k-Wave

1. laughingrice
   

   Member
   Joined: May '13
   Posts: 5

   I'm trying to simulate a linear array clinical ultrasonic transducer emitting a plane wave (64 central elements active on a 128 element transducer). Simulation is running in 2D due to time and memory constraints at the moment.

   My problem is that just activating the pressure sources along the transducer face in a homogenous medium creates circular waves either at each element, if I make independent elements, or at the end points, if I just make a continuous line source. This waves travel across the transducer face and show up in the output traces.

   Running the same measurement with an actual probe, those waves traveling across the probe face can barely be seen, if at all, as signal does not propagate well sideways inside the probe. I'm guessing that it's a combination of independent piezo elements which I think are air separated, and attached to a matching layer.

   There is a little discussion about the physical structure here
   http://www.ndk.com/en/sensor/ultrasonic/basic02.html
   and here
   http://www.echocardiographer.org/Echo%20Physics/BasicTransducers.html

   Anyone have an example of modeling a real probe in a simulation to get a more realistic wave field?
   How would I go about distributing the sources, and should I use pressure or velocity sources, if I do a piezo / air / matching layer construction in the domain?

   Thanks

   Posted 6 years ago #

2. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   Hi laughingrice,

   You could try using a velocity source instead of a pressure source. This will produce a dipole field, which will make the wave-field more directional. If the source is aligned with the grid, there will also be no response detected at the adjacent elements.

   Alternatively, you could model the source conditions using another model (e.g., FEM) that accounts for the actual source construction, record the acoustic field just in front of the source, and then input this into k-Wave.

   Accurately modelling the layered structure of the piezo, backing, and air in k-Wave will be pretty difficult due to the small scale structures and the large impedance mismatches.

   Brad.

   Posted 6 years ago #

3. jbishop
   

   Member
   Joined: Oct '18
   Posts: 5

   Simulation of heating in an air-backed transducer and its substrate was of interest to me, and I too wondered just how well k-Wave could do with it not having access to any FEM solver. Without yet knowing how well the result matches experimental data, I can say so far it is certainly plausible looking. It required a lower CFL number of < 0.1 before the air gap would stabilize. As a result of the low CFL number and fine resolution needed, compute times in 2D matlab were getting too long. I saw the other post on the c++/gpu release for 2D: that will be of immediate use to me.

   Posted 4 years ago #

4. bramth
   

   Member
   Joined: Mar '19
   Posts: 5

   Dear Bradley,
   When using such a dipole transducerin a velocity-source setting (by setting source.p_mask to source.u_mask and source.p to source.ux), why can we not scale the input the same way as with a pressure source? (E.g. the transducer.source_strength, when scaling the tone_burst input with this value (and the impedance).) If I do scale the input, I get 'black' wavefronts (when not using C++ or CUDA implementations). Kind regards, Bram

   Posted 4 years ago #

5. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   The particle velocity is typically several orders of magnitude smaller than the pressure. In 1D, they are related by the acoustic impedance (sound speed times density), so to get the same relative amplitude you will need to divide by the amplitude of the velocity source by the impedance.

   Posted 4 years ago #

6. bramth
   

   Member
   Joined: Mar '19
   Posts: 5

   That is true, thanks. Is it by the way a plotting artefact that the B-mode scan example using linear array, when plotting the simulation, shows black intensities and something that seems PML penetration in the beginning? The "black" wave progression continues afterwards. See for the result after a few time steps https://imgur.com/a/4ME4RBZ and https://imgur.com/a/gNVeMEs.

   Posted 4 years ago #

7. bramth
   

   Member
   Joined: Mar '19
   Posts: 5

   It seems setting the 'PlotScale' to either the max source intensity or to 'auto' solves the problem. Still, I wonder what happens in those low intensity regions.

   Posted 4 years ago #

8. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   Hi bramth,

   Some of the wave energy will propagate through the PML. However, with appropriate settings this should be -80 dB less than the incoming wave (see the user manual if you want more details). This means your simulation is only accurate to a few decimal places, but that should be enough for most practical purposes.

   Brd.

   Posted 4 years ago #

RSS feed for this topic

Reply

You must log in to post.