I came across an issue running thermal sims with a ring transducer. It would appear that if I change only Nx/Ny that my temperature (T1 and T2) changes. It may be an issue with how extractAmpPhase(...) is working/ how Q is being calculated. But I just took the default example and changed it to use a ring array.
Note: When I change Nx/Ny I also change dx/dy so the grid is the same size in [mm]. Also a pdf of the results and the code can be see at the link below.
I've updated the slides in the above link. It looks like the raw pressure field isn't the same for the last recorded field (@ t_end). Seeing as t_end is the same for all simulations and the waveform is the same I would expect the pressure field to be the same as well.
I thought this was an issue with createCWSignals(...), so I switched to a sine wave for slides 7&8 with no improvement.
Turns out the t_end for Nx/y=256 and Nx/y=512 aren't exactly the same in kgrid.t_array. So I updated the slides in the link to show the raw pressure field @ t=50usec (slide 9).
Another update, I've plotted the time trace of the pressure at the focus (slide 10). It would appear as Nx/y increases and dx/y decreases (in order for the physical space to stay the same size) the pressure at the center decreases. Is this expected?
So if we increase the fineness of the grid would the element size have to increase as well? If so I don't 100% understand why this is the case.
I've updated the slides again. I recalled from another post I made (http://www.k-wave.org/forum/topic/input-pressure-signal-vs-sensor-pressure-signal) that we have to use dirichlet BC to get a specific pressure distribution. This doesn't seem to 100% solve the issue of pressure (and therefore temperature) varying with grid size, but it is an improvement.
Is there a recommended way of doing this that avoids these varying pressure/temp issues? or is dirichlet the answer?
Hi Alexander,
Are you using an equal number of point sources in each case? A difference will arise in this case because of the scaling of the source. There are a few forum posts discussing this. I found this one but I recall there being others.
Dirichlet will give you a different problem as all of the sources will also act as scatterers.
Hope that helps,
Brad.
Hi Dr. Treeby,
I can confirm I used a fixed number of elements with:
`
ring = makeCartCircle(radius, 256);
[source.p_mask, order_index,reorder_index]= cart2grid(kgrid, ring);
`
^ In slides 10+11 you can see how the pressure waveform changes with Nx & Ny. Slide 11 shows the dirichlet BC and while the waveforms aren't exactly the same with changing Nx/y, they are closer than with the default BC.
You must log in to post.