Hi all,
I have a problem of NAN in my example_us_bmode_linear_transducer_modified.m that can be downloaded from here:
https://www.dropbox.com/s/v5bfybo4i3n36f5/example_us_bmode_linear_transducer_modified.m?dl=0
The problem is with sensor matrix at line 291 that is full of NAN,
but for me is not clear the cause.
thank you so much for your help.
Gg
Hi Gg,
Sorry for taking so long to get back to you. It seems you have discovered a small bug in k-Wave. When you set a transmit or receive apodization with only one active element in the transducer, getWin may return NaN (this depends on the exact distribution selected). This is what is happening in your code. For an apodization of 'Hanning', the apodization values used within the code are set to NaN. I have fixed this for the next release. In the meantime, I'd suggest setting the apodization to 'Rectangular' (the default), or setting the numerical apodization values manually. For a single active element, the apodization is not meaningful in any case.
Hope that helps,
Brad.
Hi,
I am running into a similar problem, but I am not using the transducer class with apodization. I am collecting a synthetic aperture dataset using an array of pressure sources and the same array of sensors. I.e., for each transmit event I am running kspaceFirstOrder2DG() for a single active source element and collecting on all sensor elements. It seems to work for some smaller sims, but for my larger sim my sensor traces have a few nonzero values followed by only NaN. I don't know if the above solution will help in my case? Hoping you can help me discern what is going wrong, thanks!
Best regards,
Omar
Hi Omar,
Generally NaNs suggest an instability, so I would start there by making your time step (or equivalently the CFL) smaller.
Brad.
Thank you for the lead, Brad!
There is an instability of some sort, but bumping down the time step/CFl did not help in my case (e.g., I decreased the CFL by factor of 100 to no avail). Interestingly, it seems that the low density values in my heterogeneous maps may be the cause. Keeping everything the same, but bumping up those lower density values results in signals that I expect (i.e., no NaN values in my dataset). Do you have an idea of why this may be the case? FYI - my map's density initially ranged from approx. 1-2500 kg/m^3, and sound speed ranged from approx. 1400-2500 kg/m^3.
Best regards,
Omar
Quick update: I realized that the minimum density in my original map was not in kg/m^3, after fixing, the sim runs well.
You must log in to post.