k-Wave

1. Anthony
   

   Member
   Joined: Apr '13
   Posts: 96

   Hi everyone,

   I am currently trying to write my own HDF5 input file for the C++ code and I only have a problem with ddx_k_shift_pos_r and its family (ddx_k_shift_neg_r, ddy_k_shift_pos, ddy_k_shift_neg etc...).

   How to compute it? I have only found this formula in the matlab code (in kspaceFirstOrder3D.m):
   ddy_k_shift_pos = ifftshift( 1i*kgrid.ky_vec );
   ddy_k_shift_neg = ifftshift( 1i*kgrid.ky_vec );

   More precisely, I have 3 questions:
   - it does not give the same result as when I read an input file created with kspaceFirstOrder_saveToDisk.m what makes me feel that my formula is not the right one, but I don't see anything else in the code...
   - what is the difference between ..._neg and ..._pos?
   - why is there a ..._r suffix along the x dimension? Is the formula different?

   Sorry for the question, but I have fought with this the whole afternoon.
   Regards,
   Anthony

   Posted 11 years ago #

2. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   Hi Anthony,

   The formulae you mention are for when the staggered grid scheme is not used. This is an undocumented option that is only used for testing. The lines of code you need are directly above these, where

   ddx_k_shift_pos = ifftshift( 1i*kgrid.kx_vec .* exp(1i*kgrid.kx_vec*kgrid.dx/2) );
   ddx_k_shift_neg = ifftshift( 1i*kgrid.kx_vec .* exp(-1i*kgrid.kx_vec*kgrid.dx/2) );
   ...

   The difference between _neg and _pos is the direction of the spatial shift term (negative or positive). The _r suffix is used to show that only the first half of the Fourier data in the x-dimension is stored (see the _saveToDisk code for more details). This is because the FFT routine used in the C++ code is the real-to-complex FFT. This doesn't calculate the redundant part of the output for the FFT over the first dimension.

   Hope that helps!

   Brad.

   Posted 11 years ago #

3. Anthony
   

   Member
   Joined: Apr '13
   Posts: 96

   Thanks for your explanation,

   it seems to work now!

   Anthony

   Posted 11 years ago #

4. Anthony
   

   Member
   Joined: Apr '13
   Posts: 96

   Hi everyone,

   Just a little precision that might save some time if someone needs to do the same : as ddx_k_shift_pos_r (and its family) are complex values, you need to be careful when saving it with matlab into a HDF5 file.

   The sequence must be stored as :

   [real(ddx_k_shift_pos_r(1), imag(ddx_k_shift_pos_r(1),... ,real(ddx_k_shift_pos_r(Nx_r), imag(ddx_k_shift_pos_r(Nx_r)].

   i.e. a vector of size 2*Nx_r.

   Anthony

   PS : another trick (not related to ddx_k_shift_pos_r), don't forget to scale the source terms as explained in "kspaceFirstOrder_scaleSourceTerms.m".

   Posted 11 years ago #

RSS feed for this topic

Reply

You must log in to post.