conditions go away and waves just propagate from that moment as if u_mask was zero matrix.<br /> 2) Some strange vibrations occur on the left side of scattering surface. Strange thing that this effect is not symmetrical as there is nothing like that on the right side of the sample.</p> <p>Here is my code:<br /> <pre><code>% create the computational grid Nx = 160; % number of grid points in the x (row) direction Ny = 160; % number of grid points in the y (column) direction dx = 0.025; % grid point spacing in the x direction [m] dy = 0.025; % grid point spacing in the y direction [m] kgrid = makeGrid(Nx, dx, Ny, dy); % define the properties of the propagation medium medium.sound_speed = 345*ones(Nx,Ny); % [m/s] Air medium.alpha_coeff = 1.1795*ones(Nx,Ny); % [dB/(MHz^y cm)] medium.alpha_power = 1.4484; %Air (?) medium.density = 1.1864; %kg/m3 Air PML_x=20; PML_y=20; % create the time array dt = 1.5e-5; total_time = 0.05; kgrid.t_array = 0:dt:total_time; % define source point(s) source.p_mask = zeros(Nx, Ny); source.p_mask(Nx/2,Ny/2) = 1; % define Ricker source fcent = 1500; t0 = 0.002; source_amp = 0.18e-2; source.p = source_amp*-sqrt(2)*pi*fcent*((sqrt(2)*pi*fcent*(kgrid.t_array-t0)).^2 - 1) ... .*exp(-0.5*(sqrt(2)*pi*fcent*(kgrid.t_array-t0)).^2); % Filter source_notfiltered = source.p; medium_filter.sound_speed=345; source.p = filterTimeSeries(kgrid, medium_filter, source.p); source.u_mask = zeros(Nx, Ny); source.u_mode=&#39;dirichlet&#39;; source.u_mask(PML_x+1:PML_x+8,PML_y+1:Ny-PML_y-1)=1; source.ux(1:nnz(source.u_mask))=0; source.uy(1:nnz(source.u_mask))=0; sensor.mask = [0;0]; sensor.record = {&#39;p_rms&#39;}; sensor_data = kspaceFirstOrder2D(kgrid, medium, source, sensor, &#39;DataCast&#39;, &#39;gpuArray-single&#39;, ... &#39;PlotScale&#39;, [-0.1, 0.1], &#39;PlotSim&#39;, true, &#39;DataRecast&#39;, true, &#39;LogScale&#39;, true, &#39;PMLSize&#39;,[PML_x,PML_y]);</code></pre>